JP2003529847A - Construction of component management database for role management using directed graph - Google Patents

Abstract

SUMMARY A system and method for monitoring and controlling various hardware and software components in both a computer system and a set of linked computer systems (70, 72, 74). A method that allows the component to communicate with the component management database (14 in FIG. 1) is used for the component, and the component management database is used by the configuration manager (40). A component can represent its parameters, its relationship to other components, and its performance criteria. The configuration manager can use this information to monitor and control these components to maximize system or network availability.

Description

Detailed Description of the Invention

The present invention relates to the field of increasing the availability of computer systems and network computer systems.

The present application is based on provisional application number 60 / 194,375 and filed on April 4, 2000.
Priority is given on the 4th of a month.

Current interface buses are designed to meet standards that allow the addition and removal (insertion and removal) of hardware components that are inserted into these buses without first removing power. When an insertion or extraction event occurs, a signal is generated that signals this event. When this event signal is used across multiple operating systems and multiple hardware platforms, there has never been a solution that can use this event signal to maintain the integrity of the operation of the system. The present invention provides such a solution. Applications that can determine the interdependence of software and hardware components of the system, monitor the operational status of these components, and manage the movement of these resources as needed to maximize system performance. Finally, the standard's ability to insert and withdraw resources while maintaining a powered up environment can be fully utilized. This solution is scalable to resource management and migration regardless of the type of hardware and software resources present in the system.

This situation is further complicated in a client / server network environment.
While this environment may be less expensive than mainframe systems, the management and support issues associated with this environment are becoming more complex. These problems are exacerbated by spreading multiple applications across multiple hardware systems. Administrators seeking to maintain a high level of network availability require ever more sophisticated tools to monitor network performance and, if problems arise, fix them. The present invention provides these tools.

By maintaining a database of system components and their various interdependencies and monitoring the performance and operational status of these components, the system can be managed to provide a high level of system availability. Can maximize the "uptime" of the system. The component itself is
It uses a distributed messaging system to communicate with each other and a dynamic configuration manager that holds a database of system components. These messages include the state and interdependencies of each component. During system operation, changes in component states are communicated to the Dynamic Configuration Manager, which is responsible for moving available resources to maximize system availability (uptime).

The same type of management is available in multiple computer network systems. In this system, each computer system including hardware, operating system, application, and communication function is also called “node”. In this case, if the microprocessor running the dynamic configuration manager itself becomes unavailable, the activity of the dynamic configuration manager can be forwarded to another microprocessor node on the network.

[0007]

DETAILED DESCRIPTION OF THE INVENTION

The features of the invention believed to be novel are set forth in detail in the appended claims. The invention, and further objects and advantages of the invention, can best be understood by reference to the following description taken in connection with the accompanying drawings. In the drawings, like reference numbers identify like elements.

Management software requires a detailed understanding of the types of managed components installed in the system and the relationships between those components. Software must dynamically track the topology of these managed components and the individual configuration of each component. To address a wide range of components within a typical system, the software recognizes CPU modules, I / O cards, peripherals, applications, drivers, power supplies, fans, and other system components,
You have to manage these. If the configuration changes, the system must take appropriate action to load or unload the driver and inform other components that may be affected by this change. The interdependence of various components is essential information for building a highly available system.

In order to be effectively managed, components need to be represented in one standard central repository. This component management database must contain information about each component and information about the relationship between components. For example, a daughter board connected to an I / O card represents a parent-child relationship. Also, the component table must be able to identify the group of components that share responsibility for a given operation. Finally,
The component management database must store information about actions that need to be taken on components that depend on 1) components removed from the system or 2) other components removed from the system.

As managed components are added and removed, the system needs a mechanism to track these events. If the component type and location are fixed, the system can periodically poll the component to determine its presence. However, as component types and locations change, the system requires a more intelligent way of identifying components. In the preferred embodiment, the component must be able to identify and reveal itself to the system and describe its functional capabilities so that the management software understands the component's function in advance. Remove the need. This mechanism is essential for enabling hot swap and non-resident (short term) components.

