JP2006085209A - Deployment method of computer system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem in a current deployment method wherein a latency time is generated up to start of a business because an added server cannot start the business while copying from a master disk to a system disk of each server blade is executed, and it is started after completing the copying to start business processing. <P>SOLUTION: In a computer system including a plurality of server modules and at least one or more master disks storing operating systems and applications, which are mutually connected through a backplane, each server module including a processor, a memory, and a system disk for storing operating systems and applications, when a new server module is connected to the backplane, the processor of this server module reads an operating system and application from any master disk followed by storing in the memory of the server module for execution, and also stores the operating system and application read from the master disk in its system disk. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a computer system that distributes operations to a plurality of servers to perform processing, and more particularly to a server deployment method.

  In recent years, a computer called a blade server has been commercialized as a new form of server. A blade server is a backplane (or network switch) that connects power supplies, cooling fans, and server blades that are shared by multiple server blades, each blade functioning as one server, called a server blade. The server can be installed at a higher density than a conventional rack-mount server by being inserted into a chassis having a rack. With a blade server, when a server is added, simply insert the server blade into the chassis and connect to other server blades in the same chassis as well as servers, storage, and networks outside the chassis. The type server also has the feature that it is very easy to install, without the need for cable connection required for each server installation and expansion.

  In blade servers, deployment of each server blade from a single management terminal to a large number of server blades (in the sense of frontline deployment, operating states and applications installed on a server blade inserted in the chassis is a state where operations can be performed. And the assignment of tasks to each server blade can be performed in a centralized manner, thereby enabling easy operation management of the entire blade server system.

  Currently, the following methods are generally used for blade server deployment.

  First, in advance, install the operating system and applications necessary to process a job on the system disk of one server blade as usual, such as using an installation CD-ROM. After installation is complete, copy the contents of the system disk to another disk. This copied disk will be the master disk that will be required in the future when this task is assigned to the server blade. This master disk is created for each type of business to be assigned.

  When a server blade is added, a file is copied from the master disk corresponding to the task assigned to this server blade to the system disk of the added server blade using the management terminal. Since the master disk has been created with the operating system and applications already installed, the operating system and applications are installed on the system disk of the added server simply by copying.

  The information specific to each server blade, such as the server name and IP address necessary for identifying the last added server blade on the system, is set using the management terminal to complete the deployment. After the deployment is complete, start the added server blade and start business processing.

JP 2001-134482 A

  In the current deployment method, the added server cannot start a business while copying from the master disk to the system disk of each server blade (server module). Since it starts after copying is completed and starts business processing, there is a waiting time until the business starts.

  The blade server has features such as easy addition of server blades (server modules) and the ability to operate and manage a large number of server modules from a single management terminal. Adding more servers and changing the assignment of tasks to servers than conventional cabinet-type or rack-mount servers, such as adding as many server blades as necessary and dynamically changing the assignment of tasks to each server blade To be done. However, the conventional deployment method always requires a waiting time for disk copy, and there is a problem that it is impossible to quickly expand the business system and change the allocation of system processing capacity to each business.

  The present invention starts each server blade before the copying of the operating system and application from the master disk to the system disk is completed, and shortens the time from the addition of the blade server to the start of the work, thereby realizing a quick start of the work. For the purpose.

  The present invention is a system in which a plurality of server modules and at least one master disk storing an operating system and applications are connected via a backplane, and each server module stores a processor, memory, an operating system and applications. In a computer system equipped with a disk, when a new server module is connected to the backplane, the processor of the server module reads the operating system and application from one of the master disks and stores it in the memory of the server module. And the operating system and application read from the master disk at the same time are stored in the system disk. A deployment system of the computer system.

  That is, the file read from the master disk into the memory is simultaneously written into the system disk of the server blade, and is copied from the master disk to the system disk in parallel with the startup process.

  Each server blade can be started before the operating system and application copy from the server blade master disk to the system disk is completed, and it is possible to shorten the time from server addition to the start of the business and realize a quick business start. Become.

  FIG. 1 shows a block diagram of the entire system for realizing the present invention, and shows an embodiment in a blade server as an example. The configuration of this system will be described below.

  The server blade (server module) A10 includes a processor 101, a memory 102, an I / O controller 103, a network I / F 105, a disk I / F 106, and a system disk A107. The I / O controller 103 has an internal memory 104 for storing information unique to each server blade such as access destination information and initial setting information. FIG. 2 shows an information storage area of the built-in memory 104, which is composed of a master disk information storage area 1041 and an initial setting information storage area 1042.

  In the example shown in FIG. 1, the system disk A107 of the server blade (server module) A10 is directly connected to the server blade A10 via the disk I / F 106, but may be connected via the network via the backplane 40. is there.

  The server blade A10 is connected to the master disk A50, the master disk B60, and the management terminal 70 via the backplane 40. In addition to the server blade A10, a plurality of servers such as a server blade B20 and a server blade Z30 can be connected to the backplane 40. By connecting a plurality of backplanes 40 in parallel, a plurality of servers exceeding the number that can be connected via one backplane may be connected.

  The master disk A50 is a disk that stores an operating system and applications for performing business A, and the master disk B60 is a disk that stores an operating system and applications for performing business B. The master disk A50 and master disk B60 are disks in which files serving as operating system and application masters installed on system disks inherent to servers connected to the backplane 40 are stored. The contents of the master disk can be read from each connected server blade, but the contents cannot be changed. In addition to the tasks A and B, the backplane 40 can be connected to a plurality of master disks storing operating systems and applications for performing other tasks.

