ibm z/vse v4.1, v4.2 and v4.3 performance … › eserver › zseries › zos › vse › ...©2010...

70
© 2010 IBM Corporation IBM z/VSE V4.1, V4.2 and V4.3 Performance Considerations IBM System z

Upload: others

Post on 07-Jul-2020

3 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM z/VSE V4.1, V4.2 and V4.3 Performance Considerations

IBM System z

Page 2: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

2 24.Nov.10

TrademarksThe following are trademarks of the International Business Machines Corporation in the United States, other countries, or both.

The following are trademarks or registered trademarks of other companies.

* All other products may be trademarks or registered trademarks of their respective companies.

Notes: Performance is in Internal Throughput Rate (ITR) ratio based on measurements and projections using standard IBM benchmarks in a controlled environment. The actual throughput that any user will experience will vary depending upon considerations such as the amount of multiprogramming in the user's job stream, the I/O configuration, the storage configuration, and the workload processed. Therefore, no assurance can be given that an individual user will achieve throughput improvements equivalent to the performance ratios stated here. IBM hardware products are manufactured from new parts, or new and serviceable used parts. Regardless, our warranty terms apply.All customer examples cited or described in this presentation are presented as illustrations of the manner in which some customers have used IBM products and the results they may have achieved. Actual environmental costs and performance characteristics will vary depending on individual customer configurations and conditions.This publication was produced in the United States. IBM may not offer the products, services or features discussed in this document in other countries, and the information may be subject to change without notice. Consult your local IBM business contact for information on the product or services available in your area.All statements regarding IBM's future direction and intent are subject to change or withdrawal without notice, and represent goals and objectives only.Information about non-IBM products is obtained from the manufacturers of those products or their published announcements. IBM has not tested those products and cannot confirm the performance, compatibility, or any other claims related to non-IBM products. Questions on the capabilities of non-IBM products should be addressed to the suppliers of those products.Prices subject to change without notice. Contact your IBM representative or Business Partner for the most current pricing in your geography.

Adobe, the Adobe logo, PostScript, and the PostScript logo are either registered trademarks or trademarks of Adobe Systems Incorporated in the United States, and/or other countries.Cell Broadband Engine is a trademark of Sony Computer Entertainment, Inc. in the United States, other countries, or both and is used under license therefrom. Java and all Java-based trademarks are trademarks of Sun Microsystems, Inc. in the United States, other countries, or both. Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the United States, other countries, or both.Intel, Intel logo, Intel Inside, Intel Inside logo, Intel Centrino, Intel Centrino logo, Celeron, Intel Xeon, Intel SpeedStep, Itanium, and Pentium are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United States and other countries.UNIX is a registered trademark of The Open Group in the United States and other countries. Linux is a registered trademark of Linus Torvalds in the United States, other countries, or both. ITIL is a registered trademark, and a registered community trademark of the Office of Government Commerce, and is registered in the U.S. Patent and Trademark Office.IT Infrastructure Library is a registered trademark of the Central Computer and Telecommunications Agency, which is now part of the Office of Government Commerce.

For a complete list of IBM Trademarks, see www.ibm.com/legal/copytrade.shtml:

*, AS/400®, e business(logo)®, DBE, ESCO, eServer, FICON, IBM®, IBM (logo)®, iSeries®, MVS, OS/390®, pSeries®, RS/6000®, S/30, VM/ESA®, VSE/ESA, WebSphere®, xSeries®, z/OS®, zSeries®, z/VM®, System i, System i5, System p, System p5, System x, System z, System z9®, BladeCenter®

Not all common law marks used by IBM are listed on this page. Failure of a mark to appear does not mean that IBM does not use the mark nor does it mean that the product is not actively marketed or is not significant within its relevant market.

Those trademarks followed by ® are registered trademarks of IBM in the United States; all others are trademarks or common law marks of IBM in the United States.

Page 3: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

3 24.Nov.10

The information contained in this document has not been submitted to any formal IBM test and is distributed on an "AS IS" basis without any warranty either express or implied. The use of this information or the implementation of any of these techniques is a customer responsibility and depends on the customer's ability to evaluate and integrate them into the operational environment. While each item may have been reviewed by IBM for accuracy in a specific situation, there is no guarantee that the same or similar results will be obtained elsewhere. Customers attempting to adapt these techniques to their own environments do so at their own risk.

In this document, any references made to an IBM licensed program are not intended to state or imply that only IBM's licensed program may be used; any functionally equivalent program may be used instead.

Any performance data contained in this document was determined in a controlled environment and, therefore, the results which may be obtained in other operating environments may vary significantly. Users of this document should verify the applicable data for their specific environments.

It is possible that this material may contain reference to, or information about, IBM products (machines and programs), programming, or services that are not announced in your country. Such references or information must not be construed to mean that IBM intends to announce such IBM products, programming or services in your country.

Disclaimer

Page 4: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

4 24.Nov.10

Contents

§ Hardware support

§ z/VSE V4.1 Considerations

§ z/VSE V4.2 Considerations

§ z/VSE V4.3 Considerations

§ Sizing a system for z/VSE

§Miscellaneous Considerations

Page 5: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

5 24.Nov.10

z/VSE V4.1, V4.2 and V4.3 Hardware support

§ z/VSE V4.1, V4.2 and V4.3 runs on the following machines– IBM System z10 BC or z10 EC– IBM System z9 BC or z9 EC (formerly z9-109)– IBM zSeries: z800, z900, z990, z890– z/VM V5.2 (or later) is a prerequisite for running z/VSE V4 under

VM.

Page 6: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

6 24.Nov.10

Supported VSE Releases

11/26/2010z/VSE 4.3

11/26/201010/17/2008z/VSE 4.2

04/30/201110/17/200803/16/2007z/VSE 4.1

06/30/2002(out of service)

09/29/200006/25/1999VSE/ESA 2.4

12/31/2003 (out of service)

12/14/200109/29/2000VSE/ESA 2.5

12/31/2001(out of service)

06/30/200007/12/1997VSE/ESA 2.3

03/31/2006(out of service)

03/14/200312/14/2001VSE/ESA 2.6

02/28/2007(out of service)

09/30/200503/14/2003VSE/ESA 2.7

07/31/2009(out of service)

05/31/200803/04/2005z/VSE 3.1

End of ServiceEnd of MarketingAvailableVSE Release

Page 7: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

7 24.Nov.10

Running z/VSE V4 under z/VM

§ z/VM V5.2 (or later) is a prerequisite for running z/VSE V4.1, V4.2 or V4.3 under z/VM

–If you IPL z/VSE V4.1, V4.2 or V4.3 in a guest system of z/VM version 4 or z/VM 5.1, you may experience severe performance problems

–Because of that the following message is issued during IPL:•0J86I WARNING: VM RELEASE NOT SUPPORTED BY VSE 4.1 - Z/VM 5.2 OR LATER REQUIRED

–If you receive this message, you must urgently upgrade your VM system to z/VM 5.2 or a later release.