To accomplish this, the component can be enabled by the publish and subscribe feature with the dynamic configuration manager. When a component is loaded or inserted into the system, the component sends its identity to the configuration manager. The configuration manager queries the component and determines its type and function. The component is then placed in the list of managed components and properly monitored. Each different component or class of components can have its own set of methods that can be invoked.
When a component is removed, the configuration manager triggers the appropriate action. For cards, this action may include unloading the driver and transferring the operation to the redundant card.

The management software must provide a mechanism for system components such as cards, drivers, and applications to communicate with each other within the system or with other system components within the cluster. Distributed message services carry these messages. This service uses a "publish and subscribe" model. The software provides client, server, and routing functionality. To send the message, the component passes the message to management software. If a new publisher appears, inform all subscribers that the new publisher has registered. This message service provides global event classes and event names to enable the routing of messages across "bridged" networks. Instead of using broadcast packets, which can be blocked by firewalls and routers, the message service sets up a connection with each system. With these individual system connections, messages are routed to the correct system without interruption.

FIG. 1 is a high-level diagram showing how components are managed. When a component is added, changed or removed (12), the component management database (14) is updated to reflect that fact. The database is constantly monitored (16) for changes that meet predetermined criteria. If the change in the monitored component meets this criterion, then the event can be sent (issued) to the listening (subscribing to) event or the detector (22 and 24).
. If the event being subscribed to meets other predetermined criteria, execute the state or detector script (26). This script is a component management database 1
It has the function of changing 4.

FIG. 2 illustrates a preferred embodiment of the present invention that can be used in web trading applications. The component management database, the configuration management function, and the role management function are provided by the EMP-manager block 40. EM
P (embedded management process) is an application executed on the operating system 44. EMP has a number of APIs (Application Program Interfaces) that provide functions that the system can call to implement component management, configuration management, and role management processes. The application 42 uses these APs
Written using I. The driver software 46 that provides the interface to other hardware components can also be written to take advantage of the functionality provided by the API. A board 48 such as a network interface card (NIC) controlled by the driver 46 may be incorporated in the component management database and appropriately managed using predetermined operation rules.

FIG. 3 shows a state machine that abstracts the events that a component can react to. A state machine can not only respond to an event, but can generate other actions and reactions other than the event that triggered the reaction. The reaction generated is determined by the state of the component as it receives the event. State S0 exists when the card is currently inserted into the appropriate operating system bus. Event E1 occurs when the card begins to be removed from the bus. An instruction is sent to the component management database to set its state to "extracting". An instruction is then sent to change the state of their children (components that depend on the card to operate correctly) to "spare." Event E2 occurs when the card is removed. The state of the card is defined as "completed", a command is sent to the database to reflect this state and a "track" command is set. A "track" command is a piece of data that remains in memory to reflect the sequence of operations that affect the components listed in the database. By examining the recorded trace events, the history so far can be restored.

Insertion event E3 is very similar to the extraction event: the instruction issued by the state wants to issue a request to reload the driver needed to operate the card as the state is placed in the database. Reflect that. When the load is going well, the component requests a status update in the database to reflect its existence and operation.

