CA7 Workload Automation: Improving performance in job submission between Pred and Suc
Article ID: 132527
Updated On:
Products
CA 7 Workload Automation
Issue/Introduction
CA Workload Automation CA 7® Edition tracks the progress of jobs on the operating system using SMF for z/OS work and pseudo SMF for cross-platform work. Records are generated for the job start, step end, data set create, and job end. ICOM is the facility that picks up the feedback and communicates it to CA7ONL.
By default, all ICOMs write SMF feedback to the communications data set (COMMDS) .
Sometimes, base on changes done to production jcls loaded in CA7 DB (for example moving from mono-step jobs to multi-steps jobs) , it could happen that CA7ONL main task takes too many seconds completing a job and submitting the successor in the chain.
How is it possible to reduce this interval time, improving the product submission and tracking performance?
By default, all ICOMs write SMF feedback to the communications data set (COMMDS) .
Sometimes, base on changes done to production jcls loaded in CA7 DB (for example moving from mono-step jobs to multi-steps jobs) , it could happen that CA7ONL main task takes too many seconds completing a job and submitting the successor in the chain.
How is it possible to reduce this interval time, improving the product submission and tracking performance?
Environment
Z/OS - Workload Automation CA7 Edition
Resolution
One thing that could help improving the product submission and tracking performance, is to use the X=',XCF=NOTIFY' ICOM parm.
The PARM= data for ICOM can use a keyword of X= , and using XCF=NOTIFY could help this condition.
Having short running jobs that use trigger streams means that the "next" job will wait for the completion of the previous job. The completion data is gotten using SMF data, and this is "sent back" by ICOM(s). There can be a few seconds between ICOM writing the data to the COMMDS and CA-7 knowing the data is there to be processed. Using the NOTIFY option lets CA-7 know (very quick) there is data there without CA-7 checking every few seconds to see if the data is there.
The Workload Automation documentation (System Guide and User Guide) provides details about how it is possible to activate this ICOM parm/feature
Here a quick summary:
1. Allocate and format the XCFDS dataset (each ICOM its own dataset).
2. Allocate and format the XCFCKPT dataset.
3. Change the CA7ONL init parm SMF setting “SMFXCF=YES”.
4. Change the ICOM task to start with X=',XCF=NOTIFY'.
5. Change CA7ONL and ICOM tasks, to insert the above datasets.
6. Restart all involved tasks ( CA7ONLY TYPE=ERST restart is OK).
The PARM= data for ICOM can use a keyword of X= , and using XCF=NOTIFY could help this condition.
Having short running jobs that use trigger streams means that the "next" job will wait for the completion of the previous job. The completion data is gotten using SMF data, and this is "sent back" by ICOM(s). There can be a few seconds between ICOM writing the data to the COMMDS and CA-7 knowing the data is there to be processed. Using the NOTIFY option lets CA-7 know (very quick) there is data there without CA-7 checking every few seconds to see if the data is there.
The Workload Automation documentation (System Guide and User Guide) provides details about how it is possible to activate this ICOM parm/feature
Here a quick summary:
1. Allocate and format the XCFDS dataset (each ICOM its own dataset).
2. Allocate and format the XCFCKPT dataset.
3. Change the CA7ONL init parm SMF setting “SMFXCF=YES”.
4. Change the ICOM task to start with X=',XCF=NOTIFY'.
5. Change CA7ONL and ICOM tasks, to insert the above datasets.
6. Restart all involved tasks ( CA7ONLY TYPE=ERST restart is OK).
Feedback
Yes
No
Powered by