§ Note: It is not required to run z/VSE under z/VM, you can also run z/VSE in an LPAR

Page 8: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

8 24.Nov.10

VSE Server Support

No

No

No

No

No

No

No

No

Yes

Yes

Yes

Yes

Yes

z/VSE 4.3

No

No

No

No

No

No

No

No

Yes

Yes

Yes

Yes

Yes

z/VSE 4.2

NoYes (PTF required)

Yes (PTF required)

Yes (PTF required)

Yes (PTF required)

Yes (PTF required)

Yes (PTF required)

IBM zEnterprise 196

No

No

No

No

No

No

No

No

Yes

Yes

Yes

Yes

z/VSE 4.2

NoYes (PTF required)

Yes (PTF required)

Yes (PTF required)

Yes (PTF required)

Yes (PTF required)

IBM System z10 BC/EC

No

No

No

No

No

No

No

No

Yes

Yes

Yes

z/VSE 4.1

NoYes (PTF required)

Yes (PTF required)

YesYesIBM System z9 BC/EC (z9-109)

YesYesYesYesYeszSeries 800, 900

NoYes (PTF required)

Yes (PTF required)

YesYeszSeries 890, 990

YesYesYesNoNoES/9000 – 9221, 9121, 9021 (out of Service)

YesYesYesNoNoP/390 and R/390 (out of Service)

YesYesYesNoNoS/390 Parallel Enterprise Server G2 / G1 (out of Service)

YesYesYesNoNoS/390 Integrated Server

YesYesYesNoNoS/390 Multiprise 2000

YesYesYesNoNoS/390 Parallel Enterprise Server G3/G4

YesYesYesYesYesS/390 Multiprise 3000

YesYesYesYesYesS/390 Parallel Enterprise Server G5/G6

VSE/ESA 2.4/2.3 (*)

VSE/ESA 2.5 (*)

VSE/ESA 2.6 (*)

VSE/ESA 2.7 (*)

z/VSE 3.1 (*)

IBM Server

(*) Note: Although earlier releases technically run on selected servers, these releases are Out-of-Service anyway.

Page 9: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

9 24.Nov.10

VSE Hardware Support

Yes (PCICA, CEX2C, CEX2A, CPACF)

YesYesz/VSE 4.3

Yes (PCICA, CEX2C, CEX2A, CPACF)

YesYesz/VSE 4.2

Yes (PCICA, CEX2C, CEX2A, CPACF)

YesYesz/VSE 4.1

NoNoNoVSE/ESA 2.5 or earlier

NoYesNoVSE/ESA 2.6

Yes (PCICA, CPACF)YesYesVSE/ESA 2.7

Yes (PCICA, CEX2C, CEX2A, CPACF)

YesYesz/VSE 3.1

Hardware CryptoOSA Express (QDIO mode)

HiperSocketsVSE Release

Yes

Yes

Yes

No

z196

Yes

Yes

Yes

No

z9 BC/EC

No

Yes

Yes

Yes

z990

YesYesNoNoCPACF

YesNoNoNoCEX2A

YesYesNoNoCEX2C

NoYesYesNoPCICA

z10 BC/ECz890z900z800CryptoCard

Page 10: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

10 24.Nov.10

Overhead Deltas for VSE Releases

Remember that you get a lot of new functions that in most cases helps you to increase VSE system performance and throughput:

Partition Balancing, PRTY SHARE (Turbo Dispatcher), FlashCopy, Buffer Hashing, Shared data Tables (CICS TS), NOPDS with larger VSIZE

These numbers are for a specific average I/O intensive workload (PACEX16)

More I/O intensive workloads can cause more overhead

VSE/ESA 2.3Standard

DispatcherCICS/VSE

VSE/ESA 2.3Turbo

DispatcherCICS/VSE

+5-10%

z/VSE 3.1Turbo

DispatcherCICS/VSE

z/VSE 3.1Turbo

DispatcherCICS TS

z/VSE 4.1Turbo

DispatcherCICS/VSE

z/VSE 4.1Turbo

DispatcherCICS TS

z/VSE 4.2Turbo

DispatcherCICS/VSE

z/VSE 4.2Turbo

DispatcherCICS TS

+5% +5% +5%

3% 5%11%

3% 5%

z/VSE 4.3Turbo

DispatcherCICS/VSE

z/VSE 4.3Turbo

DispatcherCICS TS

+5%

3%

3%

Page 11: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

11 24.Nov.10

z/VSE V4.1, V4.2 and V4.3 - z/Architecture mode

§ z/VSE V4.1, V4.2 and V4.3: – Supports z/Architecture-capable (64-bit) processors. – Executes in z/Architecture mode only. – Supports 64-bit real addressing for selected system functions. – Supports processor storage up to 8 GB (z/VSE 4.2) / 32GB (z/VSE 4.2). – The storage beyond 2 GB is managed exclusively by the z/VSE operating

system.§ z/VSE V4.1, V4.2 and V4.3 does not support:

– 64-bit virtual addressing. The size of a virtual address or data space remains restricted to 2 GB.

– For user applications, 64-bit addressing and operations that use 64-bit registers.

§ 64-bit real addressing is transparent to your user applications providing you use IBM-supplied standard interfaces.§ Customers with especially large z/VSE environments might benefit

from lower paging rates. § Many z/VSE environments might be able to run without a page data

set (using the NOPDS option).

Page 12: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

12 24.Nov.10

What does 64 Bit ‘real addressing’ mean ?

16MB(24 bit)

2GB(31 bit)

Line

Bar

z/VSE 4.1/4.2/4.3VSE/ESA 2.x and z/VSE 3.1

Addressablereal storage

x GB(64 bit)

Addressablereal storage

Addressablevirtual storage

Addressablevirtual storage

VSIZE=4GB VSIZE=4GB

Page 13: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

13 24.Nov.10

What does 64 Bit ‘real addressing’ mean ?

§ VSE/ESA V2.x and z/VSE 3.1–z/VSE 3.1 or below can address only 2 GB of processor storage

• 31 bit (real) addressing– Page data set required if VSIZE (+VIO) > ~ 2GB– No page data set required as long 2 GB processor storage is

sufficient§ z/VSE V4.1, V4.2 and V4.3

–z/VSE V4.1 can address up to 8 GB of processor storage, z/VSE V4.2 & V4.3 up to 32 GB

• 64 bit (real) addressing– No page data set required

• If processor storage >= VSIZE (+VIO)–Virtual address spaces and data spaces are still limited to 2 GB

• 31 bit (virtual) addressing• No changes required to applications

Page 14: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

14 24.Nov.10

z/VSE V4 – Example with 3 GB of real storage

map real AR 0015 AREA R-SIZE R-ADDR PFIX(BELOW) PFIX(ABOVE) AR 0015 ACTUAL LIMIT ACTUAL LIMIT AR 0015 SUP 52K 0 AR 0015 SYS-24 592K 14136K AR 0015 BG V 0K 0K 0K 0KAR 0015 F1 V 148K 400K 0K 1400K AR 0015 F2 V 32K 144K 0K 0KAR 0015 F3 V 88K 424K 0K 0KAR 0015 F4 V 0K 0K 0K 0KAR 0015 F5 V 0K 0K 0K 0KAR 0015 F6 V 0K 0K 0K 0KAR 0015 F7 V 200K 400K 1044K 2100K AR 0015 F8 V 0K 0K 0K 0KAR 0015 F9 V 0K 0K 0K 0KAR 0015 FA V 0K 0K 0K 0KAR 0015 FB V 0K 0K 0K 0KAR 0015 SYS-31 7404K 2052264K AR 0015 DYN-PA 0K 0K 0K 0KAR 0015 AVAIL 64K AR 0015 SYSTEM 25068K AR 0015 TOTAL 3145728K <----’ <----’AR 0015 AR 0015 AVAILABLE FOR SETPFIX: 13544K 2044860K AR 0015 AR 0015 1I40I READY

Page 15: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

15 24.Nov.10

Exploiting 64 Bit real storage

§ Even on VSE/ESA or z/VSE 3.1 the VSIZE could ‘theoretically’ be up to 90G§ ‘Practically’ you are limited by

– Page dataset size and number of extents– Page I/O rate

Too heavy page I/O rates makes VSE almost unusable§ With z/VSE V4 you can run with VSIZE+VIO up to approximately

8GB (z/VSE V4.1) / 32 GB (z/VSE 4.2/4.3) without a page dataset (NOPDS)

–If enough processor storage is available–No time consuming page I/Os

§ You can have more large partitions in your system

Page 16: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

16 24.Nov.10

Paging considerations

§ ‘Paging’ is another word for Page Manager activities– Assigning real pages to virtual pages– Writing pages to the page data set (page-out I/O)– Reading pages from the page data set (page-in I/O)

§ Even with 8GB real storage and NOPDS paging happens– Assigning real pages to virtual pages– Moving pages from <2G to >2GB (above the bar) and vice versa

§ ‘Paging’ as such is not bad– But page I/Os are ‘bad’ (dependent on the page I/O rate)

Page 17: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

17 24.Nov.10

Paging considerations - PMRMON

§ New SIR command: SIR PMRMON– Displays reports from the ‘Page Manager Monitor’– Shows number of page faults, page I/Os, page exchanges (31 <->

64), ...§ Usage:

–SIR PMRMON=ON (resets counters)–// run your workload–SIR PMRMON–SIR MPRMON=OFF– Output example, see next foil

§ The output displays mainly internal counters that are for evaluation by IBM support persons

Page 18: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

18 24.Nov.10

Paging considerations - PMRMON

AR 0015 PAGE MANAGER MONITORING REPORTAR 0015 (BASED ON A 0000:01:04.801 INTERVAL) AR 0015 IPFQ 31-BIT = 517126 IPFQ 64-BIT = 115185AR 0015 PSQ 31-BIT = 1500 PSQ 64-BIT = 527AR 0015 PF EXCH TOTAL = 4673 PF EXCH 31->64 = 0AR 0015 PF EXCH 64->31 = 4673 PGFLT TOTAL = 25825AR 0015 PGFLT PMGR = 0 PGFLT USER = 25825AR 0015 PGFLT IMM PO 31 = 0 PGFLT IMM PO 64 = 0AR 0015 SELCT ON PSQ 31 = 0 SELCT ON PSQ 64 = 0AR 0015 SELC R=1 MAX 31 = 0 SELC R=1 MAX 64 = 0AR 0015 RECLAIMS = 0 NPSQ LOW = 0AR 0015 PGOUT I/O TOTAL = 0 PGIN I/O TOTAL = 0AR 0015 PGOUT I/O UNC. = 0 PGOUT I/O PRE. = 0AR 0015 LRA PGM CHECK = 3904 TFIX 64-BIT FR = 768AR 0015 1I40I READY

Description of values see next foil

Page 19: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

19 24.Nov.10

Paging considerations - PMRMON

Total number of page-in I/OPGIN I/O TOTALTotal number of unconditional page-out I/OPGOUT I/O UNC.Total number of pre-page-out I/O PGOUT I/O PRE.Number of special operation program exceptions on LRA LRA PGM CHECKNumber of TFIX requests with page frames above 2G involvedTFIX 64-BIT FR

Total number of page-out I/OPGOUT I/O TOTALNumber of times the available page frames are below a critical minimum NPSQ LOWNumber of times page frames are reclaimed from the page-out queueRECLAIMSMaximum number of cycles on a page selection event for page frames above 2GSELC R=1 MAX 64Maximum number of cycles on a page selection event for page frames below 2GSELC R=1 MAX 31Number of page selection events on page frames above 2GSELCT ON PSQ 64Number of page selection events on page frames below 2GSELCT ON PSQ 31

Number of page faults forcing an immediate page-out of a page frame above 2GPGFLT IMM PO 64Number of page faults forcing an immediate page-out of a page frame below 2GPGFLT IMM PO 31Number of page faults handled immediately without activating page manager task PGFLT USERNumber of page faults handled by page manager taskPGFLT PMGRTotal number of page faultsPGFLT TOTALNumber of times a page frame is exchanged from above 2G to below 2G (e.g. TFIX) PF EXCH 64->31Number of times a page frame is exchanged from below 2G to above 2G (e.g. instead of page-out I/O)PF EXCH 31->64Total number of times a page frame exchange is requestedPF EXCH TOTALNumber of page frames above 2G in page selection queuePSQ 64-BITNumber of unused page frames above 2GIPFQ 64-BITNumber of page frames below 2G in page selection queuePSQ 31-BITNumber of unused page frames below 2GIPFQ 31-BIT

Page 20: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

20 24.Nov.10

z/VSE V4.2 Performance Considerations

§More tasks–z/VSE V4.2 now supports up to 512 tasks

§PAV Performance Considerations–PAV = Parallel Access Volumes

§CPU Balancing–Automatically start and stops CPUs according to the current workload

Page 21: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

21 24.Nov.10

z/VSE V4.2 – More tasks

§ z/VSE 4.2 supports up to 512 concurrent tasks– the previous limit was 255 concurrent tasks. – Additional task IDs will be in the range X'0100' to X'0200'. – A QUERY SYSTEM command has been provided to display information

on the current task usage§ Please note:

– The maximum number of tasks per partition remains 32. – The default maximum remains 255. If you wish to increase the maximum

number of tasks, you must use the SYSDEF command to increase it to any number up to 512.

§ Benefits:– The workload resulting from CICS and batch programs can now be more

efficiently distributed– Previously, some customers might have been reluctant to perform

migration from CICS/VSE to the CICS Transaction Server for VSE/ESA because of its more intensive usage of VSE tasks (compared to CICS/VSE). This inhibitor has now been removed.

– Using more tasks have no measurable performance impact (neither negative nor positive)

Page 22: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

22 24.Nov.10

z/VSE V4.2 – More tasks

§ The SYSDEF SYSTEM command is used to activate the new tasks support system-wide.

§ NTASKS=nnn|MAX – nnn specifies the total number of tasks (old and new) that can be

allocated to the system and application programs. This is possible only during BG ASI processing. The value of nnn can be between 255 and 512.

§ TASKS=ANY|OLD– This is the system-wide default mode for subtask attaching, it can be run

at any time after IPL:• If TASKS=ANY is specified, new or old tasks can be attached as subtasks.

However, primarily new tasks will be attached. • OLD specifies that only old tasks will be attached as subtasks.

§ Use QUERY SYSTEM to display what has been specified with SYSDEF SYSTEM

Page 23: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

23 24.Nov.10

z/VSE V4.2 – More tasks

§ The TASKS specification in the SYSDEF SYSTEM command can be overwritten with the TASKS parameter in the JCL EXEC statement:

–When the SYSDEF SYSTEM command is used with TASKS=OLD, application programs which want to make use of the new tasks, must be started with parameter TASKS=ANY in the EXEC statement.

–When the SYSDEF SYSTEM command is used with TASKS=ANY, but an application program needs to be restricted to old tasks, the parameter TASKS=OLD has to be specified in the EXEC statement.

Page 24: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

24 24.Nov.10

z/VSE V4.2 – Parallel Access Volumes (PAV)

§ Parallel Access Volume (PAV) is an optional licensed featureon the IBM System Storage DS6000/DS8000 series§ Parallel access volumes:

–Are managed by creating multiple addresses (aliases) for a single logical device.

–Allow your z/VSE system to access volumes in parallel. –Enable more than one I/O operation to be processed for a single

logical device. –Depending upon workload, might represent a significant

performance improvement by the storage unit over traditional I/Oprocessing.

–Can significantly reduce device queue delays to busy devices (all I/O operationsare normally processed from a single queue).

–Allow your system to access a single volumefrom a single z/VSE host with multiple concurrent requests.

Page 25: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

25 24.Nov.10

z/VSE V4.2 – Parallel Access Volumes (PAV)

§ To gain a performance benefit from PAV, it is required that:–the workload can utilize multiple paths (sub channels) to the same

volume–A single job that starts an I/O and waits until its completion before

it starts another I/O does not benefit from PAV by itself–However, running multiple such jobs concurrently in multiple

partitions can utilize parallel access to the same volume.

§ Besides the simple ‘Start I/O and wait’ behavior, more sophisticated programs (like CICS) may be capable to start multiple I/Os and wait for its completion asynchronously

–This behavior also is eligible for benefiting form PAV, if the I/Os are for the same volume

§ Jobs that use multiple subtasks that perform I/O operations may benefit from PAV

§ Running multiple jobs concurrently in separate partitions and access the same volume are also good candidates for PAV

Page 26: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

26 24.Nov.10

z/VSE V4.2 – Parallel Access Volumes (PAV)

§ Candidates for PAV volumes– In general, all volumes that contain files, catalogs of data in general that

are used from multiple jobs can benefit from PAV• Without PAV, I/Os might get queued and therefore delayed in case of

parallel access to the same volume• PAV allows parallel access to these files from different jobs without

disturbing each other. – VSAM Catalogs– Shared VSAM Clusters– VSE Libraries

§ Besides volumes that contain files that are shared between multiple jobs, also volumes that contain heavily used system files can benefit from PAV

– Although there might not be any parallel access to those system files as such, there may be parallel access to other files residing on the same volume

• Without PAV, I/Os might get queued and therefore delayed in case of parallel access to the same volume

• PAV allows parallel access to these volumes without being disturbed by access to the system files.

– POWER Spool files– Workfiles– Log files

Page 27: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

27 24.Nov.10

z/VSE V4.2 – Parallel Access Volumes (PAV)

§ Besides the general PAV setup, no changes in application programs or JCL are required

–PAV is enabled in z/VSE via the SYSDEF SYSTEM command:

–Define the PAV base devices to z/VSE via ADD statement–You must NOT define your PAV-alias volumes to z/VSE!

• An ADD statement for an alias device will be ignored

§ z/VM support for PAV (since z/VM 5.2)–Dedicated PAV devices –PAV Mindisks. –For more information see here:

http://www.vm.ibm.com/storman/pav/pav2.html

Page 28: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

28 24.Nov.10

z/VSE V4.2 – Parallel Access Volumes (PAV)

Duration vs. NoPAV over #Aliases

0,00%10,00%20,00%

30,00%40,00%

50,00%60,00%

70,00%

80,00%

90,00%100,00%

No PAV 1 3 5 7#Aliases

FCOPY - COPY IDCAMS - REPRO LIBR - DEL+COPY:

Conclusion: Dependent on the workload, PAV can dramatically increase the I/O throughput and reduce the job duration (e.g. 1/3 of the duration when running without PAV), if it is not limited by other factors (like LOCKs).

PAV performance measurements:

Page 29: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

29 24.Nov.10

z/VSE V4.2 – CPU Balancing

§ Using CPU Balancing can reduce the multiprocessing overhead–When CPU Balancing is active, the Turbo Dispatcher only selects

CPUs that are required for the current workload–CPU Balancing might also reduce the overhead or performance-

degradation for I/O-intensive workloads that are running in one partition only (the remaining partitions are idle) when multipleCPUs are active

§ With CPU Balancing active, Turbo Dispatcher inspects the CPU utilization periodically

–When the the utilization falls below a certain threshold, CPUs are dynamically stopped

–When the the utilization increases over the threshold, CPUs are dynamically started

§ This results in only as many CPUs being active that are utilized by the current workload

Page 30: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

30 24.Nov.10

z/VSE V4.2 – CPU Balancing

§ CPU Balancing is activated using the SYSDEF TD command

§ INT=nn - nn defines the interval in seconds, after which the CPUs utilization is to be inspected.

– When 0 is specified, CPU balancing will be deactivated– The default value is 0. – The value must be in the range of: nn=0..99

§ THR=nn - nn defines the threshold value in percent– An additional CPU is activated, when the CPU utilization is larger than nn– The default value is 50– The value must be in the range of: nn=10..99

§ You can use the QUERY TD command to obtain information about thecurrent status of CPU balancing

Page 31: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

31 24.Nov.10

z/VSE V4.3 Performance Considerations

§ Support for 4 digit device addresses§ GETVIS constraint relief§ Dynamic start/stop of CPUs§ 1 MB frames for data spaces§ Fast Path to Linux on System z§ Queue-I/O Assist (QIOASSIST)§ Crypto Express3 and AP queue interrupt support§ FICON Express8§ SNMP Agent support§ VSAM Redirector executed in subtask under CICS

Page 32: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

32 24.Nov.10

z/VSE V4.3 – Support for 4 digit device addresses

§ z/VSE now supports 4 digit device addresses (that is, physical addresses) of up to X'FFFF'§ This support is implemented as follows:

– z/VSE applications, messages, commands, and so on, do not address a device by the physical address (pcuu), but instead by the VSE address (cuu)

– VSE addresses are in the range from X'000' to X'FFF'– To each physical address (pcuu) there is a corresponding VSE address (cuu)– If the physical address is less than or equal to X'FFF', the VSE address (cuu)

is equal to the physical address (pcuu)– If the physical address is higher than X'FFF' (and therefore outside the range

of VSE addresses), the physical address (pcuu) and VSE address (cuu) will be different

§ There is no performance impact by using 4 digit addresses

Page 33: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

33 24.Nov.10

z/VSE V4.3 – GETVIS below constraint relief

§ Several components have been updated to reduce the usage of GETVIS below the line:– I/O control blocks– VSAM code and buffers/control blocks– DL/1 1.12

§ Allocating I/O Control Blocks Above 16 MB Line– The IODEV supervisor parameter specifies the number of Input/Output (I/O) devices the

z/VSE supervisor will support– Using the IODEV parameter, you can instruct the z/VSE supervisor to allocate various

Input/Output (I/O) control blocks in the 31-bit shared addressing area (below the 2 GB line) instead of in the 24-bit shared addressing area (below the 16 MB line)

– You can set IODEV to either:• 1024, which means the z/VSE supervisor will allocate I/O control blocks in the 31-bit

shared area. This is the default value• 1023, which means the z/VSE supervisor will allocate I/O control blocks in the 24-bit

shared area (as in previous z/VSE releases). This provides compatibility for Vendor programs that will only run when I/O control blocks are allocated in the 24-bit shared area

– Note: If you run z/VSE with IODEV set to 1024, any Vendor-supplied programs that have not been modified so they address control blocks moved to the 31-bit shared area will no longer work! In such situations, contact the appropriate Vendor.

Page 34: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

34 24.Nov.10

z/VSE V4.3 – GETVIS below constraint relief

§ VSAM GETVIS constraint relief– VSE/VSAM modules have been relocated from the 24-bit to 31-bit SVA storage area– VSE/VSAM internal control blocks have been moved from 24-bit to 31-bit partition

GETVIS– VSE/VSAM storage is now acquired from GETVIS sub-pools, therefore allowing real-time

monitoring of GETVIS usage– User-generated VSE/VSAM control blocks (RPL, ACB, EXLST), as well as action macros,

can now be located-in / executed-from 31-bit partition GETVIS.– DLBL statement enhancements:

• You can now specify allocated space in blocks and cylinders for VSE/VSAM Managed-SAM clusters

• You can now use the BUFDAT=RMODE31 parameter to allocate buffers in 31-bit partition GETVIS, without the need to recompile legacy applications.

§ DL/1 1.12 GETVIS constraint relief– z/VSE 4.3 includes a new release (Version 1 Release 12.1) of DL/I VSE providing storage

constraint relief in the 24-bit area of a CICS/DLI partition– In a CICS/DLI online environment, the following DL/I resources can be allocated in 31-bit

storage above the 16 MB line:• Load PSBs and their transaction working copies• HD ESDS buffers.• DL/I action modules and user exit routines.

Page 35: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

35 24.Nov.10

z/VSE V4.3 – Dynamic Starting/Stopping of CPUs

§ z/VSE allows you to start CPUs in that were not online at IPL§ These can be either:

– CPUs that were in a “standby” state at IPL– CPUs that were added to the z/VSE LPAR profile after IPL

§ Note that z/VSE uses the term CPU to refer to Central Processors (CPs).

§ Using the HMC/SE “Logical Processor Add” task, you can add CPUs to an LPAR after IPL. This feature is available on IBM z10 and z196 platforms.

– When CPUs are added to an LPAR, z/VSE automatically updates its CPU configuration.

§ You can then use the:– SYSDEF TD,STARTSBY=cpuaddr command to set CPUs that are in a “standby” state to

an “online” state and start these CPUs– SYSDEF TD,STOPSBY=cpuaddr command to change the CPU state from “online” to

“standby”.§ These functions allow you to exploit z10 and z196 technology and update the CPU

configuration depending on workload needs

§ Note: “Standby” CPUs do not consume any CPU share of their LPAR

Page 36: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

36 24.Nov.10

z/VSE V4.3 – Support for 1 MB Frames for Dataspaces

§ The 1 MB frame support (also referred to as “large page support” or “1 MB page support”) allows 1 MB of virtual storage to be backed with 1 MB frames in processor storage§ The 1 MB frame support is available:

– on IBM z10 and z196 platforms– in LPAR mode only.

§ z/VSE uses 1 MB frame support for data spaces.– Long-running applications accessing data spaces frequently might benefit from improved CPU

utilization.– 1 MB frame support for data spaces might help you to exploit large processor storage to a

greater extent. § z/VSE implements this support transparently:

– In order to take advantage of z/VSE's 1 MB frame support, you are not required to perform any actions.

– z/VSE decides internally when and how data spaces will be backed with 1 MB frames. Therefore, a data space can consist of both “fixed” and “pageable” virtual storage.

– If the z/VSE system runs out of processor storage, one or more 1 MB frames used for backing data spaces will be made available again for use by the system.

§ Data space support in z/VSE is configured using the SYSDEF DSPACE command.– The default size of a data space (DFSIZE) is 960 KB. – All data spaces that you define with a maximum size of at least 960 KB will benefit from

1 MB frame support providing there is sufficient processor storage available.

Page 37: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

37 24.Nov.10

z/VSE V4.3 – Fast Path to Linux on System z (LFP)

§ The Linux Fast Path uses an IUCV connection between z/VSE and Linux, where both systems run in the same z/VM-mode LPAR on IBM z10 or z196 servers

– It allows selected TCP/IP applications to communicate with the TCP/IP stack on Linux without using a TCP/IP stack on z/VSE

– All socket requests are transparently forwarded to a Linux on System z system running in the same z/VM

§ On Linux on System z, the LFP daemon must run– This daemon fulfills all socket requests by forwarding them to the Linux TCP/IP stack.

§ The fast path to Linux on System z provides standard TCP/IP socket APIs for programs running on z/VSE:

– LE/C socket API via an alternative $EDCTCPV.PHASE (IJBLFPLE)– EZA SOCKET and EZASMI interface via an alternative EZA interface phase IJBLFPEZ– CSI’s (Connectivity Systems, Incorporated) assembler socket interface via the SOCKET

macro– Other than the basic socket API, no other tools are provided.

§ Performance increase due to:– Less overhead for TCP/IP processing on z/VSE (TCP, sequence numbers and

acknowledging, checksums, resends, etc)– More reliable communication method (IUCV) compared to HiperSockets, which is a

network device, with all its packet drops, resends, etc.

Page 38: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

38 24.Nov.10

z/VSE V4.3 – Fast Path to Linux on System z (LFP)Communication using TCP/IP• Data is passes from the application

to the TCP/IP stack partition (involves dispatching)

• TCP/IP builds IP packets (including TCP, checksums, sequence numbers, etc) and sends it through the OSAX device driver

• Linux TCP/IP stack receives the packets and passes it to the application

• Application receives the data• Response is going the reverse way

Communication using Linux Fast Path• Data is passes from the application

to the LFP LFP builds IUCV packets and sends it through to IUCV channel

• Linux IUCV device driver receives the packets and passes it to the LFP daemon LFPD.

• LFPD translates it into the appropriate socket calls against the TCP/IP stack

• TCP/IP passes the data to the application• Response is going the reverse way

Page 39: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

39 24.Nov.10

z/VSE V4.3 – Queue-I/O Assist (QIOASSIST)

§ The z/VM function queue-I/O assist (QIOASSIST) provides performance improvements for V=V guests using real adapters and real networking devices that use the Queued Direct I/O (QDIO) facility

§ z/VSE exploits the queue-I/O assist function for:– OSA Express devices (CHPID type OSD)– HiperSockets devices (CHPID type IQD)

§ To use the queue-I/O assist (QIOASSIST) function in z/VSE, use the z/VM CP command:– SET QIOASSIST ON

§ After you have enabled the queue-I/O assist function, each z/VSE DEFINE LINK,TYPE=OSAX command will then automatically exploit the queue-I/O assist function

§ If you do not wish to use the queue-I/O assist (QIOASSIST) function in z/VSE, disable it using this z/VM CP command:

– SET QIOASSIST OFF

Page 40: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

40 24.Nov.10

z/VSE V4.3 – Crypto Express3 - AP queue interrupt support

§ z/VSE supports the Crypto Express3 card (a secure coprocessor) in both coprocessor and accelerator mode§ The Crypto Express3 card is available with IBM System z10 and zEnterprise 196

– This Crypto Express3 generation of cryptographic cards will provide faster SSL acceleration

– The Crypto Express3 card is recognized during the startup of the Security Server partition (default FB) in the same way as other supported cards

§ Notes:– z/VSE uses the Crypto Express3 card with the same functionality as Crypto Express2.