The configuration management database shown in FIG. 4 represents one of many ways in which the information present in the database can be represented. The address field 50 of the database is the global IP address of the listed component. The fact that the IP address is used enforces the fact that this information can be used across multiple networks, including the Internet, not just a particular network. In the preferred embodiment, the communication protocol used to send and receive information between networks is TCP / IP. TCP
The preferred API for accessing / IP is the socket interface.
The address using the TCP / IP socket is the Internet address (IP_addr
ess) and port number 52. The port is an entry point to an application existing in the host 54 (network machine). The database also provides the name 56 of the cluster (a collection of interconnected computers used as a single integrated computing resource). Then there is the management role 58 assumed by the host, and the last field shown is the desired management role 60 that the system seeks to obtain. In the preferred network embodiment, the protocol used is HTTP, which establishes a client / server connection, sends and receives parameters including returned files, and interrupts the client / server connection. . The language used to write documents using the HTTP protocol is HTML. In the preferred embodiment, a copy of the component management database information is generated by the small footprint web server and made available to other nodes in the system. The web server runs on an operating system that also runs the component management database system. The information and messages that need to be sent over the network using the TCP / IP protocol are first described in XML (Extensible Markup Language) using tags that are specifically defined to identify the parameters of the monitored and controlled components. Be translated into For example, keep the component management database in the dynamic memory of the processor board, keep a copy of it on your computer or network hard drive, and copy another copy of the XML markup language to client components on other linked networks. Can be used to send.

In a preferred embodiment of the present invention, a cluster of components can be managed by executing a common component management database instruction for each branch of the cluster. This allows the cluster to be managed centrally. Each branch of the cluster can find each other and communicate on the network. To make these sets of instructions a single entity, a single cluster name and communication port are assigned to them. The commands begin to send their presence towards each other as soon as the system is booted. Once all have communicated, the instructions begin selecting cluster-wide managers. You can pre-select a manager,
It may be selected dynamically by a nomination and "voting" process. When a cluster manager is selected, other entities become clients of that manager. If no manager is selected, activate the timing mechanism that selects a cluster manager from that group. This algorithm ensures that the cluster always has the right manager. When a new entity joins the cluster, all other entities reunited to determine the most appropriate manager according to preselected criteria. The managing cluster entity is lost, among other things, a communication port for sending and receiving information about the functional status of the managed entity, the amount of time between status updates that the manager can tolerate, and before the client considers it to be “lost”. It receives configuration information from the client entity, including the number of possible continuous status updates, and what the manager must issue when the client is determined to "lose". All this and other information is stored in the cluster manager's database. Each client also maintains a cluster database, which stores information about itself and the cluster manager.

When the cluster manager receives this information from the client, the cluster manager initiates normal operations including maintaining a connection with the client, monitoring the client's status, and routing published cluster events to the subscribing application. To do. The client, on the other hand, sends database information to the manager, responds to status requests, detects if the cluster manager is lost or corrupted, and selects a new cluster manager if the cluster manager is found to be lost. Initiate normal operations, including joining and issuing messages routed by the cluster manager to the subscribing entity.

FIG. 5 illustrates three operating systems that can be used to manage components. Machine A70 has machine entities B72 and C7
Nominated by 4 as Manager. Entity B and entity C are client entities. The machine in this configuration controls three types of components: an electronic circuit board 76 (also known as a card), drivers 78, which are the interface between the board and the application, and applications such as 80. The broken line shows critical dependence, and the solid line shows non-critical dependence. Double line 86 is machine C
Machine B has the function of taking over and controlling the board 82 when 74 fails. A critical dependency is that if one component fails due to a failure, all other components that can fail due to the failed component have so-called critical dependencies. Is. The board 76 critically depends on the operating system (O / S) 70. Double line 86 indicates that the operating system of machine B can take over board 82 if the operating system of machine C fails.

FIG. 6 illustrates how the component management database can be configured to generate events upon component failure. There is a host IP address 92, a host name 94, and a cluster listening port 96 defined as a network port through which the component management system sends and receives broadcast messages. This port is the same for all component management systems in the cluster. Then there is the heartbeat period 98, expressed in milliseconds, the reciprocal of which is the heartbeat pulse to be generated per second. A heartbeat is a periodic signal sent from one component to another, indicating that the signaling unit is still functioning correctly.
Next is the heartbeat port 100, which is the network port through which the component management database receives heartbeats from the cluster manager.
The next field is heartbeat retries 110, which is the number of consecutive heartbeats sent to the component management system that must be lost before the cluster manager considers the client component management system to be lost. Is. The last field 120 tells the system what should be issued when the number of heartbeat retries is reached.

