Michael_D. 1100004WAH Visits (3002)
DB2 10 can reduce the total DB2 CPU demand from 5-20% when you take advantage of all the enhancements. Many CPU reductions are built in directly to DB2, requiring no application changes. Some enhancements are implemented through normal DB2 activities through rebinding, restructuring database definitions, improving applications, and utility processing. The CPU demand reduction features have the potential to provide significant total cost of ownership savings based on the application mix and transaction types.
Improvements in optimization reduce costs by processing SQL automatically with more efficient data access paths. Improvements through a range-list index scan access method, list prefetch for IN-list, more parallelism for select and index insert processing, better work file usage, better record identifier (RID) pool overflow management, improved sequential detection, faster log I/O, access path certainty evaluation for static SQL, and improved distributed data facility (DDF) transaction flow all provide more efficiency without changes to applications. These enhancements can reduce total CPU enterprise costs because of improved efficiency in the DB2 10 for z/OS.
DB2 10 includes numerous performance enhancements for Large Objects (LOBs) that save disk space for small LOBs and that provide dramatically better performance for LOB retrieval, inserts, load, and import/export using DB2 utilities. DB210 can also more effectively REORG partitions that contain LOBs.
This IBM Redbooks® publication® provides an overview of the performance impact of DB2 10 for z/OS discussing the overall performance and possible impacts when moving from version to version. We include performance measurements that were made in the laboratory and provide some estimates.
Keep in mind that your results are likely to vary, as the conditions and work will differ.
In this book, we assume that you are somewhat familiar with DB2 10 for z/OS.
See DB2 10 for z/OS Technical Overview, SG24-7892-00, for an introduction to the new functions.
Michael_D. 1100004WAH Visits (2951)
PM42528: SUPPORT FOR DELETING A MEMBER FROM A DATA SHARING GROUP
To use this support:
- The APAR or PTF providing delete member support must be applied to all members. Since deleting a member requires all members to be stopped, there is no pre-conditioning APAR / PTF.
- The member being deleted must be quiesced with no outstanding units of work, active utilities or retained locks.
- There should be no objects in restricted states. Use the -DISPLAY DATABASE(*) RESTRICT to verify.
- All surviving members must be DB2 10 New Function Mode
- The member to be deleted must be quiesced at some point before the surviving members are stopped so that the quiesced state is saved in all the surviving members' BSDSs.
- Stop all members of the data sharing group.
- Make backup copies of all BSDSs.
- Run the change log inventory (DSNJU003) DELMBR control statement against all the group members' BSDSs to deactivate the member that is to be deleted.
- Restart the surviving members of the group.
- When the logs of the member to be deleted are no longer needed, proceed.
- Stop all members of the data sharing group.
- Make backup copies of the BSDSs.
- Run the change log inventory (DSNJU003) DELMBR command against all the group members' BSDSs to destroy the member that is to be deleted.
- Restart the surviving members of the group.
- After all surviving members have been restarted, the logs and BSDS of the deleted member are no longer needed.
DELETE data sharing member related APAR's and PTF's
APAR : PM31003, PM31004, PM31006, PM31009
PTF : UK67512, UK67958, UK69286, UK65750
agburke 060001QPDN Visits (2861)
Workfiles have changed quite a bit from V8 -> V9. When we moved from DB2 V8 to V9 we combined the Temp. database (DGTTs) and Workfile database, and began favoring 32k table spaces if the records were over 100 bytes in length. Customers faced issues with runaway DGTTs eating up valuable workfile table spaces and impeding other work.
So IBM introduced a zParm to reestablish a deliniation between the two uses.
We then came out with a best practices informational APAR based on what customers had seen.
With the advent of DB2 10 V8 customers were introduced to this zParm after the fact, and were not always ready for the implications.
As part of the best practices we suggest when going into V9 or V10 size the 4k workfile buffer pool (BP7?) and the 32k workfile buffers (BP32k7?) equally in CM mode. You will need to create more 32k workfile tablespaces as well sinc not just DGTTs will use these now. The amount of space used by 4k and 32k tables for workfiles, as well as when DB2 wanted a 32k table, but was not able to get one is exposed in the Statistics report.
Once you get a handle on how much the 32k tables are used, increase them. In the field a 75 / 25 split is usually seen where 75% of the time 32k tables are favored, and hence 75% of the workfile space should be allocated to 32k table spaces.
In V10 DGTTs favor Partition By Growth tables first, then table spaces with >0 secondary quatities, then lastly those with 0 secondary quatities. So as a soft separation you might want to add some PBGs to the workfile database if DGTTs have been an issue.
timmzzz 060000RTR7 Visits (2799)
Customer X experienced Partition-By Growth (PBG) size increase with APPEND YES. Assuming the data is sparsely distributed across the PBG partitions. The following ONLINE REORG redistributed the data to the first 3 partitions. However, after the SWITCH PHASE, concurrent SQL would still append data to the last partition, leaving many partitions empty in between.
KevinHarrison 1000005N52 Visits (2761)
Traditionally customers submitted requests for requirements and enhancements thru their IBM contacts on their respective client teams. Now customers may manage and track their requests using a new process.
DB2 for z/OS Request for Enhancements (RFE) Community for Customer Requirements - Now Live!
The DB2 for z/OS
Request for Enhancements (RFE) Community enables customers to directly submit,
manage and track their requirements online. Customers can additionally access
requirements that others have submitted to vote on, comment and watch in a
social media paradigm. It provides customers with greater accessibility to the
requirements that are of interest to them. IBM's Rational, Tivoli and WebSphere
have already adopted RFE with positive customer feedback. All you need to get
started is an IBM developerWorks IBM ID. Please use the DB2 for z/OS RFE to
submit customer requirements going forward.
flodubois 270000K6H5 Visits (2750)
Below is a set of optimiser-related DB2 system parameters (ZPARMS) and their V10 default values. If the setting of any of these system parameters in your environment does NOT match the V10 default, then please re-evaluate the setting before migrating to DB2 10 for z/OS. If you need special assistance from IBM, please open a problem record (PMR).
MACRO ZPARM DEFAULT V10
DSN6SPRM OPTIOWGT ENABLE
DSN6SPRM OPTIXIO ON
DSN6SPRM OPTXQB ON
DSN6SPRM STATCLUS ENHANCED
agburke 060001QPDN Visits (2680)
CPU increase is always a customer concern, expecially when it occurs across a release boundary. Many of our customers in DB2 10 have seen a CPU reduction in the DB2 address spaces, and much of this can be attributed to the zIIP eligibility of asynchronous reads/ writes. However there have been some circumstances where the new monitoring or storage allocation behavior has affected CPU negatively.
Here are some APARs to address CPU degredation at the address space level.
PM56363 (HIPER): DIST TCB time increase due to remote SIGNON requ
Michael_D. 1100004WAH Visits (2669)
Higher CF CPU utilization in a DB2 10 for z/OS data sharing environment during Delete_Name processing. Delete_Name requests/sec can be significantly higher in DB2 10 for z/OS than in DB2 9 for z/OS. In V10, when the pageset/partition becomes non GBP dependent, the Delete_Name process deletes only data entries to avoid cross invalidation processing at that time, and to allow cleanup of the directory
entries later when other pages are registered. The V9 Delete_Name process deleted both data and directory entries. In DB2 10 for z/OS, the Delete_Name requests/sec can be significantly higher because CFCC processing is not as efficient when only data entries are deleted.
PM51467: CF DELETE_NAME PERFORMANCE IN DB2 10 FOR Z/OS http
GZJ 1100006WMT Visits (2638)
It is a common belief that DB2 10 and 11 for z/OS can only use 1MB size pageable large pages, other than for buffers, if the CEC has Flash Express installed. For example, the text for APAR PM85944 infers just that. However, this is not true. The only requirement is that the CEC be SCM-capable (SCM = Storage-Class Memory). In other words, it does not matter whether Flash Express is actually installed or not. So if a customer is running on a zEC12 CEC without Flash Express, DB2 could request, and be given, a 1MB size pageable large page, residing in a 1MB size large page frame. However if that page needs to get paged out for some reason, at that point it will be broken down into 4KB page frames. To put it another way, pageable large pages are available on a zEC12 capable CEC with the caveat that if Flash Express is not installed, then if those pages are ever paged out they will be demoted to 4KB page frames. 1MB pageable large pages that are demoted to 4KB page frames as they paged out will never be coalesced: that is, they will remain 4KB page frames for the remaining life of the IPL.
Michael_D. 1100004WAH Visits (2547)
System z Integrated Information Processor (zIIP)
The IBM zIIP is available on all IBM zEnterprise, IBM System z10 and IBM System z9 servers. The zIIP can support many technologies, and can help implement, integrate, and optimize new workloads on System z.
DB2 z/OS zIIP related performance:
PM12256: DRDA PERFORMANCE IMPROVEMENT USING TCP/IP
OA35146: NEW FUNCTION - ALLOW NON-CLIENT PREEMPTABLE SRBS TO JOIN/LEAVE AN ENCLAVE AFTER IT HAS BEGUN – z/OS APAR for PM12256PM2
DB2 z/OS informational APAR II14219 (DB2 z/OS ZIIP EXPLOITATION "SUPPORT USE" INFORMATION)
flodubois 270000K6H5 Visits (2537)
Migrating an SQL stored procedure from external to native is not as simple as a DROP/CREATE. You need to understand the release incompatibilities related to SQL stored procedures, examine your external SQL procedure source code, and make any necessary adjustments. This APAR can help you do that. It provides sample job DSNTEJ67 which initiates the process of converting source for an external SQL procedure into source for a native SQL procedure. REXX services, native SQL PL and the HOST(SQLPL) checkout precompiler are combined to extract, inspect, analyze and convert external SQL procedure source code. The appropriate set of native SQL procedure options are applied and a listing of the modified SQL procedure source code is produced.
