JP5821692B2 - File sharing system, file write-back control program, and file write-back method - Google Patents

Description

  The present invention relates to a file sharing system, a file write-back control program, and a file write-back method.

  Parallel computer systems are used in applications such as HPC (high performance computing). In a parallel computer system, a shared file system that can access common data from a plurality of computing nodes is used. Today, demands for computing performance to be realized by a parallel computer system are increasing, and the number of computing nodes is increasing. In addition, the capacity of data handled with the calculation performance has increased, and the number of storages (storage devices) used in the file system and the number of local area networks have also increased. As the required capacity increases, the use of tiered storage has become common.

  In the tiered storage, a storage area provided by a solid state drive (SSD) having a small capacity and a high speed reading / writing and a storage area provided by a large capacity hard disk having a large capacity and a low speed reading / writing are used. Hereinafter, a storage area having a small capacity and capable of high-speed reading / writing is referred to as a cache layer, and a storage area having a large capacity and low-speed reading / writing is referred to as a base layer.

  When using a shared file system that supports tiered storage, specify the target file before starting the job, and read the specified file from the base layer to the cache layer before starting the job. After the processing requested by the job is performed on the temporarily stored file, the cache layer file updated by the execution of the job is written back from the cache layer to the base layer.

  In addition, in the use of HPC, the file server manages only the file metadata in the base layer in order to eliminate the file server overhead and improve the input / output performance. A method is used in which direct input / output is issued from each file server to individual storages constituting the base layer without using a file server for managing the file. Storage that constitutes the file server and the base tier is connected via a storage area network, and the storage area network allows direct access from each file server to the individual storage that constitutes the base tier, and is necessary. It also allows information transmission between file servers.

  As a file system taking such a method, there are Parallel Network File System (pNFS) and GSTORAGEFS (trademark). In order to use these file systems, it is necessary that the computing node can be directly connected to the storage used by the file server.

  However, since there is a limit to the number of local area networks that can be handled by one computing node and the number of hosts that can be connected to one storage node, data that exceeds the size that can be handled by the number of storages that can be connected to one computing node In order to handle, the calculation nodes are divided into a plurality of groups, and a file server is allocated to each group of calculation nodes.

An example of such a configuration is shown in FIG. In FIG. 8, P _{1 to} P _i are client machines constituting each computation node, Q _{1 to} Q _j and H are storage area networks, R _{1 to} R _j are local area networks, and S _{1 to} S _j are file servers. T ₁ through T _j is the file server _S 1 to S _j of each of the SSD and high speed rotation type to provide a cache layer such as a hard disk storage (hereinafter, the cache layer), W-based layer in the file server _S 1 to S _j Storage (hereinafter referred to as base layer) such as a large-capacity hard disk.

Each computing node P ₁ -P _i can be directly connected only to the cache layer connected to the same storage area network used by the file server mounted on itself. For example, the computation node P ₁ can directly access only the cache layer T ₁ connected to the storage area network Q ₁ used by the file server S ₁ mounted on the computation node P ₁ , and the computation node P 1 ₁ can not be directly accessed cache layer _T 2 through T _j.

File server S ₁ to compute node P ₁ is used, the file compute node P ₁ is to be read from or written before the compute nodes P ₁ uses its files, read from the base layer W in the cache layer T ₁ After the computation node P ₁ manipulates the file on the cache layer T ₁ , if the file on the cache layer T ₁ is changed, the changed file is stored in the cache layer T 1. ₁ is written back to the base layer W. Further, since the time of the file transfer between the base layer W and the cache layer T ₁ is not only written back files in the base layer, it is necessary to update the metadata for managing the file, moving file server that manages the target file, for example, in a situation where the meta data of the file is the file server S ₂ manages requests an update of the metadata to the file server S _2.

However, since different jobs are generally executed in the plurality of calculation nodes P _{1 to} P _i , the file is written back to the base layer W, and the input / output from the calculation node to the file server that manages the write back target. Requests may collide. For example, when the file server S ₁ is writing back the file on the cache layer T ₁ back to the base layer W in the situation as described above, that is, the file server S ₁ updates the metadata to the file server S ₂ . If an input / output request is given to the file server S ₂ from the computation node P ₃ or the node P ₄ when requesting the file, the input / output request and the metadata update process collide on the file server S _2. Such a situation will occur.

  In this way, when the write-back and the input / output request from the computing node collide, the load on the base layer used by the file server that has received the request or the file server that has received the request increases.

  When the load on the file server or the base layer increases, input / output with respect to the calculation node is delayed, which causes a disadvantage that the performance of the calculation node decreases.

  As a technique related to the problem of load acting on a computer or storage, a file management method disclosed in Patent Document 1 and a network storage virtualization method disclosed in Patent Document 2 are already known.

  However, Japanese Patent Laid-Open No. 2004-228561 is merely disclosed regarding hierarchy management, metadata, access frequency for each hierarchy, and data movement between hierarchies.

On the other hand, Patent Document 2 describes storage load evaluation, file handle configuration, and guarantee of file handle uniqueness, and file access frequency is high based on disk usage rate and network bandwidth usage rate. By identifying the file, that is, the file that causes the excessive load, and moving the file that causes the high load to the storage that is in a relatively low load state, the load of the entire storage is leveled and the computer or storage Attempts have been made to reduce the load acting on the system.
However, the configuration does not adjust the timing of file operations such as file reading and file writing back, but simply attempts to level the load by adjusting the storage location of the problematic file in the base layer. It's just a thing.
The invention described in Patent Document 2 is based on the premise that a file with a high access frequency can be freely moved to a storage with a low load at that time, between the base layer and the cache layer. It is not related to reading and writing back of files between storages with different uses.

JP 2006-302012 A (paragraphs 0014, 0016 to 0019, 0029) JP 2004-54721 A (paragraphs 0030, 0032, 0042 to 0044, 0052, 0054, 0056 to 0058)

  SUMMARY OF THE INVENTION An object of the present invention is to provide a file sharing system, a file write-back control program, and a file write-in system in which file write-back from a cache layer to a base layer via a file server and an input / output request from a client machine serving as a computation node do not collide It is to provide a return method.

The file sharing system of the present invention connects file servers having a plurality of client machines constituting a computer node so as to be able to transmit information, is managed by each of the file servers, and has a base layer for each of the file servers. A base layer providing storage to be provided and a cache layer providing storage for each file server that provides a cache layer to each of the file servers are connected to each of the file servers via a storage area network, and each of the clients A file server connected to the client machine reads a file requested by the machine from the base layer providing storage via the storage area network, and temporarily stores the file in the cache layer providing storage. In the file sharing system in which the processing on the cache layer providing storage is performed on the file on the cache layer providing storage and then the file on the cache layer providing storage is written back to the base layer providing storage. In particular, to achieve
For each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, sequentially updates and stores the information, and communicates with other file servers connected to the file server. Usage status management means for sequentially reading and storing each of the loads stored in each as a load for each file server;
Refer to the load of the file server stored in the usage status management means and the load of each file server, and manage it in the base layer providing storage managed by each file server according to the load size Transfer control means for setting the write-back priority in the order of increasing load of the file server to be
Write back execution means for writing back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the transfer control means. It has the structure characterized by the above.

The file write-back control program of the present invention connects file servers having a plurality of client machines constituting a computer node so as to be able to transmit information, is managed by each of the file servers, and is based on each of the file servers. A base layer providing storage for providing a layer, and a cache layer providing storage for each file server providing a cache layer to each of the file servers, connected to each of the file servers via a storage area network, A file server connected to the client machine reads a file requested by each client machine from the base layer providing storage via the storage area network, and temporarily stores the file in the cache layer providing storage. A file for a file sharing system in which a file on a cache layer providing storage is processed according to a request from a local machine and then a file on the cache layer providing storage is written back to the base layer providing storage. In order to achieve the same purpose as described above,
A microprocessor included in each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, and sequentially updates and stores it in the storage area of the memory of the file server, and other file servers connected to the file server Usage status management means for sequentially reading each of the loads stored in each of the other file servers by information transmission of the above, and sequentially updating and storing them in the other storage areas of the memory as loads for each file server,
Refers to the load of the file server stored in the storage area of the memory and the load of each file server stored in the other storage area of the memory, and is managed by each file server according to the size of the load Transfer control means for setting the write-back priority in the order of increasing load of the file server that manages the storage for the base layer provided, and
According to the write-back priority set by the transfer control means, it functions as a write-back execution means for writing back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server. It has the structure characterized by this.

