|  |
Document Author:
|
Liz Holland
|
|
Document ID: |
TD103548 |
|
Doc. Organization: |
Techline Americas |
|
Document Revised: |
02/05/2010 |
|
Product(s) covered: |
DB2 for z/OS and OS/390; IBM eServer zSeries; WebSphere Application Server for z/OS; z/OS; z/OS.e |
|
|
|
|
|
|
|
| Abstract: PROJECTCPU can aid in determining the number of zAAP and zIIP engines required to satisfy a specific customer's usage. This parmlib option is available with the correct level of software. The PROJECTCPU=YES parameter enables z/OS to collect usage as if a specialty engine was installed, when the target workload is being run. This projection capability can be run at any time, including on a production environment. |
The first specialty engine for the mainframe was the Internal Coupling Facility (ICF), introduced in 1997. This provided a hardware capability required by sysplex, while avoiding the cost of an external box. More specialty engines have been developed that allow customers to increase the computing power of their zSeries, z9 and z10 systems, while avoiding increased software licensing costs.
The ability to do accurate capacity planning for the specialty engines' target workloads has evolved over time. With the release of z/OS 1.8 and its associated level of RMF, there have been significant improvements in both the accuracy and simplification of data collection. z/OS 1.8 introduces a new keyword in SYS1.PARMLIB, and RMF reports have a new field that show the projected CPU busy of an engine that is not yet installed on the machine.
It is important to be clear about what the term 'target workloads' means. Each specialty engine was designed to dispatch a particular type of software. IFL engines were the second specialty engine deployed. They are for z/VM or Linux workloads, but are not part of this capacity discussion.
zAAP engines were the next addition, and were originally targeted to dispatch only Java workloads.
zIIP engines are designed to support workloads that use enclave SRBs. DB2 version 8, XML parsing, and IPSEC translation are some exploiters of zIIP engines.
zAAP engines have been assigned more target workloads, so some XML parsing will run there if the invoker is in TCB mode.
Prior to z/OS 1.8 and DB2 Version 8, collecting data for a capacity planning study could involve making changes to each JVM that was running, and using various tools provided for DB2. Enhancements to z/OS and RMF mean it is no longer necessary to use product level tools for these workloads, but only to collect SMF data. If new workloads are enabled, there may be specific techniques for collecting that data.
The following is a checklist of items to complete to be sure you can take advantage of the simplification for DB2, Java and XML.
Install z/OS 1.8 or later.
Install enabling code on z/OS 1.6 and z/OS 1.7. The enabling code is
FMID JBB77S9 for z/OS 1.6 and JBB772S for z/OS 1.7. Please search in IBMLink on the keyword "zIIP/K" for the latest maintenance information.
- In SYS1.PARMLIB member IEAOPT00, code:
PROJECTCPU=YES (Please see the note at the end of the document for zAAP on zIIP cases.)
Activate the new setting with the operator command: SET OPT=00
- No changes to JVM required to project zAAP usage through SMF data.
- DB2 Version 8 or higher IS required to project zIIP usage through SMF data.
- Contact IBM at the capacity planning portal to request a complete sizing study. http://w3-03.ibm.com/support/americas/techline/sizewise.html
- Examine data in RMF reports. Use RMF report options:
SYSRPTS(WLMGL(SYSNAM(<system_name>)))
SYSRPTS(WLMGL(POLICY))
This will yield a single Workload Activity Report for the period. It will consolidate all work in the designated system and allow you to easily estimate the zAAP and/or zIIP workload across all servers on the z/OS image. Following are three examples of Workload Activity Reports.
Example 1)
In this RMF report, APPL% AAPCP is the field that reports projected zAAP consumption. APPL% IIPCP is the field that reports projected zIIP consumption.
This report shows that the DDFPOS workload uses 132.10% of a single CPU (The system is a 4-way)
It shows that 73.90% of a CPU could have been running on a zIIP processor. No zAAP or zIIP engines are installed.
Example 2)
This machine has one zAAP engine installed. At the time that it is running 101.96% busy (an indication of more work in queue), there is enough work to run another zAAP engine 54.97% busy, IF a second zAAP engine WERE installed. (AAPCP)
Example 3)
This final example is a little more difficult to explain. There are 5 zAAP engines installed, and they are not at 100% busy. Still, some work that is zAAP eligible has gone to a general purpose engine. (AAPCP is 2.04%) What would be the reason for this? And, is it a concern?
These numbers indicate more requests have come in at one time than there are zAAP processors to service the requests. This snapshot of the arrival rate is interesting, but since the overall system is lightly loaded, there is no reason to think another zAAP engine (or any other kind of engine) is needed.
Additional information for zAAP on zIIP
1. If the RESERVE field in the PR/SM LP Image Profile Processor definition screen is blank for zIIP/zAAP then you need to specify PROJECTCPU=YES in IEAOPTxx
2. If the RESERVE field in the PR/SM LP Image Profile Processor definition screen is not blank for zIIP/zAAP then you will always get projection, even if PROJECTCPU=NO is specified in IEAOPTxx
3. If you cannot specify RESERVE (machines older than z9) then you need to specify PROJECTCPU=YES in IEAOPTxx
|
Classification: |
Hardware |
Category: |
Performance |
|
|
|
|
Platform(s): |
IBM System z Family |
S/W Pillar(s): |
Information Management; WebSphere; System z |
O/S: |
z/OS |
Keywords: |
AAPCP, IIPCP, projectcpu |
|
|
| Is this your first visit to Techdocs (the Technical Sales Library)? |
 |
Learn more |
|
|
|
