Performance Problems with iSeries Applications
- 1. Eliminate the Confusion Eliminate the Confusion!
- 2. Squeezing Bowling Ball down Garden Hose Optimize & Reduce Workload!
- 3. Different Approach Different Approach…
- 4. Interactive Response Time 69% - Interactive response time is usually OK, but random slowdowns occur. Can’t find cause – solution appears to be a $100,000 interactive feature card.
- 5. Daily GANTT Chart - Interactive
- 6. CPU (seconds) by Subsystem
- 7. Number of I/Os by Subsystem
- 8. Skipped Records by Job/User
- 9. Skipped Records by File/Fields
- 10. Long Running Batch Jobs 57% - CPU used less than 50%, but batch jobs take a long time & don’t use CPU efficiently. Have been told that a bigger system with more CPU is needed.
- 11. Daily GANTT Chart – Long Batch
- 12. Temporary Disk Space Jumps 48% - Disk space jumps from 70 to 90%, could be from temp space used by jobs but have not been able to prove which job is doing it.
- 13. Disk Navigator – Size of QTEMP by Job
- 14. Long Nightly Processing 44% - Night processing runs long, one CL program calls 100s of other programs and it takes 6 hrs to run.
- 15. Job Overview – One Process
- 16. Job Overview – Several Processes
- 17. Application Analyzer – By Program Function
- 18. Application Analyzer – By OS/400 Function
- 19. Application Analyzer – Open/Close
- 20. Java & JDBC Killing System 43% - New web app uses Java & JDBC, when web lookup pages are hit, user waits 2 minutes. System was fine until new app went live.
- 23. Waiting for Pick Slips 43% - Overall CPU is only 30% used, yet users wait 6 minutes for pick slips to print. When batched, it takes 45 minutes to start printing, which delays product shipping.
- 24. CPU 30% Used, Jobs in Queue
- 25. # of Jobs by Job and User
- 26. Jobs Waiting in Job Queues
- 27. Waiting 250 Minutes, Runs 1 Second
- 28. Job Logs Wasting Resources 40% - 500,000 pages of job logs generated every day, logs rarely used, wasting disk space, CPU & I/O.
- 29. # of Pages by Queue, Job & User
- 30. Job Description & End Code
- 31. Message Logging
- 32. Job Runs 30,000 Times 38% - Our batch job runs 30,000 times per day, consumes 1 CPU second each time it runs. Ends up being 30% of CPU consumed, constantly spikes system all day.
- 33. CPU by Job Name
- 34. # Jobs by Job Name
- 35. HA Can’t Keep Up 30% - Journal receivers are 10 GB per day, HA can’t keep up.
- 36. Journal Analyzer
- 37. Q&A To receive MB Software’s FREE performance evaluation: 1. Register at http://www.mb-software.com 2. Download and install software on AS/400 3. Call (800) 713-5034 to schedule your free performance review session What To Expect : After allowing performance data to accumulate for one full business day, you will log into a secure area of MB Software’s Web site and have a conference call to discuss our findings and how you can gain performance improvements. It’s that easy! Questions & Answers