The file write-back method of the present invention connects file servers having a plurality of client machines constituting a computer node so as to be able to transmit information, and is managed by each of the file servers, and has a base layer in each of the file servers. A base layer providing storage that provides a cache layer to each of the file servers, and a cache layer providing storage for each of the file servers that is connected to each of the file servers via a storage area network, A file server connected to the client machine reads a file requested by the client machine from the base layer providing storage via the storage area network, and temporarily stores the file in the cache layer providing storage. File writing for a file sharing system in which processing on the cache layer providing storage is performed on the file on the cache layer providing storage, and then the file on the cache layer providing storage is written back to the base layer providing storage. In order to achieve the same purpose as described above,
Each of the file servers refers to the load of the file server and the load of each file server, and manages the file server in the base layer providing storage managed by each file server according to the size of the load. Set the write-back priority in order of ascending load
Each of the file servers writes back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the file server. It has the structure characterized by having made it.

In the file sharing system, the file write-back control program, and the file write-back method of the present invention, each file server refers to the load of the file server and the load of each other file server, and uses each file server. Depending on the size of the load acting on each file server in response to a request from a client machine, the base layer provision storage managed by each file server is increased from the one with the smallest load on the file server. By setting the write-back priority in the order of 冪, each file server manages the files on the cache layer providing storage of the file server according to the write-back priority set by the file server. Since the data is written back to the storage provided by the base layer, the client Priority write-back processing of the file for the base layer provided for the storage managed by the file server high load at that time is acting in response to a request from Tomashin is set automatically lowered.
Therefore, file write-back is executed preferentially from the storage provided by the base layer managed by the file server with a relatively low load, and file management processing required accompanying file write-back In other words, the opportunity for processing related to updating metadata to be performed on a heavily loaded file server is reduced. As a result, input / output requests and metadata requests from client machines on each file server are reduced. Since the probability that the update process overlaps and decreases, the load on each file server is reduced and the performance of the entire file sharing system is improved.

It is the block diagram shown about the outline of the structure of the file sharing system of one Embodiment to which this invention was applied. It is a functional block diagram showing an outline of a configuration of a file server in the file sharing system of the embodiment. It is the block diagram which showed the outline of the structure of the file server in the file sharing system of the embodiment. FIG. 4A is a conceptual diagram showing the table means used by the file server in a simplified manner. FIG. 4A shows a self-data storage table and other data storage tables for storing the number of used files corresponding to each file server. FIG. 4B is a diagram showing a logical configuration, and FIG. 4B is a diagram showing a logical configuration of a priority storage table for storing the priority of file write-back. 6 is a flowchart showing an outline of processing executed by a microprocessor of a file server when the file server receives a file handle open request from a client machine. It is the sequence diagram which showed simply about an example of the communication performed between a client machine, its file server, and another file server. 6 is a flowchart showing an outline of processing executed by a microprocessor of a file server when the file server receives a file handle close request from a client machine. It is the block diagram shown about the general structural example of the file sharing system which divided | segmented the calculation node into several groups and allocated the file server to the calculation node of each group.

  Next, embodiments for carrying out the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing an outline of a configuration of a file sharing system 1 according to an embodiment in which the present invention is applied to a parallel computer system.

File sharing system 1 of this embodiment, a client machine _E 11 _{~EE nk} constituting the computer node, and the file server _A 1 to A _n to provide a shared file system on the client machine _E 11 _{~EE nk,} the file server A ₁ to a _n each base layer or base layer provided for storage to provide a region for storing various files (storage device) C ₁ -C _n and the cache layer or files on each file server a ₁ to a _n each file server a ₁ to a _n for each of the cache layer provides for storage to provide a region for temporarily storing and a (storage device) B ₁ ~B _n.

Of these, the client machine _E 11 _{to E 1j} is connected to the file server _{A 1} via a local area network _{D 1,} also the client machine _E n1 _{to E nk} in the file server _{A n} via the local area network _{D n} It is connected.

The file server A ₁ to A _n are mutually signaling connected via a local area network F.

File server _A 1 to A _n each storage _C 1 -C _n for the base layer provided to provide a base layer of, respectively, are individually managed by the file server _A 1 to A _n. That is, assuming that x = 1 to n, the file server A _x manages the metadata of the file stored in the base layer providing storage C _x (where x is an integer).

File server _A 1 to A _n and the base layer provided for the storage _C 1 -C _n is connected via a storage area network G, if the file handle even apparent, each of the file server _A 1 to A _n, base layer even can read a file from one for the storage C ₁ -C _n provides, also, can be written back file to any of the base layer provided for the storage C ₁ -C _n.

File server _A 1 to A _n each file server _A 1 through each file server _A 1 to A _n for each of the cache layer provides for storage _B 1 .about.B _n be a storage area network G to provide a cache layer to A _n of it is connected to the, only the file server a ₁ to a _n, each storage B ₁ for providing cache layer subscript corresponding .about.B _n can access direct. That is, assuming that x = 1 to n, the file server A _x uses the cache layer providing storage B _x .

In this file sharing system 1, a file server connected to a client machine reads a file requested by each client machine E _{11 to} E _1j from any base layer providing storage via a storage area network G, Temporarily store the file in the cache layer providing storage corresponding to the file server, that is, the file server that received the request from the client machine. This is a file sharing system in which a file server writes back a file on a cache layer providing storage to a base layer providing storage via a storage area network G.

Thus, for example, the file server A ₁ client machine E ₁₁ is the base layer provided for storage under conditions which stores metadata of the file of the base layer provided for the storage C ₁ manages the base layer provided for the storage C ₁ in the case where the request file in the C ₁ to the file server a ₁ acquires the file handle of the file server a ₁ immediately its files connected to the client machine E ₁₁ via the local area network D _1, Based on this file handle, the file server A ₁ reads the corresponding file from the base layer providing storage C ₁ via the storage area network G, temporarily stores the file in the cache layer providing storage B ₁ , and the client processing in response to a request from the machine E ₁₁ Sense from subjected to the cache layer provides for storage B ₁ on the file, the file server A ₁ is written to the base layer provides for storage C ₁ the file cache layer provides for storage B ₁ via the storage area network G Will return.
Here, the file server A ₁ can immediately obtain the file handle of the corresponding file because the file server A ₁ itself manages the metadata of the file in the base layer providing storage C ₁ .
In this case, when the file is written back, the file server A ₁ updates the metadata of the corresponding file by its own internal processing.

Further, for example, the file server A _n client machine E ₁₁ is the base layer provided for storage under conditions which stores metadata of the file of the base layer provided for the storage C _n to manage the storage for providing base layer C _n in the case of requesting the files in C _n in the file server a ₁ is that the file server a ₁ which is connected to the client machine E ₁₁ via the local area network D ₁ immediately acquires the file handle of the file can not be, first, the file server a ₁ queries the file handle of the file to the file server a _n via a local area network F, acquires the file handle of the file the file server a ₁ from file server a _n And then this file handle The file server A ₁ directly accesses the base layer providing storage C _n via the storage area network G, and the file server A ₁ reads the corresponding file from the base layer providing storage C _n, and the file temporarily stored in the cache layer provides for storage B _1, request processing in accordance with the from is subjected to a cache layer provides for storage B ₁ on the file from the client machine E _11, the file server a _1, the cache layer The file on the providing storage B ₁ is written back to the base layer providing storage C _n via the storage area network G.
Here, the reason why the file server A ₁ cannot immediately obtain the file handle of the file is that the file server A ₁ does not directly manage the metadata of the file in the base layer providing storage C _n .
In this case, when the write back of the file, the file server A ₁ is, with respect to the file server A _n for managing the storage C _n for providing the base layer, and request the update of the metadata of the file through a local area network F , so that the file server a _n updates the metadata of the file by its internal processing.

The cache layer providing storages B _{1 to} B _n are configured by a storage such as an SSD (Solid State Drive) having a small capacity and high speed reading and writing, and a high-speed rotating hard disk, and the base layer providing storages C _{1 to} C _n are It is composed of a large-capacity hard disk or the like that has a large capacity and relatively low reading / writing speed.
Cache layer provides for storage _B 1 .about.B _n is cache layer provides for storage for each file server _A 1 to A _n. As already described, for example, the cache layer providing storage B ₁ is a cache layer providing storage dedicated to the file server A ₁ and cannot be directly accessed by the file server _An or the like.
On the other hand, the file server _A 1 to A _n respectively connected to the base layer provided for the storage _C 1 -C _n via a storage area network G is the base layer for both the file server _A 1 to A _n it is possible to provide, file server a ₁ to a _n is direct access to the storage C ₁ -C _n for both the base layer provided are possible.

Figure 2 is a functional block diagram showing an outline of the configuration of the file server _A 1 to A _n one file server _{A 1} representing a file server _A 1 to A _n for example, the file server _A 1 to A _n All the configurations are the same.
In FIG. 2, only one client machine E _1i representing these among client machines E _{11 to} E _1j connected to the file server A ₁ via the local area network D ₁ is shown, and the storage area network Only one base layer providing storage C _m representing these among the base layer providing storages C _{1 to} C _{n that} can be directly accessed by the file server A ₁ via G, and the local area network Description of D ₁ , local area network F, and storage area network G is omitted (where 1 <i <j, 1 <m <n, m is an integer).

The file server A ₁ includes a shared service providing unit U ₁ and a load management unit V ₁ .

Among these, the shared service providing unit U ₁ is for providing a shared file system to the client machines E _{11 to} E _1j , and provides, for example, an NFS daemon or a file sharing service for processing a request from the client. A CIFS server program (Common Internet File System) or the like is generally known.
More specifically, the main part of the shared service providing unit U ₁ in this embodiment is composed of a file sharing unit U ₁₁ and an input / output unit U ₁₂ .
File sharing unit _{U 11} processes the protocol of the shared file system to provide at the request from the client machine _E 11 _{to E 1j} client machine _E 11 _{to E 1j,} provides cache layer provides for storage _{B 1} and base layer requests the output unit U ₁₂ so as to input and output process directed to the use storage C ₁ -C _n, input output U ₁₂ is based on the result of the protocol processing by the file sharing unit U ₁₁ having received the I / O requests are issued to the cache layer providing storage B ₁ and the base layer providing storages C _{1 to} C _n to read / write data, and the result is returned to the file sharing means U ₁₁ , and further the file sharing means U ₁₁ Transfers this result to the requesting client machines E _{11 to} E _1j .
Further, the input / output unit U ₁₂ reads a file from the base layer providing storages C _{1 to} C _n to the cache layer providing storage B ₁ in response to a request from the file sharing unit U ₁₁ , or cache layer providing storage B _{1 has a} function of writing back a file to the base layer providing storages C _{1 to} C _n .
Structure and function of此those associated with the file sharing unit U ₁₁ and the input-output unit U ₁₂ is already known.

On the other hand, the load management unit V ₁ was, file server A ₁ and cache layer of the file server A ₁ grasps the load of the file server A further connected via a local area network F in _one file server A ₂ to A _n This is for controlling the transfer of files between the providing storage B ₁ and the base layer providing storages C _{1 to} C _n .
Load management unit V ₁ of the this embodiment is equivalent to the number of files that the file server A ₁ is open in response to a request from the client machine E ₁₁ to E _1j provided in the file server A ₁ to the load on the file server A ₁ with sequentially updated and stored as a file using the number of CA _1, other connected to the file server a ₁ via a local area network F by communication with the file server a ₂ to a _n of the other file server a ₂ to a _n corresponding file using the number _CA 2 to CA _n of each file server _a 2 to a _n to the load of the file server _a 2 to a _n each file using the number _CA 2 to CA _n stored in each successive loading comprising a usage management means V ₁₁ for sequentially updated and stored as the value. The load management unit _{V 1} was a file using the number CA ₁ of the file server _{A 1} stored in the use condition management unit _{V 11} files using number _CA 2 to CA _n for each file server _A 2 to A _n reference, in accordance with the magnitude or load of large and small files used number, the file server a ₁ to a _n which manages it in the file server a ₁ to a _n for storage provides the base layer are managed by the C ₁ -C _n It comprises transfer control means V ₁₂ from those consisting of the load small sets the write-back priority order of the large-made ones Noboribeki.
These usage management means V ₁₁ and the transfer control unit V ₁₂ is a unique component to此embodiment.
The usage status management means V ₁₁ of this embodiment particularly monitors requests input from the client machines E _{11 to} E _1j to the file sharing means U ₁₁ of the shared service providing unit U ₁ , and client machines E _{11 to} E _1j. from one to 1 increments the value of the file using the number CA ₁ of the file server _{a 1} when the file sharing unit _{U 11} open request for the file handle is entered, a file sharing unit _{U 11} from the client machine _E 11 _{to E 1j} When a file handle close request is input, the server has a function of updating its own server file usage number that decrements the value of the file usage number CA ₁ of the file server A ₁ by one.
The reason why the load of a file server with a small number of file usage is small and the load of a file server with a large number of file usage is large is in a typical data input / output pattern of a program used in HPC (high performance computing). In general, processes such as opening a file handle and reading a large file together and closing the file, or opening a file handle and writing a large file together and then closing the file are performed. This is because the I / O load usually increases while the file handle is open.
Incidentally, CIFS (Common Internet File System) or the like is generally known as an example of a shared file system that can easily detect an open / close file handle from the file sharing unit U _11. Even if a shared file system that does not notify the state such open / close, configured to implement the notification of the open / close the file sharing unit U ₁₁ to use state management unit V ₁₁ by expansion or change of protocol It is technically easy.
Other Also, the use state management unit V ₁₁ is the absent files to be open request or close the target base layer provides for storage C ₁ managed by the file server A _1, is connected to the file server A ₁ file server a ₂ through the information transmission and to a _n of the other file server a ₂ either to a _n, for example, need to request the closing of the file server a _y filehandle or file that manages the appropriate file when produced, and transfers the current value CA ₁ of the number of files using the file server a ₁ along with the request for closing the request and the file of the file handle to another file server a _y, from another file server a _y Applicable file handle and file close completion notification and the file on the file server Receives the current value CA _y number use, by rewriting the value CA _y file using the number of the current value CA _y the file server _{A y} received, received with the current value CA _y, the file server _{A y} file Another server file usage number update function for updating and storing the value of the usage number CA _y (where 1 <y <n, y is an integer)).
Usage management means V ₁₁ further upon receiving a request for closing the file handle or file by communication with other file server A ₂ to A _n, the file server A together with the appropriate file handles and closing completion notification having a file number of used transfer function of transferring _one current value CA ₁ file number of used file handles and file close request will originate in the file server a ₂ to a _n of.

It has already been described that the input / output means U ₁₂ has a function of writing back a file from the cache layer providing storage B ₁ to the base layer providing storages C _{1 to} C _n in response to a request from the file sharing means U ₁₁ . are as, in this embodiment, is managed files in the cache layer provides for storage B ₁ of the file server a ₁ by each file server a ₁ to a _n in accordance with write-back priority set by the transfer control unit V ₁₂ As the write back execution means for writing back to the base layer providing storages C _{1 to} C _n , a known input / output means U ₁₂ is used.

Figure 3 is a block diagram showing an outline of the configuration of the file server _A 1 to A _n one file server _{A 1} representing a file server _A 1 to A _n as an example.

The file server A ₁ is a normal computer composed of a personal computer or a workstation, and stores a microprocessor 2 for arithmetic processing, a ROM 3 storing a control program for the microprocessor 2, and various parameters. Non-volatile memory 4, RAM 5 used for temporary storage of various data, built-in hard disk 6, input / output circuit 10 for connecting monitor 7, keyboard 8 and mouse 9, and client an interface 11 for connection to a local area network _{D 1} having a machine _E 11 _{to E 1j,} an interface 12 for connection to a local area network F with a file server _a 2 to a _n, the base layer provides Storage C ₁ It comprises an interface 13 for connection to the storage area network G with a -C _n.

In addition to the server program for processing the requests from the client machines E _{11 to} E _1j to provide the file sharing service, the hard disk 6 includes the microprocessor 2 of the file server A ₁ as usage status management means V ₁₁ , usage status. local server file usage number updating function implementing means in the management unit V _11, other server files using number updating function implementing means in the usage management unit V _11, the file number of used transfer function implementing means in the usage management unit V _11, the transfer control means V ₁₂ stores a file write-back control program for functioning as a write-back execution means.

The hard disk 6 also stores metadata and the like for managing the files stored in the base layer providing storage C ₁ managed by the file server A ₁ .

A part of the RAM5 storage area which is one of the memory of the file server A ₁ is utilized as a self-data storage table 14 for storing the current value CA ₁ of the number of files using the file server A _1, also another part of the storage area of the RAM5 stores in association with other file server _a 2 to a file number of used each stored in _{n CA} 2 to CA _n to each of the file server _a 2 to a _n Therefore, it is also used as the other data storage table 15 for this purpose.

FIG. 4A shows a simplified logical configuration of the data tables 14 and 15. Own data stored in the table 14 and other data storage table 15 values for each file server _A 1 to A _n file using the number _CA 1 to know at that time to CA _n of each file server _A 1 to A _n The name, that is, the identification information is stored in one-to-one correspondence.
FIG. 4B shows an example of the logical configuration of the priority storage table 16. The priority storage table 16 stores the metadata write-back priority for the set of file server and base layer providing storage.

Figure 5 is a flowchart showing the outline of the processing by the microprocessor 2 executes when the file server A1 receives an open request for the file handle from the client machine E ₁₁ to E _1j.

Next, referring to FIG. 5, using state management unit V _11, its own server file usage number updating function implementing means in the usage management unit V _11, other server files using number updating function implementing means in the usage management unit V ₁₁ , specific processing operations of the microprocessor 2 that functions as a file using the number of transfer function implementing means in the usage management unit V ₁₁ will be described.
The contents of the local data storage table 14 and other data storage table 15 is assumed to be initialized at power up of the file server A _1.

First, one of the application programs of the client machines E _{11 to} E _1j is activated (step s1), and the value of the file handle required to open the target file is set as the file sharing means U of the file server A _{1. 11} (step s2).
Here, as an example, it is assumed that the client machine E _1i obtains the value of the file handle from the file server A ₁ .

Microprocessor 2 that functions as a local server file usage number updating function implementing means in the usage management unit V ₁₁ detects the input of the open request from the client machine E _1i to the file sharing unit U _11, the file server A ₁ memory The current value CA ₁ of the file usage number stored in the own data storage table 14 of the RAM 5 is read, incremented by one, and this value is re-stored in the own data storage table 14. The value of the current value CA ₁ is updated and stored (step s3).

The file sharing means U ₁₁ of the file server A ₁ searches the path name of the file requested by the client machine E _1i (step s4), the requested file is on the file server A ₁ own local file system, That is, it is determined whether or not the cache layer providing storage B ₁ or the base layer providing storage C ₁ exists (step s5).

If the step s5 determination result revealed that there is a purpose of the file on the local file system of the file server A ₁ itself becomes true, file sharing means U ₁₁ is the file handle of the requested file _Is immediately transmitted to the client machine E _1i (step s9).
The file operation after the client machine E _1i acquires the file handle is the same as the conventional one. For example, the file server A ₁ reads the target file from the base layer providing storage C ₁ via the storage area network G. , processing the temporarily stored files in the cache layer provides for storage B _1, or update rewrites the file by performing processing according to a request from the client machine E ₁₁ in a file cache layer provides for storage B ₁ Will be done.

On the other hand, if the decision result in the step s5, it became clear that there is no target file on becomes false file server A ₁ own local file system, file sharing means U ₁₁ is the path name of the file depending via a local area network F any of the other file server a ₂ to a _n, for example, but would require the file handle to the file server (a _y) that manages a corresponding file, this time, the other server files Since the microprocessor 2 functioning as the usage number update function realization means reads the current value CA ₁ of the file usage number of the file server A ₁ from the own data storage table 14 and attaches it to the file handle request, the file server A _y indicates the number of files used by file server A ₁ together with the file handle request. Current value CA ₁ is to be sent in (step s6).

Then, sent with the request file handle file server A _y file sharing means receiving the current value CA ₁ of the number of files using the file server A _1, the file handle is requested from the file server A ₁ to the file server A ₁ Although will be, this time, the microprocessor of the file server a file used number of transfer function in the _y realizing unit clogging file server a _y is the current value CA _y file using the number of the file server a _y, the file server a _y The file server A ₁ uses the file handle of the file server A _y together with the value of the file handle corresponding to the request. The value of the current value CA _y of the number is to be transmitted. The

Then, the microprocessor 2 that functions as another server files using number updating function implementing means in the usage management unit V ₁₁ of the file server A ₁ has the file using the number of the file server A _y together with the file handle sent from the file server A _y Current value CA _y (step s7) and overwriting the current value CA _y received this time corresponding to the name of the file server A _y in the other data storage table 15 of the RAM 5 of the file server A ₁ file server _a 2 to a _n updates and stores the file using the number CA _y which is one file server _{a y} of the (step s8).

Then, similarly to the above, the file sharing section U ₁₁ transmits a file handle of the file requested by the file handle That client machine E _1i sent from the file server A _y the client machine E _1i (step s9).
In this case, file sharing unit U ₁₁ is subjected to a process to ensure the uniqueness of a file of interest to the file handle to return to the client machine E _1i. As a specific method, there is known a method of attaching an identifier for identifying the file server _Ay to the file handle, for example, a host name, an IP address, a MAC address, etc. of the file server _Ay . Of course, it is also possible to use an identifier to which the file is directly identify the base layer provided for the storage C _y stored in place of the identifier for specifying the file server A _y.

Client machine E _1i is the same as defined above with respect to file operations after acquiring a file handle, for example, the file server A ₁ is read a file from the storage C _y for providing base layer object through the storage area network G , processing the temporarily stored files in the cache layer provides for storage B _1, or update rewrites the file by performing processing according to a request from the client machine E ₁₁ in a file cache layer provides for storage B ₁ Will be done.

Here, an example of communication performed between the client machine E _1i , the file server A _1, and the file server A _y will be briefly described with reference to FIG.

Figure 6 is the client machine E _1i and the file server A ₁ and the file server if the client machine E _1i has issued a request to open the file of the file path name to the server _{A 1 "/ export / a /} b / c" As an example of communication performed between A _y , the file with the path name “/ export / a / b / c” does not exist on the local file system of the file server A _{1 and} is stored on the base layer providing storage _Cy . It is the sequence diagram shown about the case where it exists.

First, when the client machine E _1i opens a file, a file handle open request “/ export / a / b / c” is transmitted to the file server A ₁ (sequence T1 / FIG. 6, steps s1 to steps in FIG. 5). corresponding to s2).

When the file server A ₁ receives an open request for the file handle, it is incremented by 1 the current value the value of CA ₁ in number of files using the file server A ₁ itself (corresponding to step s3 of sequence T2 / 5 of FIG. 6) , the sequence of steps T3 / 5 of the file server a _y with respect to "/ export / a / b / c" to be open request sends the current value CA ₁ of the number of files using the file server a ₁ (FIG. 6 corresponding to s6).

Then, the file server A _y file using the number of transfer function implementing means a file handle FH and the file server A current value CA _y file using the number of the _y file "/ export / a / b / c" to the file server A ₁ It responds (sequence T4 in FIG. 6).

Then, the file server A ₁ returns the file handle FH received from the file server A _y to the client machine E _1i (sequence T5 in FIG. 6).

Next, the file server A1 client machine E ₁₁ to E _1j with reference to FIG. 7 will be briefly described process of the microprocessor 2 executes upon receiving a close request for the file handle from.

First, any application program of the client machines E _{11 to} E _1j closes the file (step t ₁ ), specifies a file handle to the file sharing means U ₁₁ of the file server A ₁ , and performs processing related to the closing of the file. Execution is requested (step t2).

Microprocessor 2 that functions as a local server file usage number updating function implementing means in the usage management unit V ₁₁ detects an input of a close request from the client machine E _1i to the file sharing unit U _11, the file server A ₁ memory By reading the value of the current value CA ₁ of the file usage number stored in the own data storage table 14 of the RAM 5 and decrementing it by 1, and re-storing this value in the own data storage table 14, the number of file usages updates and stores the current value the value of CA ₁ in (step t3).

The file sharing means U ₁₁ of the file server A ₁ searches the file handle of the file requested by the client machine E _1i (step t4), corresponding to the requested file handles files of the file server A ₁ itself It is determined whether or not the file exists on the local file system (step t5).

If the decision result in the step t5 has revealed that there is a purpose of the file on the local file system of the file server A ₁ itself becomes true, file sharing means U ₁₁ is the file handle of the requested file _Is immediately notified to the client machine E _1i of the completion of the closing process (step t9).
The file to be closed was temporarily stored from the base layer providing storages C _{1 to} C _n to the cache layer providing storage B ₁ , and the file was subjected to processing such as rewriting or updating. In this case, it is necessary to write back this file to the base layer providing storages C _{1 to} C _n of the reading source. This write back processing is performed by a transfer request from the transfer control means V ₁₂ to the output means U ₁₂ . It will be executed after waiting for output.