  The management terminal 70 is a terminal that manages each server blade connected to the backplane 40. The management terminal 70 can detect connection of each server blade to the backplane 40 and set initial setting information via the backplane 40.

  The procedure for carrying out the present invention will be described below.

  FIG. 3 shows a procedure from addition of a server blade to activation of the server blade.

  When a server blade A10 is newly added to the system (300), the management terminal 70 detects that the server blade A10 is connected to the backplane 40 via the backplane 40 (301).

  The management terminal 70 transmits master disk information indicating which master disk is used and initial setting information unique to the server blade A10 to the I / O controller 103 of the server blade A10. The master disk information includes, for example, a master disk A50 storing an operating system and applications used in the business A when the business A is allocated to the server blade A10, and an operating system used in the business B when the business B is allocated. The master disk B 60 in which the application is stored changes depending on the work assigned to the server blade A 10. Here, it is assumed that the operation A is allocated and the information of the master disk A50 is transmitted as master disk information (302).

  The I / O controller 103 stores the master disk information and the initial setting information in the master disk information storage area 1041 and the initial setting information storage area 1042 of the internal memory 104 of the I / O controller 103. The initial setting information can be stored not only in the built-in memory 104 but also as a file on the system disk A 107. The I / O controller 103 completes the storage of these pieces of information and notifies the management terminal 70 that the server blade A10 is ready to start (303).

  After confirming that the server blade A10 is ready for activation, the management terminal 70 instructs the processor 101 of the server blade A10 to start activation (304).

  The processor 101 that has received the start-up instruction requests the I / O controller 103 to read the necessary file from the system disk A 107 in order to read the file such as the operating system from the system disk A 107 and store it in the memory 102. I do.

  FIG. 4 shows a file access control procedure by the I / O controller. The processor 101 requests a file from the I / O controller 103 (401), and the I / O controller 103 checks whether the requested file exists in the system disk A 107 via the disk I / F 106 (402).

  Immediately after the server blade A10 is newly added to the system, the system disk A107 is empty and no operating system file is stored, so the I / O controller 103 reads the requested file from the system disk A107. I can't. If the requested file does not exist on the system disk A 107, the I / O controller 103 refers to the master disk information storage area 1041 in the internal memory 104, and which master disk is used to start the server blade A10. Confirm. Here, as an example, since the master disk information storage area 1041 stores information that the master disk to be used is the master disk A50, the I / O controller 103 accesses the master disk A50 via the network I / F 105. The requested file is read (403). The I / O controller 103 transmits the file read from the master disk A50 to the processor 101. At the same time, the I / O controller 103 stores the file read from the master disk A50 in the system disk A107. For this reason, as the startup process proceeds and the file is read from the master disk A50, the file copy from the master disk A50 to the system disk A107 progresses at the same time. Since the file once read from the master disk A50 is read from the system disk A107 for the second time and thereafter, reading of the file from the master disk A50 gradually decreases (404).

  The processor 101 stores the file received from the I / O controller 103 in the memory 102 and executes it (406). In this way, the operating system is executed, and then the business application is read and executed in the same manner.

  Since the I / O controller 103 controls whether the file is read from the system disk A 107 or the master disk A 50, the processor 101 recognizes that all the files transmitted from the I / O controller 103 are read from the system disk A 107. To do. For this reason, in the server blade A10, the operating system and applications necessary for starting and business processing are preinstalled in the system disk A107 in spite of the fact that the files necessary for starting are not actually present on the system disk A107. It is possible to start up and start business processing in the same manner as in

  When there is a write request from the processor 101 to the system disk A 107, the I / O controller 103 performs a write process only on the system disk A 107 and does not perform a write process on the master disk A 50. Thus, the temporary file or the like temporarily created while the server blade A10 is executing the operating system or application is always stored only in the system disk A107, and the contents of the master disk A50 are not changed.

  Further, the I / O controller 103 always checks the copy status of the file from the master disk A50 to the system disk A107, and after the server blade A10 starts the business process, it is still copied from the master disk A50 to the system disk A107. When there are no files, the files are copied in the background such as access processing to the outside from the system disk A 107 or the network I / F 105 from the processor 101. When copying of all files from the master disk A50 to the system disk A107 is completed, the I / O controller 103 notifies the management terminal 70 via the backplane 40 that the installation of the server blade A10 to the system disk A107 is completed. Notice. Thereafter, the server A10 can perform the operation A without accessing the master disk A50.

1 is a block diagram illustrating an entire system realized by the present invention. It is the figure which showed the information storage area | region which the internal memory mounted in an I / O controller has. It is the flowchart which showed the procedure from the addition of a server module to starting. It is the flowchart which showed the procedure of the file access control by I / O controller.

Explanation of symbols

10 Server blade A
40 Backplane 50 Master disk A
70 Management terminal 101 Processor 102 Memory 103 I / O controller 104 Built-in memory 105 Network I / F
106 Disk I / F
107 System disk A

Claims (1)

A plurality of server modules and at least one master disk storing an operating system and applications are connected via a backplane, and each server module includes a processor, memory, a system disk storing an operating system and applications. When a new server module is connected to the backplane, the processor of the server module reads the operating system and application from one of the master disks, stores it in the memory of the server module, and executes it simultaneously. An operating system and an application read from the master disk are stored in the system disk. Arm of the deployment system.

Images