Performance improvements have now been delivered in the latest CICS TS V5.3 open beta, here is a quick summary of the function recently delivered
Operational efficiency has been a key focus for CICS TS Version 5 and continues to evolve in the CICS TS V5.3 open beta, where there are many improvements worth knowing about. This time around we have targeted a reduction in the CPU usage of CICS routing and data owning regions, to help you continue to lower the overheads of running work in CICS TS.
Given the different ways you are able to drive work into CICS we have focused our efforts between general improvements that help all workloads, and some specific areas that have the biggest impact. From conversations with customers and survey information we found that web and web services based workloads, namely workloads that handle requests using the EXEC CICS WEB API or those that use the CICS pipeline processing to transform requests into the language structures that programs need, are both popular and growing ways of driving work into CICS. Therefore improving the efficiency of these areas will have a beneficial impact for a broad range of users.
How have we done this?
To gain improvements that affect all workloads there have been some internal code algorithm improvements, exploitation of hardware instructions, and control block alignments to improve cache usage. Two areas where these changes have had a beneficial impact are when trace and or monitoring are active in a CICS region. Also, for many workloads that use MRO, the algorithm used when a large number of sessions has been defined has been made more efficient.
For web and web services workloads we looked at the communications layer to determine how things could be made more efficient. A key optimization in this area is that the web attach transaction, CWXN, is no longer needed in many scenarios. Not running the web attach transaction results in reduced CPU usage as well as fewer monitoring records being written to SMF.
For external web connections security is vital and so we spent some effort in this area too. As of CICS TS V5.3 open beta the SSL processing in CICS switches TCB fewer times during SSL handshakes thereby saving CPU. An alternative security set up is to use the AT-TLS support in IBM z/OS Communications Server to provide SSL support. This configuration allows CICS TS to be AT-TLS aware when setting up a transaction’s security context and our testing of this has shown additional reductions in CPU usage per request.
Give it a go!
So for those that try out the CICS TS V5.3 open beta, they are likely to find that the address space monitoring records for their CICS routing regions show them using less CPU, and that benefits can be seen in other types of CICS region too.