On the other hand, when it becomes clear that there is no target file is the determination result becomes false on the file server A ₁ own local file system in step t5, file sharing means U ₁₁ is managing a file object to other file servers, for example, but would require the closing of the file to the file server a _y, this time, the microprocessor 2 files of the file server a ₁ which serves as another server files using number updating function implementing unit Since the current value CA _{1 of the} number is read from the own data storage table 14 and attached to the close request, the file server A _y sends the current value CA _{1 of the} number of used files of the file server A ₁ together with the file close request. Is transmitted (step t6).

Then, file sharing means of the file server A _y having received the current value CA ₁ of the number of files using the file server A ₁ with close request of the file, and opens the file handle of the file requested by the file server A ₁ Closed While the completion of the process will be sent to a file server a _1, this time, the microprocessor of the file server a _y file using the number of transfer functions in the implementation means, that the file server a _y is, the file server a _y file using the number of Since the current value CA _y is read from the own data storage table generated in the RAM that is the memory of the file server A _y and attached to the completion notification of the closing process, the file server A ₁ has the closing process. completion notification file server a _y file using the number of the current along with the So that the CA _y is transmitted.

Then, the microprocessor 2 that functions as another server files using number updating function implementing means in the usage management unit V ₁₁ of the file server A _1, together with notification of completion of the closing process which is transmitted from the file server A _y of the file server A _y By receiving the current value CA _{y of the} number of used files (step t7) and overwriting the current value CA _y received this time corresponding to the name of the file server A _y in the other data storage table 15 of the RAM 5 of the file server A ₁ , updates and stores the file server _{a y} file using the number CA _y is one of the other file server _a 2 to a _n (step t8).

Then, finally, file sharing means U ₁₁ transmits a completion notification of the closing process is transmitted from the file server A _y the client machine E _1i (step t9).
Similarly to the above, when the file to be closed is temporarily stored in the layer providing storage B ₁ and it is necessary to write back the file, the write back processing is performed from the transfer control means V _12. waiting for output of the transfer request to the output unit U ₁₂ is to be executed.

In this embodiment, as shown in FIG. 5 and FIG. 7, as an example, when the file is opened and when the file is closed, the file server A ₁ transmits the file usage number CA ₁ to the other file server A _y and other files. Although described example of retrieving server a _y file using the number of CA _y, it is also possible to perform the exchange of files using the number CA _1, CA _y when the file is opened or only when the file is closed only.
Furthermore, not only the timing at the time of opening and closing of the file, but also the configuration in which the usage status management means V ₁₁ exchanges the file usage numbers CA ₁ and CA _y with other file servers A _y at predetermined intervals. Good. The file number of used CA _1, CA _y by regularly changed every predetermined period, the file server A ₁ is the local file system in the file open / close and other file server A ₂ to A _n managed to manage open click / Rose file handle files or if happened uneven usage even often access to the local file system of the file server a ₁ is properly updated.
Also, the file usage numbers CA ₁ and CA _y are exchanged at another request timing other than the open / close timing transmitted to the file server A ₁ as an extension of the system call on the client machines E _{11 to} E _1j You may send it. Even when the file usage numbers CA ₁ and CA _y are exchanged at timings other than opening / closing, even if the actual input / output load in the file server A ₁ is uneven from opening to closing, it is used. The situation is reflected.

Above, the file server A ₁ and the other file server A ₂ to A _n of pick up the relationship between one file server A _y representative has been described the number of files used exchange, file server A ₁ and all the other File usage also regarding the relationship between the file servers A _{2 to An} and the relationship between the file server A _y and other file servers A _z (where 1 <y <n, 1 ≦ z ≦ n, y ≠ z) All processing related to the exchange of numbers is the same as this.

Next will be described the processing operation the microprocessor 2 which functions as a transfer control unit V ₁₂ and the input-output unit U _12.

The microprocessor 2 that functions as the transfer control means V ₁₂ in the file server A ₁ updates the file usage number CA ₁ of the file server A ₁ and the file server each time the contents of the own data storage table 14 or the other data storage table 15 are updated. a ₂ refers all to a _n file using the number of _CA 2 to CA _n of, each time, to sort the names of each file server _a 1 to a _n according to the file number used, made large from those made small files used number Set the write-back priority to the base layer providing storages C _{1 to} C _n in the order of ascending from the one with the load of the file server acting on the request from the client machine to the largest one, its priority is updated in priority storage table 16 of the RAM5 a memory of the file server a ₁ stores That.
Accordingly, if, be such that the value of the file server _A 1 to A ₄ of the file using the number _CA 1 to CA ₄ is shown in its own data storage table 14 and other data storage table 15 shown in FIGS. 4 (a) Assuming that the number of files used is 4 (number of files used by file server A ₄ ) <number of files used 5 (number of files used by file server A ₂ ) <number of files used 8 (number of files used by file server A ₃ ) <file usage From the relationship of the number 10 (number of files used by the file server A ₁ ), the setting of the write-back priority in the priority storage table 16 is the file with the highest write-back priority as shown in FIG. level managing server a ₄ base layer provided for the storage C ₄ managed _by the write back priority the second highest is the file server a ₂ That the base layer provided for the storage C _2, writeback priority base layer provided for the storage C ₃ to the third highest is managed by the file server A _3, the write back the lowest priority is the file server A ₁ The base layer providing storage C ₁ is managed.

  In other words, the base layer provided storage managed by a file server with a low load expressed by the number of files used has a higher write-back priority, and the base layer is managed by a file server with a high load expressed by the number of files used. This means that the write-back priority is set so that the write-back priority becomes lower as the providing storage becomes lower.

Then, the microprocessor 2 which functions as a transfer control means V ₁₂ is the cache layer provides for storage B ₁ of the base layer provided for the storage C ₁ -C _n file server A ₁ from for updates and rewriting of the contents of the file If there are any files that have been read, when these files are written back to the base layer providing storages C _{1 to} C _n , they correspond to the file server that manages the metadata of these files. The input / output means U ₁₂ functioning as a write back execution means is controlled so as to write back the file to each of the base layer providing storages C _{1 to} C _n according to the priority set in the priority storage table 16 at the time of .
Therefore, for example, the file managed by the file server A _{1 in} the cache layer providing storage B ₁ of the file server A ₁ (file whose file write-back destination is the base layer providing storage C ₁ ), and the file server A ₂ (The file write-back destination is the base layer providing storage C ₂ ) and the file managed by the file server A ₃ (the file write back destination is the base layer providing storage C _3). File) and a file managed by the file server A ₄ (file whose file write-back destination is the base layer providing storage C ₄ ) exist at the same time, and the write-back priority is set in the priority storage table 16 In the situation shown in Fig. 4 (b), a write-back request for these files is issued at the same time. If it is generated, the priority order of write-back is, in descending order, the file managed by the file server A ₄ , the file managed by the file server A ₂ , the file managed by the file server A ₃ , and the file server A _1. Files managed by.

Then, the input / output unit U ₁₂ functioning as the write back execution unit receives the output of the transfer request from the transfer control unit V ₁₂ , and the file server A ₁ with the priority according to the contents of the priority storage table 16 at that time. file, written back to each of the file server _a 1 to a storage for the base layer provides managed by _{n C} 1 -C _n cache layer provides for storage _{B 1} of.

As a specific priority differentiation, for example, the files on the cache layer providing storage B ₁ are simply and continuously transferred to the base layer providing storages C _{1 to} C according to the priority set in the priority storing table 16. There is a method of writing back to _n .
In this case, for example, four files managed by the file server A ₁ , two files managed by the file server A ₂ , and files managed by the file server A ₃ are stored on the cache layer providing storage B _1. If there are three files and one file managed by the file server A ₄ , in the case of the priority setting as shown in FIG. 4B, the file server A _{4 is} first set immediately after the transfer request is generated. Are first written back to the base layer providing storage C ₄ , and then immediately _two files managed by the file server A ₂ are written back to the base layer providing storage C ₂ , and then the file server three files managed by a ₃ are written back to the base layer provides for storage C _3, then four files based managed by the file server a ₁ It will be written back to the layer provided for the storage C _1.

