Splitting online systems to improve availability

Splitting the CICS® system into two or more separate address spaces may lead to improved availability. If CICS failures are being caused by application program errors, for example, separating out the failing application can improve overall availability. This can also give virtual storage gains and, in addition, can allow you to use multiprocessors and MVS™ images more efficiently. See Splitting online systems: virtual storage for more information. A fuller account can be found in the System/390® MVS Sysplex Application Migration Guide (GC28-1211).

The availability of the overall system may be improved by splitting the system because the effects of a failure can be limited or the time to recover from the failure can be reduced.

The main ways of splitting a system for availability are to have:

Having only one FOR in a system, or logical subset of a system, can reduce the operational difficulties of restarting a system. It is possible to split the regions in different ways to those described so far, by having many regions all of which own some terminals, some applications, and some files and databases. This type of splitting is very complex to maintain and operate, and also needs careful monitoring to ensure that the performance of the overall system is optimal. For these reasons, a structured approach with each of the regions having a clearly defined set of one type of resource is recommended.

Limitations

Splitting a CICS system requires increased real storage, increased processor cycles, and extensive planning. These overheads are described in more detail in Splitting online systems: virtual storage.

Recommendations

If availability of your system is an important requirement, both splitting systems and the use of XRF should be considered. The use of XRF can complement the splitting of systems by automating the recovery of the components.

When splitting your system, you should try to separate the sources of failure so that as much of the rest of the system as possible is protected against their failure, and remains available for use. Critical components should be backed up, or configured so that service can be restored with minimum delay. Since the advantages of splitting regions for availability can be compromised if the queueing of requests for remote regions is not controlled, you should also review Managing queues for intersystems sessions.

Related tasks
Operating CICS in a multiregion environmentMVS and DASD: improving performance
Reducing MVS common system area requirements
Making CICS nonswappable
Increasing the CICS region size
Using job initiators
Tuning the region exit interval (ICV)
Using LLA (MVS library lookaside)
DASD tuning

Related concepts
Multiregion operation
Introducing CICSPlex® SM
[[ Contents Previous Page | Next Page Index ]]