In between I wrote 1,100 blog entries. I hope, I posted some interesting topics. Please let me know, if there is anything that I should address.
In this blog entry I summarized my recommendations for performance comparisons, that you may consider before migrating to new hardware - be it a new processor or new devices. It's always good to have history data available in case of problems.
If you have that material, it is much easier to compare the performance characteristics old versus the new environment.
See some recommendations below:
- Look for a repeatable workload, that processes similar amount of data.
- Collect reference data:
- Measure and collect most important workload parameters , e.g. performance, I/O activity / timings, CPU / elapsed time, console logs, etc.
- Apply required PTFs for the new processor or devices (see PSP buckets)
- Do not change the processor and software at the same time. If possible, delay changes to hardware infrastructure (controller, devices, network, ...)
- E.g. if you upgrade to a new processor, verify the performance characteristics on the new environment before you continue with the software upgrade, best in production. If everything is OK, continue with other upgrades.
- We have a white paper online that describes tasks for hardware and software upgrades: z/VSE release & hardware upgrade - here.
- Hardware related information is also on the z/VSE hardware and release status PDF - here.
Collection of workload parameters:
- performance reports from performance monitors provided by vendors
- console logs
- network data
- z/VSE console commands
- Commands to get additional information
- SIR - to get general information about the system
- SIR SMF - to gather I/O information per device (I/O-count, timings - queued, connect, msecs/ I/O)
- SIR MON - to retrieve workload counters (SVCs, bound conditions, some internal counters, ...)
- SYSDEF TD,RESETCNT - to reset counters
- QUERY TD - to retrieve used CPU time, CPU utilization, non-parallel share, elapsed time, ...
- Some of those commands are described in the z/VSE Hints and Tips book - here.
To use above commands for measurements, you may use the following sequence:
- SIR MON=ON -starts monitoring (system services)
- SIR SMF=ON - starts monitoring for I/O related counters
- SYSDEF TD,RESETCNT - resets counters
- <monitor interval - e.g. 1 or more hour at peak>
- SIR MON=OFF - stops monitoring
- QUERY TD - displays CPU counters
- SIR MON - displays SVC counters
- SIR SMF. - displays I/O counters
- To start next interval, begin with 1.
- when finished: SIR SMF=OFF
z/VSE performance tools:
- z/VSE CPUMON (CPU Monitor) tool - good tool to monitor the CPU utilisation:
- is based on SIR MON command and QUERY TD to get CPU utilization of a workload, may run days
- measures in user defined intervals
- monitored data can be loaded into a spreadsheet. Generate a line chart. It shows workload peaks.
- z/VSE Health Checker tool - to retrieve, display, and analyze performance relevant data from a z/VSE system.
- The tools can be downloaded from here.
If you migrate from ECKD to SCSI devices, our measurements showed that the I/O may be faster to a SCSI disk, but it may require a bit more CPU time because of the emulation FBA to SCSI. We have SCSI white paper too - z/VSE SCSI Support and Migration Options
Have a good weekend.