Further, in accordance with the setting of the priority order of the base layer providing storages C _{1 to} C _n as the write back destination, the waiting time until the start of the write back is changed so as to increase as the write back priority is lower. Thus, a method of rewriting the files on the cache layer providing storage B ₁ of the file server A ₁ to the base layer providing storages C _{1 to} C _n for each priority can be applied.
In this case, as described above, one file managed by the file server A ₄ and two files managed by the file server A ₂ are managed by the file server A ₃ on the cache layer providing storage B _1. If there are three files and four files managed by the file server A ₁ and the priority is set as shown in FIG. 4B, immediately after the transfer request is generated, A file managed by the file server A ₄ is first written back to the base layer providing storage C _4, and _two files managed by the file server A ₂ are provided for the base layer after ΔT ₁ second has elapsed after the generation of the transfer request. storage C ₂ continuously written back to the file 3 is the base layer provided for the storage C ₃ managed by the file server a ₃ after the occurrence after [Delta] T ₂ seconds elapse of the transfer request Continued to write back, four files managed by the file server A ₁ after an after [Delta] T ₃ seconds elapse of the transfer request is to be continuously written back to the base layer provided for the storage C ₁ (however, [Delta] T ₁ <ΔT ₂ <ΔT ₃ ).

  However, these methods are merely examples, and it is possible to differentiate the priority of writing back by another method.

The output of the transfer request from the transfer control means V ₁₂ to which the write-back priority is assigned to the input / output means U ₁₂ is repeatedly executed at regular time intervals, and each time the input / output means U ₁₂ functions as the write-back execution means. Is activated.
Normally, the output cycle of this transfer request is set longer than the above-mentioned maximum waiting time ΔT ₃ seconds, but the write-back of the file that has not been transferred is executed using the cycle of the next transfer request. Therefore, it is not always necessary to make the output cycle of the transfer request longer than the maximum waiting time ΔT ₃ seconds.

Alternatively, instead of outputting a transfer request at regular time intervals, for providing a cache layer file server A ₁ cache layer provides for storage B ₁ from the base layer provided for the storage C ₁ -C _n to write back to file occupies A transfer request may be output when the total storage capacity of the storage B ₁ reaches a preset threshold value.

  However, these methods are merely examples, and the output timing of the transfer request can be set by another method.

When the file write-back processing from the cache layer providing storage B _{1 of} the file server A ₁ to the base layer providing storages C _{1 to} C _n is performed in this way, the file contents are updated by the write back. Te, file server a ₁ to a _n for managing the write back destination becomes the metadata of the base layer provided for the storage C ₁ -C _n files, to update metadata in the hard disk 6 of each server by its internal processing Need arises.
File that is, if the temporarily stored file in the cache layer provides for storage B ₁ of the file server A ₁ than written back to the base layer provides for storage C _1, the microprocessor 2 of the file server A ₁ is by its internal processing The metadata in the hard disk 6 of the server A ₁ is updated, and the files temporarily stored in the cache layer providing storage B ₁ of the file server A ₁ are written back to the base layer providing storages C _{2 to} C _n. For example, the microprocessor 2 of the file server A ₁ requests the base layer providing storage to which the file is written back, for example, the file server A _y managing the base layer providing storage _Cy to update the metadata. the microprocessor of the server a _y is a file server a _y hard disk It is that updating the metadata of the inner.

In this embodiment, the file server A ₁ is reference all files using number _CA 2 to CA _n for each file server A number of files using ₁ CA ₁ and the other file server _A 2 to A _n, the magnitude of the file using the number acting on each file server _a 1 to a _n ie client machine _E 11 _{to E} 1j using each file server _a 1 ~A _{n, ···,} the input-output requests from E n1 _{to E nk} depending on the size of the load for each file server a ₁ to a _n are storage C ₁ -C _n for the base layer provided for managing those often from those small files used number, that is, those small consisting of the load on the file server after setting back priority written in the order of the large-made ones Noboribeki, file server a ₁ is off according to the priority writeback set by the file server a ₁ The Irusaba A ₁ cache layer provides for storage B ₁ on the file are written back to the file server A ₁ storage for the base layer provided to A _n managed C ₁ -C _n. Therefore, a file server on which a high load is applied at that time due to an input / output request from the client machines E _{11 to} E _1j ,..., E _{n1 to} E _nk , for example, FIG. In the example of b), the priority of the file write-back process for the base layer providing storage C ₁ managed by the file server A _{1 with} the highest load is automatically set to the lowest level, and the base layer providing storage C _The file write-back process for ₁ is postponed both in order and in time.
As a result, the file management processing required in association with the write back of the file in which the base layer providing storage serving as the write back destination becomes the base layer providing storage C ₁ , that is, the processing related to the metadata update, Opportunities to be executed on the file server A ₁ having the highest load are reduced, and the input / output requests from the client machines E _{11 to} E _1j and the metadata update process are duplicated on the file server A _1. As a result, the probability of collision decreases, and inconveniences such as input / output delays and performance degradation of the computation node in the client machines E _{11 to} E _1j constituting the computation node are improved.

Since in each of the other file server A ₂ to A _n same processing as file server A ₁ is performed, similarly to the above, relatively higher at its point by input-output requests from a client machine load is to have set the priority of the rewriting process of a file for the base layer provided for the storage managed by the file server is relatively low to act, in any of the file server a ₂ to a _n, for providing a cache layer The file write-back from the storages B _{2 to} B _n is preferentially executed from the server base layer providing storage managed by the file server with a low load.
In other words, the file write-back process to the base layer providing storage managed by the file server with relatively high load, and the metadata update process accompanying it are also postponed. It reduces the probability that the process of updating the output request and metadata from the client machines on the file server a ₁ to a _n collides overlap occurs, the load of each file server a ₁ to a _n can be reduced as a whole In addition, the performance of the entire file sharing system constituted by the base layer providing storages C _{1 to} C _n is also improved.

As another embodiment, in addition to opening / closing the file handle, the shared service providing unit also transmits a request transmitted in connection with the operation of the file handle on the extension of the system call by the client machines E _{11 to} E _1j. even when the file sharing unit U ₁₁ of U ₁ is processed, using state management unit V ₁₁ of the load management unit V ₁ is to detect the occurrence of此request, the last time the file handle and the file handle is operated usage management means V ₁₁ may be sequentially updated and stored.
When such a configuration is applied, an input / output operation is performed with a file handle open request, the file handle close request is omitted, and even when a program that is left to close by the OS at the time of termination operates, an appropriate usage situation Is obtained.

  A part or all of the embodiment disclosed above can be appropriately expressed by the description shown in the following supplementary notes, but the form for carrying out the invention and the technical idea of the invention are not limited to these. Absent.

The file write-back control program may be stored on a non-temporary recording medium. As the non-temporary recording medium, for example, a DVD, CD, flash memory or the like can be used. When storing a file write-back control program in a non-transient recording medium, the file write-back control programs are read from the non-transitory recording medium by the file server A ₁ to A _n, are executed by the microprocessor It will be.

[Appendix 1]
A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for storage in the file sharing system to write back the file on the storage for providing cached layer on the base layer provided for the storage,
For each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, sequentially updates and stores the information, and communicates with other file servers connected to the file server. Usage status management means for sequentially reading and storing each of the loads stored in each as a load for each file server;
Refer to the load of the file server stored in the usage status management means and the load of each file server, and manage it in the base layer providing storage managed by each file server according to the load size Transfer control means for setting the write-back priority in the order of increasing load of the file server to be
Write back execution means for writing back a file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the transfer control means. A file sharing system characterized by that.

[Appendix 2]
The usage status management means sequentially updates and stores the number of files opened by a request from a client machine included in the file server as a file usage number representing a load and communicates with other file servers connected to the file server. Each of the file usage numbers stored in each of the other file servers is sequentially read by information transmission, and sequentially updated and stored as a file usage number representing the load for each file server,
The transfer control means refers to the file usage count of the file server and the file usage count for each file server stored in the usage status management means, and is managed by each file server according to the size of the file usage count. The file sharing system according to appendix 1, wherein the write-back priority is set in the order of increasing the number of used files of the file server that manages the base layer providing storage.

[Appendix 3]
The usage status management unit receives a file handle open request from a client machine provided in the file server and increments the value of the file usage count of the file server by 1, while the file server having the usage status management unit includes the usage status management unit. A self-server file usage count update function for updating and storing the file usage count by decrementing the value of the file usage count of the file server in response to a file handle close request from a client machine;
There is no file handle that is the target of an open request or close request in the storage provided by the base layer managed by the file server, and the information is communicated to other file servers connected to the file server. When it becomes necessary to request a file handle or file close, along with the file handle or file close request, the current value of the file usage number of the file server is transferred to the other file server, and another file Receiving the file handle and file close completion notification from the server and the current value of the file usage of the file server, and rewriting the value of the file usage of the file server that received the current value to the received current value By each file And other server files use the number of update function for updating and storing the number of file-per-use over server,
When a file handle or file close request is received by communicating information with another file server, the current value of the number of used files on that file server is displayed along with the corresponding file handle or close completion notification. The file sharing system according to appendix 2, further comprising a file usage number transfer function for transferring to a file server as a request source.