New functions and key lengths are not exploited.– Crypto Express3 support is also available with z/VSE 4.2 and PTF(s).

§ AP queue interrupt support:– When performing cryptographic operations on Crypto Express2 and Crypto Express3

cards, the calling program is notified via a hardware interrupt when a response is ready for de-queueing from a card

– The AP-queue adapter-interruption facility is a new function of IBM z10 and zEnterprise196.

– However, this facility is only available when running in an LPAR (that is, not when running under z/VM).

Page 41: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

41 24.Nov.10

§ z/VSE Monitoring Agent enables customers to monitor his z/VSE systems using standard monitoring interfaces (SNMP V1)

– It also includes an open interface, which enables customers or vendors to use own programs (plugins) to collect additional data

§ Data collected by the IBM provided plugins contains– Information about the environment (e.g. Processor, LPAR and z/VM information)– Number of partitions (static, dynamic, total, maximum)– Partition priorities– Number of CPUs (active, stopped, quiced)– Paging (page ins, page outs)– Performance counters overall and per CPU– CPU address and status– CPU time, NP time, spin time, allbound time– Number of SVCs and dispatcher cycles

§ A MIB (Measurement Information Base) is provided describing the data collected§ Standard SNMP based monitoring tools can be

used to collect, display and analyze z/VSE performance monitoring data§ z/VSE SNMP Trap client is an application

that you can use in (or add to) your batch jobs to send SNMP V1 traps to inform one or more monitoring stations/servers when, for example:

– the end of a job stream is reached.– an error has occurred during a job stream

z/VSE V4.3 – SNMP Monitoring Agent support

Page 42: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

42 24.Nov.10

Redirector EXCPAD - Overview

§ Prior to z/VSE 4.3 the VSAM Redirector was executed in the same subtask as VSAM and the application (caller)

– Redirector activities may be time consuming (network transfers, database operations, ...)• During this time, no other activities are possible for this subtask

– Under CICS, VSAM normally returns back via EXCPAD exit when waiting for an I/O• Allows CICS to perform other activities concurrently

§ Since z/VSE 4.3 VSAM executes the Redirector under a separate subtask– VSAM now also returns back to CICS via EXCPAD when waiting for Redirector

• Allows CICS to perform other activities concurrently– This capability is primarily implemented for CICS TS transactions.

• The Redirector EXCPAD is not used for VSAM files opened by CICS/VSE.

Main Task Redirector subtask

EXCPAD RedirectorVSAM

RedirectorServer DB2TCP/IP

Page 43: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

43 24.Nov.10

z/VM 5.4 Considerations

§ z/VM V5.4 removes one of the consequences of stopped virtual processors for virtual processor prioritization

§ Prior to z/VM V5.4, a guest's CPU share was distributed equally among its virtual processors

– by dividing its share value by the number of non-dedicated virtual processors in the configuration

– regardless of whether the virtual processors were in a stopped or started state

§ z/VM V5.4 performs share redistribution whenever a virtual processor is started or stopped

– and no longer includes stopped virtual processors in the calculation of how much share to distribute to each virtual processor

§ The share redistribution also continues to occur in z/VM V5.4, as in previous releases

– at the time a SET SHARE command is issued – or when the number of defined virtual processors is altered.

Page 44: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

44 24.Nov.10

z/VM 5.4 Considerations

VCPU1 VCPU2 VCPU3

<= z/VM 5.3Guest 1 – Share=30%

Other Guests

10% 10% 10%

VCPU3

10%

Total = 30 %Total = 20 %

VCPU1 VCPU2 VCPU3

z/VM 5.4Guest 1 – Share=30%

Other Guests

10% 10% 10%

VCPU3

10%

Total = 30 %Total = 30 %

15% 15%

A guest's CPU share is distributed equally among its virtual processors by dividing its share value by the number of processors, regardless of whether the virtual processors were in a stopped or started state.

z/VM V5.4 performs share redistribution whenever a virtual processor is started or stopped and no longer includes stopped virtual processors in the calculation of how much share to distribute to each virtual processor.

Page 45: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

45 24.Nov.10

Sizing a system for z/VSE

§ Sizing a system for z/VSE is different from sizing a system for z/OS

– Although z/VSE supports multiprocessing, z/VSE does not scale as good as z/OS does

• Do not use more than 3 active processors per z/VSE LPAR or z/VM Guest

§ In general, a faster single CPU is better than multiple smaller CPUs– One partition can only exploit the power of one CPU

• The largest partition (e.g. CICS) must fit into one single CPU– Dependent on nonparallel share (NPS) value

§ Additional CPUs can be useful when multiple LPARs or z/VM Guests are used

– Define only up to 3 CPUs per LPAR or z/VM Guest, even if more than 3 CPUs are available on the CEC

§ Do not use MIPS tables for capacity planning purposes– Use zPCR Tool (see page 24) instead with the CB-L workload– Use free of charge Capacity Planning Services from IBM

Page 46: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

46 24.Nov.10

Sizing a system for z/VSE – Example 1

CPU 1 CPU 2 CPU 3 CPU 4 CPU 5

LPAR 1 LPAR 2

z/VM

z/VSE

Production

z/VSE

Development

z/VSE

Test

VCPU 1 VCPU 2 VCPU 3 VCPU 4

§ Although the CEC has 5 CPUs, each VSE only sees 2 CPUs

§ Test & Development systems have no influence on production system

Page 47: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

47 24.Nov.10

Sizing a system for z/VSE – Example 2

CPU 1

(LPAR 1)

z/VM

z/VSE

Production

z/VSE

Development

z/VSE

Test

VCPU 1 VCPU 2 VCPU 3

§ z/VM does a very good job in sharing a single CPU across multiple guests systems

§ Use SET SHARE (ABSOLUTE/RELATIVE) command in z/VM to prioritize production systems

Page 48: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

48 24.Nov.10

Sizing a system for z/VSE

(*) from a single VSE-image point o view

The fastest The fastest uniuni--processorprocessor

is (almost always *)is (almost always *)the best processorthe best processor

Page 49: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

49 24.Nov.10

IBM Processor Capacity Reference for zSeries (zPCR)

§ The zPCR tool was released for customer use on October 25, 2005– http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/PRS1381– ‘As is’, no official support, e-mail to [email protected]

§ PC-based productivity tool under Windows

§ It is designed to provide capacity planning insight for IBM System z processors running various workload environments

§ Capacity results are based on IBM's LSPR data supporting all IBMSystem z processors

–Large System Performance Reference: http://www.ibm.com/systems/z/advantages/management/lspr/

§ For VSE use z/VSE workloads Batch, Online or Mixed

Page 50: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

50 24.Nov.10

VSE CPU Monitor Tool

§ Intended to help customers to measure the CPU utilizationof their VSE system over a period of time. § When you plan for a processor upgrade it is very important

to know the CPU utilization of your VSE system over a day or a week.

