Electrical computers and digital processing systems: multicomput – Distributed data processing – Processing agent
Reexamination Certificate
2000-12-07
2004-09-07
Luu, Le Hien (Department: 2152)
Electrical computers and digital processing systems: multicomput
Distributed data processing
Processing agent
C709S223000
Reexamination Certificate
active
06789101
ABSTRACT:
PRIOR FOREIGN APPLICATION
This application claims priority from German patent application number 19959181.4, filed Dec. 8, 1999, which is hereby incorporated herein by reference in its entirety.
TECHNICAL FIELD
The invention describes a system and method for automation of a computer network, in particular for automatic starting and stopping of resources in the computer network taking account of their start dependencies.
BACKGROUND ART
The interaction between various IT resources in connection with non-determinable requests and events forces the operators of computer centers to monitor, and where necessary correct, the behavior of the said systems. This activity is usually termed Operations Management. Automation as a subsidiary discipline concerns itself with the task of simplifying Operations Management.
In this, an attempt is made to analyze autonomously occurring events and to respond in accordance with an instruction. This instruction is customer-specific, since automation is not able to make the necessary decisions a priori.
A key aspect of automation is that the availability and delivery of IT resources depends on certain states of other IT resources. A specific computer center service (=an application) cannot be made available before all the necessary IT resources for it have been started. An application of this kind typically comprises several programs on different computers. These programs require data which are stored on various storage media. Very often it is also necessary to provide a comprehensive network topology which permits the individual components to communicate both internally and externally.
A further key aspect of automation relates to the fact that IT resources are required by several different applications simultaneously (shared resources). For example, a disk will typically store data for more than one application. Database software can be used for payroll accounting as well as for the product planning system (PPS), and networks are not limited to single applications.
The management of these structures with the goal of delivering and maintaining applications for a specific period of time (service window) is fundamentally supported by automation functions.
The fact that nowadays applications are distributed across several different systems and computers have been interconnected to form complex structures has made automation substantially more difficult.
In a large S/390 Sysplex network, for example, 5,000 programs can be launched in parallel. Application programs are distributed in several instances across computers in the Sysplex network. The workload is distributed across several computers, resulting in a higher throughput, and in the event of a total failure of a computer applications can continue running at reduced capacity on the remaining computers.
The dependencies between the individual IT resources are not dealt with adequately at present. For example, resources can only be started if other resources are already active, or should only be started when another resource has been stopped. The dependencies also apply to the stopping of IT resources.
Likewise, a resource could be started but could not fully perform its service without a specific resource being active/inactive.
The basic object of automation is to deliver or to terminate an application. When this object is fulfilled, automation attempts to maintain the attained state until a new goal is transmitted.
A change in goal may have many causes:
The operator wants to change.
A service window has expired or begun (driven by a calendar).
The system wants applications to be relocated or additional instances to be created as a result of capacity bottlenecks.
Such goals may be mutually contradictory and have differing levels of importance. At present automation functions barely respond to the demands of that rule. Each automation order is executed unconditionally, with no account taken of the preceding activities and without truly understanding why a resource is in its current state. As a result, it may occur that the operator stops an IT resource without knowing that the resource is still needed by another application.
Maintenance work is being carried out on a resource, for example. The allotted time for the work is not sufficient. In this case especially, the beginning of a service window must not result in the resource starting.
The automation described above is at present restricted to a single computer. Consequently, only dependencies between IT resources belonging to that computer can be defined. Typically they are programs which can be run on the computer on which the automation software is also active. Although today a large number of resources of a company are accessible over networks on different computers, a local automation software per se cannot centrally automate all remote IT resources because
the data exchange to monitor an automation process could overload the network (message traffic);
failure of the centralized automation software would result in the total failure of an entire computer network;
the number of resources needing to be automated simultaneously could result in bottlenecks in the automation software itself.
The limitation in force to date has meant that distributed applications could in no way be automated in their entirety. Rather, the automation software itself must attempt, as a client/server topology, to run processes directly on each computer locally and only escalate to a remote next highest instance when required.
This is illustrated by the following example (see FIG.
1
), wherein a Web Server application requires the following three start-dependent programs:
1. Network (for example TCP/IP)
2. Database software (for example IBM DB2)
3. Web Browser software
For capacity reasons, each program is started on a different computer. However, the Web Browser cannot be started until TCP/IP and DB2 are both active. This behavior is dependent on the implementation of the Web Browser software. Ideally, this software would simply wait until its two partners are active. But this is not always the case: it can occur that a Web Server which has started too early simply terminates again.
Starting of this application is not fully automatable at present. The process must run as follows:
1. Start TCP/IP on system
1
and DB2 on system
3
. This can be done by the automation functions of the respective systems.
2. Monitor both processes until both are active. Automatic launch of the Web Browser cannot be handled because of the limitation described here.
3. Start Web Browser.
In order to launch the overall application three orders are actually required to be entered by the operator. Experienced operators launch the Web Browser first. Assuming the automation is now capable of launching the start-dependent resources (TCP/IP and DB2) first, the overall application could be activated in this way. That works well until there are applications in which there are resources with no start relationships. Of course this component, too, can be entered somewhere in the start dependencies, but with the result that start processes are serialized.
Resources serve several applications (shared resources) and comprise several components (resource components) which can run on different computers. In the OS/390 Sysplex these components can run multiply on different computers (resource instances). This provides a better throughput, and a greater availability in the event of complete computer failures.
The automation can meet the preconditions before the desired resource is handled.
This solution does not support the following valid situations:
Resources in a start relationship do not necessarily have to be in a stop relationship in the reverse sequence. Typically, resources are more often in a start relationship than a stop relationship. With the present-day concept, when an application is terminated components are unnecessarily serialized, even though the computer capacities would permit more parallelism. The overall process takes longer.
Starting a resource should also trigger termination of
Clarke Michael
Schneider Juergen
Schueppen Walter
Terekhov Alexander
Heslin Rothenberg Farley & & Mesiti P.C.
International Business Machines - Corporation
Kinnaman, Jr. Esq. William A.
Luu Le Hien
Schiller, Esq. Blanche E.
LandOfFree
Automation system uses resource manager and resource agents... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Automation system uses resource manager and resource agents..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Automation system uses resource manager and resource agents... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3209815