[Appendix 4]
Instead of the other server file usage update function and the file usage transfer function,
The usage status management means is
Information is transmitted between the file server having the usage status management means and another file server connected to the file server at predetermined intervals, and the current value of the file usage number of the file server is set to the other file. And a server file usage number transfer update function for receiving the file usage count of the file server from the other file server and updating and storing the file usage count for each file server. The file sharing system according to appendix 3.

[Appendix 5]
The write back execution means is
According to the write-back priority set by the transfer control means, the files on the cache layer providing storage of the file server are managed by each file server in order from the base layer providing storage having the higher write-back priority. The file sharing system according to any one of Supplementary Note 1, Supplementary Note 2, Supplementary Note 3 and Supplementary Note 4, wherein the file sharing system is configured to write back to the storage for providing the base layer.

[Appendix 6]
The write back execution means is
In accordance with the write-back priority set by the transfer control means, the waiting time until the write-back start increases as the write-back priority becomes lower, and the file on the cache layer providing storage of the file server is The file sharing system according to any one of Supplementary Note 1, Supplementary Note 2, Supplementary Note 3 and Supplementary Note 4, which is configured to write back to the storage for providing base layer for each priority.

[Appendix 7]
The file sharing system according to any one of Supplementary Note 1, Supplementary Note 2, Supplementary Note 3, Supplementary Note 4, Supplementary Note 5 and Supplementary Note 6, wherein the writing back execution unit is operated at a constant time interval.

[Appendix 8]
The write back execution means is activated when the total capacity of files temporarily stored in the cache layer providing storage of the file server reaches a predetermined threshold value. The file sharing system according to any one of Appendix 3, Appendix 4, Appendix 5, and Appendix 6.

[Appendix 9]
The base layer providing storage is configured by a mass storage device, and the cache layer providing storage is configured by a high speed storage device. Supplementary Note 1, Supplementary Note 2, Supplementary Note 3, Supplementary Note 4, Supplementary Note 5, Supplementary Note 6, Supplementary Note 7 Or the file sharing system according to any one of appendix 8.

[Appendix 10]
A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back control program for file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
A microprocessor included in each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, and sequentially updates and stores it in the storage area of the memory of the file server, and other file servers connected to the file server Usage status management means for sequentially reading each of the loads stored in each of the other file servers by information transmission of the above, and sequentially updating and storing them in the other storage areas of the memory as loads for each file server,
Refers to the load of the file server stored in the storage area of the memory and the load of each file server stored in the other storage area of the memory, and is managed by each file server according to the size of the load Transfer control means for setting the write-back priority in the order of increasing load of the file server that manages the storage for the base layer provided, and
According to the write-back priority set by the transfer control means, it functions as a write-back execution means for writing back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server. A file write-back control program.

[Appendix 11]
A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back control program for file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
A microprocessor included in each of the file servers,
In response to a request from a client machine included in each file server, the number of files opened by that file server is sequentially updated and stored in the memory area of the file server as the number of used files, and other files connected to the file server Usage of sequentially reading each of the file usage numbers stored in each of the other file servers by transmitting information to the file server, and sequentially updating and storing them in the other storage area of the memory as the file usage count for each file server Situation management means,
Refer to the file usage count of the file server stored in the storage area of the memory and the file usage count for each file server stored in the other storage area of the memory, and according to the size of the file usage count Transfer control means for setting the write-back priority in the order of ascending to the base layer providing storage managed by each file server, and
According to the write-back priority set by the transfer control means, it functions as a write-back execution means for writing back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server. A file write-back control program.

[Appendix 12]
Each of the microprocessors functioning as the usage status management means,
In response to a file handle open request from a client machine provided in the file server having the usage status management means, the value of the file usage count of the file server is incremented by 1, while the client provided in the file server having the usage status management means Receiving a file handle closing request from the machine, and decrementing the value of the file usage number of the file server by 1 to update and store the file usage number in the storage area of the memory. ,
There is no file handle that is the target of an open request or close request in the storage provided by the base layer managed by the file server, and the information is communicated to other file servers connected to the file server. When it becomes necessary to request a file handle or file close, along with the file handle or file close request, the current value of the file usage number of the file server is transferred to the other file server, and another file Receiving the file handle and file close completion notification from the server and the current value of the file usage of the file server, and rewriting the value of the file usage of the file server that received the current value to the received current value By each file Other server files using number updating function implementing section to the number of files per-use over server is updated and stored in the other storage area of the memory and,
When a file handle or file close request is received by communicating information with another file server, the current value of the number of used files on that file server is displayed along with the corresponding file handle or close completion notification. The file write-back control program according to appendix 11, wherein the file write-back control program is made to function as a function for transferring the number of used files to be transferred to a requesting file server.

[Appendix 13]
Instead of the other server file usage number update function realization means and the file usage number transfer function realization means,
Each of the microprocessors is
Information is transmitted between the file server having the usage status management means and another file server connected to the file server at predetermined intervals, and the current value of the file usage number of the file server is set to the other file. Server file usage number transfer for transferring to the server and receiving the file usage number of the file server from the other file server, and updating and storing the file usage number for each file server in the other storage area of the memory The file write-back control program according to appendix 12, which functions as an update function realization means.

[Appendix 14]
The write back execution means is
According to the write-back priority set by the transfer control means, the files on the cache layer providing storage of the file server are managed by each file server in order from the base layer providing storage having the higher write-back priority. 14. The file write-back control program according to any one of appendix 10, appendix 11, appendix 12, or appendix 13, wherein the file write-back control program is configured to write back to the base layer providing storage.

[Appendix 15]
The write back execution means is
In accordance with the write-back priority set by the transfer control means, the waiting time until the write-back start increases as the write-back priority becomes lower, and the file on the cache layer providing storage of the file server is 14. The file write-back control program according to any one of Supplementary Note 10, Supplementary Note 11, Supplementary Note 12, or Supplementary Note 13, configured to write back to the storage for providing base layer for each priority.

[Appendix 16]
16. The file write-back control program according to any one of Supplementary Note 10, Supplementary Note 11, Supplementary Note 12, Supplementary Note 13, Supplementary Note 14, or Supplementary Note 15, wherein the write-back execution unit is operated at a constant time interval.

[Appendix 17]
Appendix 10, appendix 11, wherein the write-back execution means is activated when the total capacity of files temporarily stored in the cache layer providing storage of the file server reaches a predetermined threshold The file write-back control program according to any one of Appendix 12, Appendix 13, Appendix 14, and Appendix 15.

[Appendix 18]
A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back method for a file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
Each of the file servers detects a load acting on the file server in response to a request from a client machine included in each file server, sequentially updates and stores the information, and information on other file servers connected to the file server Each of the loads stored in each of the other file servers is sequentially read by transmission, and sequentially updated and stored as a load for each file server.
Each of the file servers refers to the load of the file server and the load of each file server, and manages the file server in the base layer providing storage managed by each file server according to the size of the load. Set the write-back priority in order of ascending load
Each of the file servers writes back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the file server. A file write-back method characterized by that.

[Appendix 19]
A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back method for a file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
Each of the file servers sequentially updates and stores the number of files opened by the file server in response to a request from a client machine included in each file server as a file usage number, and is connected to the file server. Each of the file usage numbers stored in each of the other file servers by sequentially transmitting information and sequentially updating and storing the file usage numbers for each file server,
Each of the file servers refers to the number of files used by the file server and the number of files used for each file server, and increases to the base layer providing storage managed by each file server according to the size of the number of files used. Set the write-back priority in the order of 冪
Each of the file servers writes back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the file server. A file write-back method characterized by that.