–Helps you to estimate the size of the new processor.

§ The VSE CPU Monitor Tool is not intended to replace any existingmonitoring product provided by partners. § It provides only very basic monitoring capabilities on an overall VSE system

level. § No details about CPU usage of certain applications are provided§ New version available (XML Output) for z/VSE Capacity Planning

§ Download– http://www.ibm.com/systems/z/os/zvse/downloads/tools.html–‘As is’, no official support, e-mail to [email protected]

Page 51: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

51 24.Nov.10

VSE CPU Monitor Tool

§ CPUMON periodically issues a TDSERV FUNC=TDINFO macro to get performance relevant data. § The data provided by the macro is the same as command QUERY TD

shows.§ The data from each measurement interval is printed to SYSLST in a

comma separated format.§ Later on this data can be imported into a spreadsheet (EXCEL)§ CPUMON runs in a VSE partition (dynamic or static). § CPUMON is started using:

// EXEC DTRIATTN,PARM='SYSDEF TD,RESETCNT‘/*// EXEC CPUMON,PARM='nn‘ nn = interval in seconds/*

§ The tool can be stopped by entering the following command:

MSG xx,DATA=EXIT xx = partition id

Page 52: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

52 24.Nov.10

VSE CPU Monitor Tool

Example CPU utilization chart

0,00%

10,00%

20,00%

30,00%

40,00%

50,00%

60,00%

70,00%

80,00%

90,00%

100,00%

2005

/11/23

-14:11

:34

2005

/11/23

-14:12

:49

2005

/11/23

-14:14

:04

2005

/11/23

-14:15

:19

2005

/11/23

-14:16

:34

2005

/11/23

-14:17

:49

2005

/11/23

-14:19

:04

2005

/11/23

-14:20

:19

2005

/11/23

-14:21

:35

2005

/11/23

-14:22

:51

2005

/11/23

-14:24

:06

2005

/11/23

-14:25

:21

2005

/11/23

-14:26

:36

2005

/11/23

-14:27

:51

2005

/11/23

-14:29

:06

2005

/11/23

-14:30

:21

2005

/11/23

-14:31

:37

2005

/11/23

-14:32

:53

2005

/11/23

-14:34

:08

Time

Util

izat

ion

%CPU

%NP

Page 53: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

53 24.Nov.10

New: z/VSE Capacity Planning Offering

§A brand new z/VSE Capacity Planning Offering is now available

–Available for Business Partners–and Customers

§Performance data collection is based on a new version of the CPUMON Tool

§Contact [email protected] ask for z/VSE Capacity Planning Support

Page 54: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

54 24.Nov.10

IBM System z9 BC (1-4 CPUs)

050

100150200250300350400450500550600650700750800850900950

10001050110011501200125013001350140014501500155016001650170017501800

A01

B01

C01

D01 E01

F01

G01

H01 I01

J01

R01

S01

T01

U01

V01

W01 X01

Y01

Z01

A02

B02

C02

D02 E02

F02

N02

O02 P02

Q02

R02

S02

T02

U02

V02

W02 X02

Y02

Z02

A03

B03

C03

D03 L03

M03

N03

O03 P03

Q03

R03

S03

T03

U03

V03

W03 X03

Y03

Z03

K04

L04

M04

N04

O04 P04

Q04

R04

S04

T04

U04

V04

W04 X04

Y04

Z04

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

Page 55: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

55 24.Nov.10

IBM System z9 EC subcapacity models (1-8 CPUs)

0

200

400

600

800

1000

1200

1400

1600

1800

2000

2200

2400

2600

2800

3000

3200

3400

3600

3800

4000

401

501

601

701

402

502

602

702

403

503

603

703

404

504

604

704

405

505

605

705

406

506

606

706

407

507

607

707

408

508

608

708

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

Page 56: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

56 24.Nov.10

IBM System z9 EC (9-54 CPUs)

0

10002000

3000

4000

5000

6000

70008000

9000

1000011000

1200013000

1400015000

1600017000

18000

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

Page 57: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

57 24.Nov.10

IBM System z10 BC (1-5 CPUs)

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

0100200300400500600700800900

100011001200130014001500160017001800190020002100220023002400250026002700280029003000

Page 58: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

58 24.Nov.10

IBM System z10 EC (Models 400, 500, 600)

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

0200400600800

10001200140016001800200022002400260028003000320034003600380040004200440046004800500052005400560058006000

401

501

601

402

502

602

403

503

603

404

504

604

405

505

605

406

506

606

407

507

607

408

508

608

409

509

609

410

510

610

411

511

611

412

512

612

Page 59: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

59 24.Nov.10

IBM System z10 EC (Models 700)

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

11000

12000

13000

14000

15000

16000

17000

18000

19000

20000

21000

22000

23000

24000

25000

26000

27000

28000

29000

30000

31000

32000

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

Page 60: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

60 24.Nov.10

IBM zEnterprise 196

Note: Do not use MIPS to do any kind of capacity planning, use the zPCR tool instead !

02000400060008000

100001200014000160001800020000220002400026000280003000032000340003600038000400004200044000460004800050000520005400056000

2817

-401

2817

-501

2817

-601

2817

-701

2817

-402

2817

-502

2817

-602

2817

-702

2817

-403

2817

-503

2817

-603

2817

-703

2817

-404

2817

-504

2817

-604

2817

-704

2817

-405

2817

-505

2817

-605

2817

-705

2817

-406

2817

-506

2817

-606

2817

-706

2817

-407

2817

-507

2817

-607

2817

-707

2817

-408

2817

-508

2817

-608

2817

-708

2817

-409

2817

-509

2817

-609

2817

-709

2817

-410

2817

-510

2817

-610

2817

-710

2817

-411

2817

-511

2817

-611

2817

-711

2817

-412

2817

-512

2817

-612

2817

-712

2817

-413

2817

-513

2817

-613

2817

-713

2817

-414

2817

-514

2817

-614

2817

-714

2817

-415

2817

-515

2817

-615

2817

-715

2817

-716

2817

-717

2817

-718

2817

-719

2817

-720

2817

-721

2817

-722

2817

-723

2817

-724

2817

-725

2817

-726

2817

-727

2817

-728

2817

-729

2817

-730

2817

-731

2817

-732

2817

-733

2817

-734

2817

-735

2817

-736

2817

-737

2817

-738

2817

-739

2817

-740

2817

-741

2817

-742

2817

-743

2817

-744

2817

-745

2817

-746

2817

-747

2817

-748

2817

-749

2817

-750

2817

-751

2817

-752

2817

-753

2817

-754

2817

-755

2817

-756

2817

-757

2817

-758

2817

-759

2817

-760

2817

-761

2817

-762

2817

-763

2817

-764

2817

-765

2817

-766

2817

-767

2817

-768

2817

-769

2817

-770

2817

-771

2817

-772

2817

-773

2817

-774

2817

-775

2817

-776

2817

-777

2817

-778

2817

-779

2817

-780

Page 61: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

61 24.Nov.10

Performance Tips

§ A partition can only exploit 1 CPU at a time– 2 CPUs do not have any benefit for a single CICS partition– Use as many partitions as required for selected n-way

§ Use/define only as many CPUs as really needed– additional CPUs create more overhead, but no benefit

§ Partitions setup– Set up more batch and/or (independent) CICS partitions– Split CICS production partitions into multiple partitions (AOR,

TOR, FOR)

§ Try to exploit Turbo Dispatcher functions– Priority settings– Partition balancing– Partition balancing groups

Page 62: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

62 24.Nov.10

Performance Tips (2)

§ 1 CPU must be able to handle all non-parallel workload

§ Non-parallel code limits the n-Way exploitation– QUERY TD: NP/TOT = NPS (non parallel share)– Measure NPS before migration– max CPUs = 0.8 - 0.9 / NPS

NPS #CPUs0.20 4.0-4.5 (4)0.25 3.2-3.6 (3)0.30 2.7-3.0 (3)0.35 2.3-2.6 (2)0.40 2.0-2.2 (2)

NPS #CPUs0.45 1.8-2.0 (2)0.50 1.6-1.8 (2)0.55 1.5-1.6 (2)0.60 1.3-1.5 (1)0.65 1.2-1.4 (1)

Page 63: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

63 24.Nov.10

Performance Tips (3)

§ Non-parallel code limits the maximum MP exploitation§ System code (Key 0) increases non-parallel share

– Vendor code can have significant impact§ Overhead increases when NP code limits throughput

§ Data In Memory (DIM) reduces non-parallel code– less system calls (I/Os) – may increase throughput– CICS Shared Data Tables– Large/many VSAM Buffers (with buffer hashing)– Virtual Disks

§ Change VSE/POWER startup to WORKUNIT=PA § Switch tracing/DEBUG off for production

Page 64: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

64 24.Nov.10

zSeries Remarks – Split cache

§ Prior to zSeries there is one cache for data and instructions§ zSeries has split data and instruction cache

§ Performance implications:– If program variables and code that updates these program variables are in the

same cache line (256 byte)• Update of program variable invalidates instruction cache• Performance decrease if update is done in a loop

– See APAR PQ66981 for FORTRAN compiler

Page 65: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

65 24.Nov.10

zSeries Remarks – Split cache - example

Better code:* prepare lengthBCTR R2,0 ADJUST FOR SS-INSTR.EX R2,MVC01...MVC01 MVC RECEIVER(*-*),SENDER

Killer example:* prepare lengthBCTR R2,0 ADJUST FOR SS-INSTR.STC R2,*+5MVC RECEIVER(*-*),SENDER

STC instruction modifies the next instruction to set the length.

Use EXECUTE instruction instead.

zSeries Performance: Processor Design Considerations:http://www.ibm.com/support/techdocs/atsmastr.nsf/WebIndex/FLASH10208

Page 66: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

66 24.Nov.10

zSeries Remarks – Split cache - example

Not causing a problem:LA R1,PHASNAME POINT AT PHASE NAMECDDELETE (1)

+* SUPERVISOR - CDDELETE - 5686-032-06+ CNOP 0,4+ BAL 15,*+8+ DC A(B'00010010')+ L 15,0(,15)+ SVC 65 ISSUE SVC FOR CDDELETE

DS 0H

Can cause a problem:WTO TEXT=DATA

+ CNOP 0,+ BAL 1,IHB0003A BRANCH AROUND MESSAGE+ DC AL2(8) TEXT LENGTH+ DC B'0000000000010000' MCSFLAGS+ DC AL4(0) MESSAGE TEXT ADDR

...+IHB0003A DS 0H+ LR 14,1 FIRST BYTE OF PARM LIST+ SR 15,15 CLEAR REGISTER 15+ AH 15,0(1,0) ADD LENGTH OF TEXT + 4+ AR 14,15 FIRST BYTE AFTER TEXT+ LA 15,DATA LOAD TEXT VALUE+ ST 15,4(0,1) STORE ADDR INTO PLIST+* SUPERVISOR - SIMSVC - 5686-032

...+ SVC 35 ISSUE SVC 35@GE00016 DS 0H

CDDELETE uses an inline flag byte,but does not modify it

WTO uses an inline parameter list,but modifies the parameter list

Note: WTO can be coded with an externalparameter list: WTO …,MF=(E,addr)

Page 67: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

67 24.Nov.10

Possible performance issues with PPRC

§ Issue may occur if–PPRC is used–VSE runs in native or in LPAR–Not all devices that are defined in IOCP are also defined in VSE

ADD statements

§ In case there is an PPRC state change, interrupts are sent to all LPARs where the related device are defined in IOCP.

–If the device is defined in VSE ADD, no problem occurs: VSE willprocess the interrupt correctly.

– If the device is NOT defined in VSE ADD, the interrupt is ignored by VSE and the interrupt is resent very quickly to that LPAR

• Results in very high channel activity (up to 100%)

§ Solution:–Define ALL devices in VSE ADD that are defined in IOCP

Page 68: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

68 24.Nov.10

VSE/POWER POFFLOAD Performance Issues

§ Caused by incompatibility between VSE/POWER tape format and new tape drives

§ 3490F empties cache for FSF used by POFFLOAD LOAD– Install DY46164/DY46245 for VSE/ESA 2.7/2.6

§ 3590 synchronizes cache with tape for each WTM– Install microcode FC0520 on A60 controller + VSE/AF APAR DY45817 + AR

command TAPE WTM=NOSYNC– Unfortunately controller A50 is to small to install FC0520

Page 69: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

69 24.Nov.10

New POWER POFFLOAD Append function

§ New APPEND option (new with z/VSE 4.1):– POFFLOAD {BACKUP|SAVE|PICKUP},

queue,tapeaddr,density,class,APPEND...– WAVV requirement WAVV200433– Requires 3592 tape unit

§ Great performance benefit when using 3592 tape unit:– Spool time dramatically reduced when using 3592 tape unit (spool time cut half)– No wasting of time for skipping existing spool entries on tape when APPENDing using

“Space End of Data” command

204 seconds307 seconds (not supported)2490E

102 secondsAbout 1 second3592

Time to Spool 206 new entries at End-of-data

Search for End-Of-Data to begin APPENDing new Offload Entries

Tape unit

Page 70: IBM z/VSE V4.1, V4.2 and V4.3 Performance … › eserver › zseries › zos › vse › ...©2010 IBM Corporation IBM System z 724.Nov.10 Running z/VSE V4 under z/VM z/VM V5.2 (or

© 2010 IBM Corporation

IBM System z

70 24.Nov.10

Documentation

§ z/VSE homepage:– http://www.ibm.com/systems/z/os/zvse/

§ z/VSE Performance:– http://www.ibm.com/systems/z/os/zvse/documentation/performance.html

§ z/VM homepage:– http://www.ibm.com/vm

§ z/VM Performance:– http://www.vm.ibm.com/perf/

§ z/VM Preferred Guest Migration Considerations– http://www.vm.ibm.com/perf/tips/z890.html

§ IBM System z Software Pricing– http://www.ibm.com/systems/z/resources/swprice/

§ IBM's MSU ratings for IBM System z– http://www.ibm.com/systems/z/resources/swprice/reference/exhibits/hard

ware.html