GZJ 1100006WMT Visits (2513)
The draft of the redbook Optimizing DB2 Queries with IBM DB2 Analytics Accelerator for z/OS , SG248005, is available.
The 'blurb' says:
The IBM® DB2® Analytics Accelerator Version 2.1 for z/OS (also called DB2 Analytics Accelerator or Query Accelerator in this book and in DB2 for z/OS documentation) is a marriage of the System z® Quality of Service and Netezza® technology to accelerate complex queries in a DB2 for z/OS® highly secure and available environment. Superior performance and scalability with rapid appliance deployment provide an ideal solution for complex analysis.
This IBM Redbooks® publication provides a broad understanding of the IBM DB2 Analytics Accelerator architecture and its exploitation by documenting the steps for the installation of this solution in an existing DB2 10 for z/OS environment. We define a business analytics scenario, evaluate the potential benefits of the DB2 Analytics Accelerator appliance, describe the installation and integration steps with the DB2 environment, evaluate performance, and show the advantages to existing business intelligence processes.
If you want to review the draft, and provide feedback on this exciting new feature, now is your chance!
agburke 060001QPDN Visits (2507)
Many customers these days are utilizing DASD mirroring solutions as well as Hyperswap technology to automate fail-over to an alternate site or to local DASD hardware in the event of a failure or disaster. z/OS APAR OA31707 was put out to aid in the event of a fail-over by ensuring any pages it might need would not be paged out to AUX.
From OA31707: "During a Hyperswap, it is possible for the system to require page fault resolution via page devices that may be part of the scope of devices being recovered by the Hyperswap. If this occurs, it is possible that a page fault will not be able to be resolved leading to deadlock and Hyperswap failures."
The downside of this is a massive amount of page fixed storage which includes the following:
The purpose of this entry is to ensure customers are aware of the effect on real storage when this function is enabled, and to plan for it in advance. A system that is already running lean on REAL storage may see increased demand paging once this function is enabled, which can lead to DB2 entering DISCARD MODE (contraction) due to the REAL
If you have page fixed your buffer pools then the vast majority of the DBM1 PRIVATE address space will never be paged to AUX either, so you could end up with a severe shortage of REAL storage on the LPAR.
You can issue the D XCF,COUPLE command to determine if the function is enabled.
Further important information about the protection provided by this APAR and the service it introduces can be found as follows:
Two other APARs relate to the REAL storage growth seen in DB2 due z/OS not reclaiming frames when CRITICAL PAGING was enabled....
agburke 060001QPDN Visits (2499)
First is the Red Alert for PM51093 that was recently posted on the Red Alert website: http
Other potential data loss HIPERs include:
Correction: APAR remains open. Corrective relief, AM55070, is available from DB2 Technical Support. The estimated PTF availability date is April 6.
Notes: This is a problem specific to freespace reuse and later rollback. There should be no latent data corruption. i.e. the only way a customer would have data corruption is if they actually encountered an abend during rollback, in which case the page would be marked broken and they should report the problem to IBM Support.
There is no need for customers to proactively check for problems. If such is desired, then any utility such as DSN1COPY, REORG, COPY or any application access would report a page marked broken.
Correction: PTF UK76352 is available (not yet RSU)
Notes: There is no actual data loss, however, the potential exists for corrupted data pages and orphan pointer records. Customers wishing to proactively check for this condition can use DSN1COPY with the CHECK option, which runs offline and is non-disruptive.
However, the IBM COPY utility would automatically detect corrupted data pages, so a normal backup cycle should be sufficient to validate data. If DB2 detects a problem, then the page may be marked broken. This can be reset by REPAIR using REPAIR LOCATE db.ts PAGE(nnnn) RESET. The IBM REORG utility will correct the page corruption.
flodubois 270000K6H5 Visits (2368)
Q. If PIECESIZE is specified at maximum of 64GB and (LARGE or DSSIZE x is specified on tablespace), then is the maximum pageset size going to be 64GB*4096 parts =256TB?
A. The number of index data sets is influenced by the number of data partitions. Initially using PIECESIZE of 64GB you could not index 128TB of data using DSSIZE 128 GB or 256 GB due to the reduced number of index data sets. APAR PM42175 was created to allow the specification of 128 GB and 256 GB PIECESIZE.
Specification of 64GB is only allowed when the number of data partitions (derived from DSSIZE) is less than the following maximum data parts.
Chart for 128GB PIECESIZE
Chart for 256GB PIECESIZE
Q. Now that DB2 can support a DSSIZE of 256GB for tablespace what is the new max size for the NPI?
Q. Is the only way to influence it - by increasing DSSIZE for the tablespace?
Q. Is there a way to calculate how much is space is actual used/left to grow based on the Catalog or RTS?