[Appendix 20]
Each of the file servers receives a file handle open request from a client machine included in the file server and increments the value of the number of used files of the file server by 1, while the file server having the usage status management unit includes In response to a request for closing a file handle from a client machine, the file usage number of the file server is decremented by 1, and the file usage number is updated and stored
In addition, there is no file handle that is the target of an open request or close request in the base layer providing storage managed by the file server, and other files are transferred by transmitting information to other file servers connected to the file server. When it becomes necessary to request the server to close the file handle or file, along with the request to close the file handle or file, the current value of the file usage number of the file server is transferred to the other file server. Received the file handle and file close completion notification from the file server and the current value of the file usage count of the file server, and received the current value of the file usage count of the file server that received the current value. By rewriting And updates and stores the number of files per-use Irusaba,
In addition, when a file handle or file close request is received through information transmission with another file server, the current value of the number of used files on the file server is displayed along with the corresponding file handle or file close completion notification. The file write-back method according to appendix 19, wherein the current value of the number of used files of the file server is transmitted to another file server by transferring the file to the file server that is the request source for closing the file.

[Appendix 21]
At each predetermined period, each of the file servers communicates information with other file servers connected to the file server, and the current value of the file usage number of the file server is transferred to the other file server. The file write-back method according to appendix 19, wherein the file usage count of the file server is received from the other file server, and the file usage count for each file server is updated and stored.

[Appendix 22]
In accordance with the write-back priority set for each of the file servers, each of the file servers, in order from the base layer providing storage with the higher write-back priority, files on the cache layer providing storage of the file server. 22. The file write-back method according to any one of appendix 18, appendix 19, appendix 20, or appendix 21, wherein the file is written back to a base layer providing storage managed by each file server.

[Appendix 23]
According to the write-back priority set for each of the file servers, each of the file servers increases the waiting time until the start of write-back as the write-back priority becomes lower. The file write-back method according to any one of appendix 18, appendix 19, appendix 20, or appendix 21, wherein the file on the layer providing storage is written back to the base layer providing storage for each priority level. .

  The present invention can be used to detect the load status of a file server in a shared file system and to control the movement of files between tiers in a shared file system using a tiered storage.

1 File sharing system 2 Microprocessor (Usage status management means, own server file usage count update function implementation means, other server file usage count update function implementation means, transfer control means, write back execution means)
3 ROM
4 Nonvolatile memory 5 RAM (memory)
6 Hard disk 7 Monitor 8 Keyboard 9 Mouse 10 Input / output circuit 11 Interface 12 Interface 13 Interface 14 Own data storage table 15 Other data storage table 16 Priority storage table A _{1 to An} File server B _{1 to} B _n Storage for providing cache layer _{C 1} ~C _n, C _m base layer provides for storage _D 1 to D _n local area network _{E 11 ~E nk, E 1i,} E 1j client machine (compute nodes)
F local area network G Storage Area Network H storage area network _P 1 to P _i compute nodes (client machine)
Q ₁ to Q _j storage area network _R 1 to R _j local area network _S 1 to S _j file server _T 1 through T _j storage (cache layer)
U ₁ ~U _n sharing service providing unit _{U 11} file sharing unit _{U 12} output means (writeback executing means)
V _{1 to} V _n Load management unit V ₁₁ usage status management means V ₁₂ transfer control means W storage (base layer)
CA ₁ Current file usage count CA _{2 to} CA _n Current file usage count of other file servers

Claims (10)

A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for storage in the file sharing system to write back the file on the storage for providing cached layer on the base layer provided for the storage,
For each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, sequentially updates and stores the information, and communicates with other file servers connected to the file server. Usage status management means for sequentially reading and storing each of the loads stored in each as a load for each file server;
Refer to the load of the file server stored in the usage status management means and the load of each file server, and manage it in the base layer providing storage managed by each file server according to the load size Transfer control means for setting the write-back priority in the order of increasing load of the file server to be
Write back execution means for writing back a file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the transfer control means. A file sharing system characterized by that. The usage status management means sequentially updates and stores the number of files opened by a request from a client machine included in the file server as a file usage number representing a load and communicates with other file servers connected to the file server. Each of the file usage numbers stored in each of the other file servers is sequentially read by information transmission, and sequentially updated and stored as a file usage number representing the load for each file server,
The transfer control means refers to the file usage count of the file server and the file usage count for each file server stored in the usage status management means, and is managed by each file server according to the size of the file usage count. 2. The file sharing system according to claim 1, wherein the write-back priority is set in the order in which the number of used files of the file server that manages the storage is provided in the base layer providing storage. The usage status management unit receives a file handle open request from a client machine provided in the file server and increments the value of the file usage count of the file server by 1, while the file server having the usage status management unit includes the usage status management unit. A self-server file usage count update function for updating and storing the file usage count by decrementing the value of the file usage count of the file server in response to a file handle close request from a client machine;
There is no file handle that is the target of an open request or close request in the storage provided by the base layer managed by the file server, and the information is communicated to other file servers connected to the file server. When it becomes necessary to request a file handle or file close, along with the file handle or file close request, the current value of the file usage number of the file server is transferred to the other file server, and another file Receiving the file handle and file close completion notification from the server and the current value of the file usage of the file server, and rewriting the value of the file usage of the file server that received the current value to the received current value By each file And other server files use the number of update function for updating and storing the number of file-per-use over server,
When a file handle or file close request is received by communicating information with another file server, the current value of the number of used files on that file server is displayed along with the corresponding file handle or close completion notification. 3. The file sharing system according to claim 2, further comprising a file usage number transfer function for transferring to a file server as a request source. Instead of the other server file usage update function and the file usage transfer function,
The usage status management means is
Information is transmitted between the file server having the usage status management means and another file server connected to the file server at predetermined intervals, and the current value of the file usage number of the file server is set to the other file. And a server file usage number transfer update function for receiving the file usage count of the file server from the other file server and updating and storing the file usage count for each file server. The file sharing system according to claim 3. The write back execution means is
According to the write-back priority set by the transfer control means, the files on the cache layer providing storage of the file server are managed by each file server in order from the base layer providing storage having the higher write-back priority. 5. The file sharing system according to claim 1, wherein the file sharing system is configured to write back to the storage for providing the base layer. The write back execution means is
In accordance with the write-back priority set by the transfer control means, the waiting time until the write-back start increases as the write-back priority becomes lower, and the file on the cache layer providing storage of the file server is 5. The file sharing according to claim 1, wherein the file sharing is configured to write back to the storage for providing the base layer for each priority. 6. system.   The write-back execution means is operated at a constant time interval, according to any one of claims 1, 2, 3, 4, 5, and 6. File sharing system.   2. The write-back execution means is operated when the total capacity of files temporarily stored in the cache layer providing storage of the file server reaches a predetermined threshold. The file sharing system according to claim 2, claim 3, claim 4, claim 5, or claim 6. A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back control program for file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
A microprocessor included in each of the file servers,
Detects the load acting on the file server in response to a request from the client machine provided in each file server, and sequentially updates and stores it in the storage area of the memory of the file server, and other file servers connected to the file server Usage status management means for sequentially reading each of the loads stored in each of the other file servers by information transmission of the above, and sequentially updating and storing them in the other storage areas of the memory as loads for each file server,
Refers to the load of the file server stored in the storage area of the memory and the load of each file server stored in the other storage area of the memory, and is managed by each file server according to the size of the load Transfer control means for setting the write-back priority in the order of increasing load of the file server that manages the storage for the base layer provided, and
According to the write-back priority set by the transfer control means, it functions as a write-back execution means for writing back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server. A file write-back control program. A storage for providing a base layer that connects a plurality of file servers having a plurality of client machines constituting a computer node so that information can be transmitted, and is managed by each of the file servers and provides a base layer to each of the file servers; A cache layer providing storage for each file server that provides a cache layer to each of the file servers is connected to each of the file servers via a storage area network, and a file requested by each client machine is A file server connected to the client machine reads from the base layer providing storage via the storage area network, temporarily stores it in the cache layer providing storage, and caches processing according to a request from the client machine. After subjecting to a file on the layer provided for the storage, a file write-back method for a file sharing system files on the storage for providing cached layer was written back to the base layer provided for the storage,
Each of the file servers detects a load acting on the file server in response to a request from a client machine included in each file server, sequentially updates and stores the information, and information on other file servers connected to the file server Each of the loads stored in each of the other file servers is sequentially read by transmission, and sequentially updated and stored as a load for each file server.
Each of the file servers refers to the load of the file server and the load of each file server, and manages the file server in the base layer providing storage managed by each file server according to the size of the load. Set the write-back priority in order of ascending load
Each of the file servers writes back the file on the cache layer providing storage of the file server to the base layer providing storage managed by each file server according to the write back priority set by the file server. A file write-back method characterized by that.

Images