IBM constantly improves z/OS Workload Management both from a functional as well as a usability point of view. The following gives a short overview of past enhancements:
WLM/SRM enhancements in z/OS V2.1
- z/Enterprise EC12 Support Enhancements:
Relaxed ZAAP-ZIIP requirements: SRM changed to relief requirements for the IEAOPTxx ZAAPZIIP (“zAAP on zIIP”) option:
New functions for hard and soft capping:
- No longer limits the zAAP on zIIP function based on the number of zAAPs and/or the number of zIIPs installed on the machine.
- The zAAP on zIIP function continues to be limited to LPARs that have no zAAP. With this information users can, for example, set a goal, based on a required response time.
- Smoother capping with WLM managed softcapping.
- When IRD weight management is active the group capacity of an LPAR may be derived by the initial weight.
- New “Absolute Capping Limit” LPAR control.
- New Classification Qualifiers and Groups:
New classification group types.
Some new and modified work qualifier types for use in classification rules in the WLM service definition.
- I/O Priority Groups:
Important service classes which are sensitive to I/O delay can now be assigned to priority group HIGH which ensures that they get always higher I/O priorities than the service classes assigned to group NORMAL.
- Other z/OS V2.1 Enhancements:
Improved granularity for resource groups.
3000 Application Environments.
Back to top
WLM/SRM enhancements in z/OS V1.13
- Enhanced reporting:
WLM reporting now provides a response time distribution for all service classes - also for service classes with an execution velocity or discretionary goal. For example, an RMF user can retrieve the response time for service class periods with velocity and discretionary goals.
With this information users can, for example, set a goal, based on a required response time.
- Enhanced processing of enclave tasks:
WLM allows enclave tasks that have subtasks that were implicitly added to that enclave to leave the enclave.
WLM also allows tasks that have non-enclave subtasks to join an enclave and have the subtasks implicitly joined to that enclave as well.
- CICS response time management enhancement:
Adjust WLM management of CICS/IMS regions to a Work Manager/Consumer model.
- Support for SLEDrunner z:
Optimized processing of IO requests in a DS8700 storage server by consideration of the WLM importance and goal fulfillment of the related work.
Back to top
WLM/SRM enhancements in z/OS V1.12
- Enhanced performance management:
Non-enclave work of queue servers and enclave servers is now performance managed towards the first service class period of the goal of the address space. The IEAOPT parameter ManageNonEnclaveWork allows to switch on or off the management of non-enclave transaction work.
- Enhanced WLM administration:
z/OS Management Facility (z/OSMF) provides a browser-based user interface for managing WLM service definitions. With the Workload management task of z/OSMF you can define, modify, view, copy, import, export, and print WLM service definitions with improved ease-of-use.
- Managing performance with the guest platform management provider (GPMP):
Starting with IBM zEnterprise System (zEnterprise), the Guest Platform Management Provider (GPMP) is the interface between the intranode management network (INMN) and the z/OS Workload Manager (WLM). It provides policy information to WLM about the platform wide performance goals of the workloads in which the z/OS system is participating. You can manage and administer the guest platform management provider with WLM. For example, you can start and stop the guest platform management provider, or display status information.
- Improved batch management:
WLM takes the resource group maximum into account in order to avoid that the projected increase in service demand exceeds the resource group maximum. When a service is already capped due to the resource group maximum, then no more initiators are started.
Back to top
WLM/SRM enhancements in z/OS V1.11
- Enhanced routing recommendations:
The WLM IWM4SRSC and IWMSRSRS services are enhanced to return detailed results and finetune the recommendations for routing work between servers in a sysplex.
- WLM supports classification of Tivoli Directory Server work:
With the new WLM subsystem type LDAP supported, Tivoli Directory Server can now classify its work and allow end users to set performance goals for different kinds of LDAP work.
- Maximum number of instances per WLM server address space extended:
The limit of server instances (the maximum number was 1000) per server address space is now removed.
- WLM supports more detailed reporting:
The number of report classes allowed in a WLM service definition is extended from 999 to 2047 which allows for more detailed reporting.
- HiperDispatch mode improved for zIIPs:
The HiperDispatch function now supports dispatching to multiple zIIP affinity nodes.
- REQLPDAT callable from unauthorized users:
The SYSEVENT REQLPDAT that provides system capacity data can now be invoked unauthorized thus improving the ease of use for the customer.
Back to top
WLM/SRM enhancements in z/OS V1.10
- Better protecting z/OS against high consumers of fixed and auxillary storage:
In order to avoid storage shortages, detection mechanisms have been added to identify a sudden growth in fixed and pageable storage. This allows to react early enough to avoid system outages. The enhancement:
New messages and automation methods allow the system to terminate address spaces with an unacceptable consumption of high fixed or pageable storage. Before termination, address spaces get logically swapped and non-swappable address spaces will be set to non-dispatchable so that they will not allocate additional storage.
- Identifies fixed and auxillary storage shortages.
- Identifies the address spaces which show the highest growth in the respective storage consumption.
- Helps the system to recover.
- Supporting enhanced ARM transaction management:
The services arm_start_transaction and arm_stop_transaction indicate the beginning and the end of an ARM transaction. Beginning with z/OS V1R10, ARM services are enhanced to use enclaves so that managing individual transactions is possible. users to set performance goals for different kinds of LDAP work.
- Managing selected components in service class SYSTEM:
To make sure that the system runs smoothly, the following address spaces are now managed in service class SYSTEM: XCFAS, GRS, SMSPDSE, SMSPDSE1, CONSOLE, IEFSCHAS, IXGLOGR, SMF, CATALOG. The *MASTER* and WLM address spaces are already managed in service class SYSTEM.
- New performance block delays:
To increase the integration of enclave exploitation in WebSphere Application Server for z/OS, a new set of performance block states is introduced. These states can be named by each application, thus allowing for more flexibility in reporting application states.
- Processing in HiperDispatch mode:
HiperDispatch mode manages the assignment of work to processors. This dispatching mode takes the book structure of the machine into account and re-dispatches a unit of work to the same processor repeatedly so that cached data continues to be available to the processor, improving performance.
Back to top
WLM/SRM enhancements in z/OS V1.9
- EWLM enhancement for asynchronous messaging:
To enable EWLM reporting for applications based on an asynchronous messaging model, the Application Response Measurement (ARM) C language interfaces are enhanced by the ARM 4.1 extensions defined by the OpenGroup. Similar extensions are provided for the WLM Execution Delay Monitoring Services.
- Improving termination of canceled jobs:
WLM makes sure that an address space being canceled gets enough CPU resources to terminate more quickly. This can eliminate the need to reset the priority of a canceled job, task, or user to speed address space termination when resolving resource contention issues.
- Promoting blocked workloads:
If the CPU utilization of a system is at 100%, workloads with low importance (low dispatch priority) might not get dispatched anymore. This could cause problems if the work holds a resource and by that holds up more important workloads. Therefore, any address space or enclave which has ready-to-run work units (TCBs or SRBs) but does not get CPU service within a certain time interval due to its low dispatch priority will be temporarily promoted to a higher dispatch priority. Address spaces which are swapped out are not considered for promotion. RMF supports this function by reporting relevant measurements.
- Advanced routing support:
WLM's routing services are enhanced to recognize the zAAP and the zIIP capacity of a central processor. This will assist TCPIP, DB2, and WebSphere in their decisions to route work to systems in a sysplex.
- Starting minimum number of server regions without delay:
Enhancements to a WLM service allow the application to control whether WLM should start server regions in parallel or sequentially. This will assist exploiters of application environments (for example, WebSphere) to shorten the start up time of WLM-managed servers.
Back to top
WLM/SRM enhancements in z/OS V1.8
- Managing the size of DB2 bufferpools:
WLM introduces a new set of interfaces to DB2 that assist DB2 in managing the size of their bufferpools. WLM dynamic bufferpool management allows WLM to collect data on how efficiently the bufferpools are being used and to tell DB2 to increase or decrease the size of their bufferpools.
As a result, there is a new subsystem wait state BPMI available which indicates that the work manager is waiting for buffer pool I/O. RMF indicates this new wait state in the subsystems state samples breakdown of the Postprocessor WLMGL report and in the Response Time Breakdown of the Monitor III SYSWKM report.
- Enhancements for routing services by accounting for abnormal conditions of servers:
This enhanced zAAP support provides CPU management for work using zAAP processing resources. This is a step towards fully integrating zAAPs as a processing resource for Java-related work at reduced costs.
- System z9 Integrated Information Processor (zIIP) support:
WLM supports the System z9 Integrated Information Processor (zIIP) -- a new type of processor for a dedicated workload. This allows you to reduce costs by offloading certain workloads, for example, DB2 tasks, from general purpose processors to zIIPs.
- EWLM/WLM service class correlation:
This enhancement allows you to associate EWLM transaction classes with local z/OS WLM service classes. By doing so, the performance administrator at the EWLM control center has access to a report that integrates the end-to-end work within the same EWLM transaction class for reporting and management purposes.
- WLM enhancement of enclave services:
This enhancement improves the performance of WLM enclave services.
Back to top
WLM/SRM enhancements in z/OS V1.7
- Support of fast processors:
This enhancement enables the SRM/WLM internal timing to match faster processors. It allows z/OS to change the speed of the processors to increase or decrease the system performance. z/OS can now recognize such processor speed changes and will adapt all speed and timing related internal calculations dynamically without requiring an IPL.
- Workload balancing:
The new server-specific workload balancing services provide a more granular view on the performance of a server in a sysplex. For its workload balancing recommendations, WLM now takes the Service Level Agreement in the WLM policy into account.
- Support for multiple subchannel sets:
WLM supports multiple subchannel sets on zSeries servers that provide support for multiple subchannel sets.
- Support for DB2 latch contention:
The 'short time promotion' is an enhancement of the enqueue management function. The enhancement makes DB2 latch contention run smoother and reduces bookkeeping. The new support allows for the DB2 resource manager to resolve DB2 latch contention by consuming the complete short promotion time without issuing an ENQRLSE when the contention ends.
Back to top