This system of managing components, nodes and clusters using a common database of information that can be duplicated and reside on multiple networks allows for effective management of multiple systems. The system can exhibit high availability with a minimum amount of downtime.

Although the present invention has been described in detail with reference to certain preferred embodiments, other embodiments are possible. Therefore, the spirit and scope of the present invention should not be limited to the description of the preferred embodiments contained herein.

[Brief description of drawings]

FIG. 1 is a flow chart showing the manner in which a component management database is used.

FIG. 2 illustrates how component management instructions interface with an operating system within a web server.

[Figure 3]   FIG. 6 illustrates a state machine that responds to insertion or extraction events.

[Figure 4]   It is a figure which shows the display mode of the information from a component management database.

[Figure 5]   It is a directed graph using critical dependencies and non-critical dependencies.

FIG. 6 illustrates how information from a component management database can be used to generate events.

─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Vogel, John             United States 98040 Washington             Martha Island 9th Firth             Toave South East 4227 F term (reference) 5B042 GA10 GA12 JJ16 KK04

Claims (12)

[Claims] 1. A method of maximizing the availability of a computer controlled system including both hardware and software components, the method comprising: a) receiving a message representative of the function of the component and the interdependence of the component. B) holding the database of functions and dependencies with the received message, and c) one or more event notifications regarding events affecting the availability of the system for successful execution. And d) managing interoperability of the components using methods that are individually determined by the type of managed component, thereby maximizing the availability of the system. The steps to achieve, and , Said method. 2. The method of claim 1, wherein the event affecting availability of the system for successful execution comprises extraction and insertion of peripheral hardware components. 3. The method of claim 2, wherein the peripheral hardware component comprises a network interface card. 4. The method of claim 1, wherein the dependency database is set during initialization of the system. 5. The method of claim 1, wherein the event affecting the availability of the system for successful execution comprises component degradation or component failure. 6. A method for maximizing the availability of a plurality of computer systems, including both hardware and software components, linked together, comprising: a) a message representative of the functionality of said components and their interdependencies. B) maintaining a database of the functions and dependencies with the received message; and c) issuing an event notification regarding an event affecting the availability of the system for successful execution. Receiving from one or more of said components, and d) managing the interoperability of said components using methods determined individually by the type of managed component, thereby maximizing the availability of said system. The steps to reach the goal Including the method. 7. The method of claim 6, wherein the event affecting the availability of the system for successful execution comprises extraction and insertion of peripheral hardware components. 8. The method of claim 6, wherein the peripheral hardware component comprises a network interface card. 9. The method of claim 6, wherein the dependency database is set during initialization of the system and the database is replicated and persisted across the plurality of computer systems. 10. The event affecting availability of the system for successful execution includes component degradation or component failure.
The method of claim 6. 11. A system that includes both hardware and software components and maximizes the availability of a plurality of computer systems linked together, comprising: a) a message representing the functionality of the components and the interdependence of the components. A) a dynamic configuration manager that receives a message, b) a distributed message service that incorporates the function of passing a message through a packet filtering and proxy firewall, and c) a component management database that holds a table of the function and the dependency using the message. , D) a set of methods for managing interoperability of said components, each determined individually by the type of component being managed, whereby said system To achieve the goal of maximizing the availability includes the set, the said system. 12. A system that includes both hardware and software components and maximizes the availability of a plurality of computer systems linked together, comprising: a) the types of components that make up the system and each component Means for configuring a database of management information including an assumed role played in the system, an actual role played by the component in the system, and the mutual relationship of each component with each other; and b) the software component and hardware. Means for setting a method for monitoring and controlling a wear component, the method being tailored to a particular class of the target component; and c) invoking the appropriate method so that the component Existence Means for receiving messages and notifications from the component so as to maximize availability of the system, and d) means for replicating data present in the database of management information and persisting it both over the system and the network. The system, including: