IBM zVM Module 9 Performance 2004 IBM Corporation

IBM ^ z/VM Module 9: Performance © 2004 IBM Corporation

IBM ^ Objectives § Describe what performance means in the z/VM environment § List and explain the performance characteristics of the z/VM system § Explain scheduling and dispatching and how they relate to performance § Describe three lists used in scheduling and dispatching: 4 Dormant list 4 Eligible list 4 Dispatch list © 2004 IBM Corporation

IBM ^ Objectives continued § Describe the planning and administration guidelines for improving § § z/VM system efficiency Explain the areas of performance that need to be considered and the major factors affecting them List and describe the CP performance facilities Describe performance guidelines in a virtual machine environment Explain the performance monitoring and analysis programs that are available to tune your z/VM system © 2004 IBM Corporation

IBM ^ Objectives continued § Describe how the INDICATE and MONITOR commands are used in performance testing § Explain why you would tune your z/VM system and the methods for doing so § List and describe the system resources that are capable of being tuned: 4 Processor subsystem 4 Paging subsystem 4 Storage subsystem 4 I/O subsystem © 2004 IBM Corporation

IBM ^ Introducing Performance § Achieving optimum system performance is an elusive goal. § This module will give you an understanding of z/VM system characteristics and introduce: 4 Performance methodology 4 Planning measures 4 Monitoring facility tools 4 Tuning © 2004 IBM Corporation

IBM ^ Real Processor Management § CP is responsible for allocating the processing power of the available real processors in order to process virtual machine instructions. § Any processor complex on which z/VM runs must have a master processor. § A dedicated processor runs only the work generated by that virtual machine’s processor. § To enhance the performance of the V=R machine, CP automatically dedicates as many processors as it can to the virtual processor belonging to the V=R machine. © 2004 IBM Corporation

IBM ^ Real Processor Management continued § When the operator IPLs CP on the real machine, CP initially uses the master processor. § The CP in V=R machines automatically dedicates available real alternate processors under the following circumstances: 4 When the V=R machine defines a new virtual processor, using the DEFINE CPU command 4 After a VARY ONLINE PROCESSOR command, making a new real alternate processor available 4 When a V=R machine enters the DETACH CPU command the detached virtual processor was dedicated © 2004 IBM Corporation

IBM ^ Virtual Machine Scheduling and Dispatching § The dispatcher selects virtual machines (guests) from the dispatch list for the processors to run. § The CP scheduler function is responsible for calculating the priorities. § A virtual machine is permitted to have processor control for a portion of the elapsed time slice each time it is dispatched. § An elapsed time slice is an interval of time that each virtual machine is allocated and during which it has the use of the real processor. © 2004 IBM Corporation

IBM ^ Virtual Machine Scheduling and Dispatching, continued § The scheduling and dispatching routines that are implemented in CP are designed to favor: 4 V=R machines 4 V=F machines 4 Interactive virtual machines over non-interactive virtual machines 4 Installation-specified virtual machines © 2004 IBM Corporation

IBM ^ Elapsed Time Slice § When a virtual machine enters the eligible list, it is assigned an elapsed time slice. § The initial value of E 1 elapsed time slice is 1. 2 seconds. § The E 0, E 2, and E 3 elapsed time slices are fixed multiples of the varying E 1 elapsed time slice. © 2004 IBM Corporation

IBM ^ Scheduling and Dispatching Lists § In order to schedule and dispatch the set of virtual machines, CP groups virtual machines according to their current execution characteristics and resource requirements into these groups: 4 Dormant list 4 Eligible list 4 Dispatch list © 2004 IBM Corporation

IBM ^ Leaving and Entering the Dispatched List § Virtual machines alternate between the dispatch list and the eligible time slice. § When dropped from the dispatch list at the end of its elapsed time slice without completing its transaction, an E 1 virtual machine becomes an E 2 virtual machine. § A virtual machine is dropped from the dispatch list and placed in the dormant list if it enters an enable wait state (becomes idle) before its elapsed time slice ends. © 2004 IBM Corporation

IBM ^ Virtual Machine Scheduling Flow © 2004 IBM Corporation

IBM ^ Four Transaction Classes: Eligible List © 2004 IBM Corporation

IBM ^ Dispatch List © 2004 IBM Corporation

IBM ^ Planning and Administration Guidelines § Performance characteristics of an operating system are dependent on such factors as: 4 Hardware 4 Number of users during peak periods 4 Functions being performed by the system 4 System parameters § The three general tasks for improving your z/VM system efficiency are: 4 Planning 4 Monitoring 4 Performing © 2004 IBM Corporation

IBM ^ Performance Considerations § There are two areas of performance to consider: 4 The performance of individual virtual machines 4 The performance of the total system § The total system performance achieved in a virtual machine environment may be equal to or better than the performance achieved in a native environment. § Current processor usage is the most important factor in determining the performance that will be achieved in a virtual machine environment. © 2004 IBM Corporation

IBM ^ Major Factors Affecting Performance § The interrelationship of the following factors affect total system performance: 4 The speed and number of paging devices 4 Real storage size 4 Real processor speed 4 Workload characteristics © 2004 IBM Corporation

IBM ^ CP Performance Facilities § CP provides a set of performance facilities that can be used by individual virtual machines to improve their performance, such as: 4 Processor dedication option 4 Virtual machine multiprocessing 4 Virtual=Real (V=R) option 4 Virtual=Fixed (V=F) option 4 System scheduling control options © 2004 IBM Corporation

IBM ^ CP Performance Facilities continued § Other performance facilities are: 4 Scheduling share option 4 Scheduling maximum share option 4 Quick dispatch option 4 Expanded storage allocation option 4 Locked pages option 4 Real channel program execution option 4 Name saved systems (NSS) 4 Saved segments © 2004 IBM Corporation

IBM ^ CP Performance Facilities continued § Other performance facilities are: 4 VM/VS handshaking 4 Interpretive-execution facility 4 Guest wait-state interpretation capability 4 Minidisk Caching 4 VM Data Spaces 4 Spool file initialization 4 File caching option for CP-accessed minidisks 4 Hot I/O Detection 4 Virtual Disk in Storage 4 I/O Throttling © 2004 IBM Corporation

IBM ^ Performance Guidelines in a Virtual Machine Environment § There are some guidelines and hints that can be used to improve the performance of a virtual machine and its applications. § A list of such guidelines is provided to help explain the performance characteristics of a z/VM system. § The guidelines cover the actions taken to increase performance of both the z/VM environment and application programs. © 2004 IBM Corporation

IBM ^ Performance Monitoring § Performance monitoring is the periodic collection of performance data and serves two major purposes: 4 Early detection of performance problems 4 Problem determination for performance problems § The INDICATE and MONITOR commands provide system performance measurement facilities for z/VM, which include: 4 Obtaining system resource usage data 4 Collecting measurement data using a z/VM monitor for later analysis © 2004 IBM Corporation

IBM ^ Performance Analysis Programs § z/VM supports additional performance analysis programs: 4 IBM Realtime Monitor VM/ESA – used to help detect and diagnose problems and analyze system performance 4 VM Performance Reporting Facility – a tool for performance measurement, analysis, and problem determination 4 Performance Analysis Facility/VM – a graphics-based facility that helps the performance analyst © 2004 IBM Corporation

IBM ^ z/VM 4. 4 Performance Toolkit § The Performance Toolkit for VM is designed to assist operators and systems programmers or analysts in the following areas: 4 System console operation in full screen mode: – The features provided have been designed to facilitate the operation of VM systems, hence improving operator efficiency and productivity. These features include: • General system output can automatically be scrolled; however additional privilege classes may be needed depending on implemented functions • Messages and actions from other virtual machines are numbered and left pending at the top of the screen until explicitly deleted • A redisplay facility allows browsing through the day’s accumulated console log, or through previous days’ logs © 2004 IBM Corporation

IBM ^ INDICATE Command § This command provides the system analyst with a “snapshot” of system activities. Subcommands include: 4 INDICATE USER EXPANDED 4 INDICATE LOAD § These commands can help determine the execution characteristics of a program with respect to the resources it uses. © 2004 IBM Corporation

IBM ^ INDICATE Command continued § Other commands that can provide a “snapshot” include: 4 INDICATE QUEUES 4 INDICATE I/O 4 INDICATE PAGING 4 INDICATE VECTOR 4 INDICATE SPACES 4 INDICATE NSS © 2004 IBM Corporation

IBM ^ Monitor Performance Using CP Monitor § The CP Monitor facility collects system performance data that can be processed by an external data reduction program. § The user controls the amount and nature of the data collected. § MONWRITE can be used as the application program that establishes communication links with CP and retrieves records. § The monitor collects user-defined sets of statistics about z/VM system operations that are saved in domains. © 2004 IBM Corporation

IBM ^ Monitor Data § CP Monitor collects data during CP operations and stores the data collected in the saved segment. § Monitor collects and reports two types of data: 4 Event data 4 Sample data, which is broken into: – Single-sample data – High-frequency sample data § The system is divided into different domains for the purpose of collecting monitor data and generating monitor records. © 2004 IBM Corporation

IBM ^ CP Monitor Commands § You would use the MONITOR command to control monitoring. § Monitor commands include: 4 SET MONDATA 4 QUERY MONITOR 4 QUERY MONDATA § Monitor utilities include: 4 MONWRITE 4 MONWSTOP © 2004 IBM Corporation

IBM ^ CP Monitor Facility § To use the CP Monitor facility, a virtual machine must run an application program to retrieve data from the monitor saved segment. § The *MONITOR system service can be broken into four parts: 4 The virtual machine makes the connection 4 *Monitor notifies all applicable virtual machines that the records are ready to be retrieved 4 The virtual machine accesses the saved segment 4 *MONITOR notifies the virtual machine that it is finished © 2004 IBM Corporation

IBM ^ Tuning Agenda § § § § Tuning z/VM Tuning Methods Controlling the Dispatch List Allocating System Resources – Allocating Processors Tuning the Processor Subsystem Tuning the Paging Subsystem Tuning the Storage Subsystem Tuning the I/O Subsystem © 2004 IBM Corporation

IBM ^ Tuning z/VM § Why tune your z/VM system? § Performance tuning is undertaken when you want to: 4 Process a large or more demanding work load without increasing the system’s configuration 4 Improve system response or throughput 4 Reduce processing costs without affecting service to your users § Tuning is basically trial-and-error, but there is a step-by-step approach to any tuning project. © 2004 IBM Corporation

IBM ^ Tuning Methods § The key states to record are: 4 Running – CPU Wait 4 I/O Wait – Simulation Wait 4 Page Wait – Test Idle 4 Console Function Wait – Dormant 4 I/O Active – Active Page Wait 4 Limit List Wait © 2004 IBM Corporation

IBM ^ Controlling the Dispatch List © 2004 IBM Corporation

IBM ^ Allocating System Resources – Allocating Processors § Allocating system resources is a preliminary step to tuning your system. § CP allocates a V=R virtual machine one processor to the system and assigns each virtual machine the same share of processor time § To remove a dedicated processor from the V=R virtual machine, enter the UNDEDICATE command, which by default has a CPU ALL operand for the V=R user ID. © 2004 IBM Corporation

IBM ^ Allocating System Resources – Using SET SHARE § This command directory statement allows you to control the percentage of system resources a user receives. § A normal share attempts to provide the least amount of resources to a virtual machine. § A maximum share attempts to prevent a virtual machine from using more than the amount of system processor resource available. The types are: 4 NOLIMIT 4 LIMITHARD 4 LIMITSOFT © 2004 IBM Corporation

IBM ^ Allocating System Resources – Using SET SHARE continued § In general, you should assign typical users relative share rather than an absolute share. § A relative share setting lets you determine the percentage of system resources a virtual machine gets with respect to other virtual machines with relative shares. © 2004 IBM Corporation

IBM ^ Allocating System Resources – SET QUICKDSP § This command operand of the OPTION directory allows you to designate virtual machines that will not wait in the eligible list when they have work to do. § The types of virtual machines that should be considered for QUICKDSP include: 4 The V=R virtual machine 4 Other virtual machines with critical response-time requirements 4 A production V=V guest that has interactive characteristics © 2004 IBM Corporation

IBM ^ Tuning the Processor Subsystem § The resources of the processors installed at your location are referred to as the processor subsystem. § Some guidelines for reducing the load on your processors are: 4 Reduce the amount of CCW translations 4 Reduce the SHARE settings § To measure the workload, use these commands: 4 INDICATE LOAD 4 INDICATE QUEUES 4 INDICATE VECTOR © 2004 IBM Corporation

IBM ^ Tuning the Processor Subsystem continued § The command to display the current interactive bias is: 4 Query srm iabias § To change the interactive bias CP use: 4 Set srm iabias m n § The SET SRM DSPBUF command lets you control the number of users in the dispatch list § This command lets you over-commit and under-commit resources of the processors and I/O devices. © 2004 IBM Corporation

IBM ^ Tuning the Paging Subsystem § Paging is often the most influential component of overall response time. § To attempt to overcome a high paging rate, you should consider: 4 Adding real storage 4 Adding expanded storage for CP to page to 4 Define more of your existing DASD as paging devices § Display paging information using: 4 INDICATE LOAD 4 INDICATE PAGING © 2004 IBM Corporation

IBM ^ Tuning the Paging Subsystem continued © 2004 IBM Corporation

IBM ^ Tuning the Storage Subsystem § Waiting for storage to become available is a common cause of slow response time § To display information use the QUERY FRAMES command § Use one of the following commands to control storage allocation: 4 SET RESERVED 4 LOCK 4 ATTACH XSTORE 4 SET SRM STORBUF © 2004 IBM Corporation

IBM ^ Tuning the Storage Subsystem continued © 2004 IBM Corporation

IBM ^ Tuning the I/O Subsystem § Tune the I/O subsystem to determine what might be causing a system bottleneck. § Some guidelines for getting the best performance out of your system are: 4 Provide more real or expanded storage 4 Keep DASD volumes at a high degree of activity 4 Balance the activity on all DASD volumes 4 Reduce the number of I/O operations required per transaction § To display I/O information, use the INDICATE I/O command. © 2004 IBM Corporation

IBM ^ Tuning the I/O Subsystem continued § A subchannel measurement block is a control block that can be associated with a given device that contains measurement data § Three measurement components are: Connect time 4 Function pending time 4 Disconnect time 4 § Some important ways to control I/O resource are: SET THROTTLE 4 I/O Priority Queueing 4 SET SRM DSPBUF 4 SET SRM DSPSLICE 4 © 2004 IBM Corporation

IBM ^ Conclusion § Performance concepts help you set performance goals. § When you configure your z/VM environment, use the INDICATE and MONITOR commands to display and compare system characteristics. § The major goal for tuning your z/VM system is to optimize performance. © 2004 IBM Corporation

IBM ^ Glossary Channel command word (CCW): an instruction to an I/O processor (channel). Performance-oriented Assembler programmers used to code their own CCWs to improve efficiency. Dispatch list: the virtual machines in this list are competing for the use of the real processors and all other scheduled system resources. © 2004 IBM Corporation

IBM ^ Glossary Dormant list: consists of the logged-on virtual machines that are not currently being considered for dispatching because they have no work to accomplish or are waiting for the completion of a long event. Elapsed time slice: is the interval of time on a real processor that a set of virtual machines are competing for. Eligible list: this list forms when there is more demand by virtual machines for real resources than are available © 2004 IBM Corporation

IBM ^ Glossary Event data: is collected and reported each time a designated system event occurs Exclusive mode: in this mode, the virtual machine is the one, and the only one, connected to *MONITOR INDICATE: this command provides the system analyst with a list of system activities MONITOR: collects data during CP operations and stores the data collected in the saved segment in the form of monitor records © 2004 IBM Corporation

IBM ^ Glossary Sample data: is collected and reported at the end of each designated time interval Share mode: any other virtual machine may connect to *MONITOR concurrently © 2004 IBM Corporation

IBM ^ References z/VM: Performance 4. 3 – SC 24 -5999 -02. http: //www. vm. ibm. com/pubs/pdf/hcsi 1 a 50. pdf URL: Bitner, Bill. “z/VM Guest Performance: z/VM for z. Series Technical Conference V 91”. April 8, 2002. Bitner, Bill. “VM Performance Monitoring of Major VM Components”. February 22, 2001. z/VM: Performance Report 4. 3. IBM Corp 2002. © 2004 IBM Corporation