   
|
|
|
SOFTBASE EVENTS
Home > Upcoming SoftBase Events
DB2 UDB FOR z/OS VERSION 8 WEBINAR:
TIPS, TRICKS, & TECHNIQUES
DATE: Tuesday, August 30, 2005
Following are answers to questions raised during and after the recent DB2 UDB for z/OS V8 Webinar - our thanks to Roger Miller of IBM for providing this information:
| # | ATTENDEE | COMPANY | QUESTION(S) |
| 1 | Charles Crockett | Wells Fargo | The V7 & V8 Utility Manual had about 7 catalog & directory objects that could not be online reorged and had to have the system standalone. Does that continue to be the case, or has there been a PTF that lifted that restriction on those objects?
How do those 7 objects interact with the list of the 18 objects that ENFM does reorgs on, as I understand that ENFM does not reorganize the entire catalog or directory?
ROGER'S ANSWER -----
The V7 book has the restriction of needing to be reorged shrlevel none. The V8 book and product do not have this restriction. There is no PTF.
|
| 2 | Angela Arettines | Keyspan | Subject: REBINDs of old DBRMs with DESCRIPTOR
In migrating from V5 to V7, we ran the DBRMCHK (DBRM Colon Finder) to identify old format DBRMs (pre 2.3) that could potentially be violating the V6 enforceable rule that a colon must precede a host variable. We deferred the modification of a number of programs that violated this rule only when it involved descriptor-names without a colon. e.g. EXEC SQL FETCH SQLCURG USING DESCRIPTOR SQLDAG; etc. We were able to do this because of APAR PQ26922, which allowed rebinds of old DBRMs with descriptor-names and missing colons.
Question: Is this exception to the host variable/colon rule still in place in V8? If not, we would need to modify all of these programs now and would want to take advantage of the early prep time.
Thank you.
Angela
ROGER'S ANSWER -----
That exception is still in place. This would be an incompatible change and would be identified in the Installation Guide, if it existed.
|
| 3 | Suzanne Alexander | Peabody Energy Corp. | Where can I find more info from the performance guru? I didn't understand his name? Something like 'Akira'?
ROGER'S ANSWER -----
Akira Shibamiya is the performance guru. He is the key contributor to many efforts on performance. Start on the DB2 Support Web Page. Put Akira in the additional search terms, choose order by date – newest first and click go. This morning there were 33 items, with 9 from 2005.
|
| 4 | Angela Arettines | Keyspan | Subject: Multi-row FETCH benefits with SELECT *
Can we expect to see performance gains if we make use of the multi-row FETCH in V8 NFM even in programs that issue SELECT *? Many of our existing applications were designed years ago with object-oriented methodology in mind and the developers coded SELECT * in many modules. The DB2 V8 Performance Topics reports on performance savings using multi-row fetch with two different numbers of columns (5 and 20). If a SELECT * ultimately selects no more than 20 columns within an application, could that program see comparable benefits to those reported in the V8 Perf Topics Redbook even though a SELECT * is being used?
ROGER'S ANSWER -----
What matters is the number of columns, not the coding. SELECT * is not a good practice, but selecting every column by name would have the same performance. For any fetch, the number of columns is a significant part of the time. For the best improvements, return only the columns and rows needed with multirow fetch.
|
| 5 | Angela Arettines | Keyspan | In preparing Application Development groups for what the level of involvement needed from them would be throughout the entire V7 to V8 migration process (so they can budget their resources ahead of time), would you identify which steps/points in time you believe they should be available to support the migration? e.g. What should we expect (or at least hope for) from them when (1) All subsystems are still on V7 and pre-planning for V8 has begun (e.g. release incompatibilties) (2) When Test subsystems are migrated to V8 CM (but no QA or prod yet) (3) When prod subsystems are first moved to CM (4) When test subsystems are moved to NFM (5) When prod subsystems are moved to NFM (e.g. monitoring CPU of apps to identify new performance problems)
ROGER'S ANSWER -----
Here is my opinion. There are many differences in terminology and in tasks. As soon as possible, it’s very helpful to have all application people know about the incompatibilities. If there are reserved words being used as column or table names, that information helps. During the planning stages, it’s helpful to have an introduction to the overall process and SQL improvements for the lead developers and managers, so they can plan. Application people need to be involved and informed as the process proceeds. The key for performance monitoring is about the changes and information about the historical performance, plus changes in the performance monitoring tools. Then the new function and advantages need to be understood.
|
| 6 | Allan Smith | Kentucky Farm Bureau Insurance | We are planning to migrate our DB2 v7 to v8 in the near future but have discovered that several of our vendor packages are using some sql reserved words (ie. partition, security, sequence, summary) for column names. What kind of problems may we encounter if we were to go ahead migrate to DB2 version 8?
ROGER'S ANSWER -----
If the vendors use delimiters in their SQL, “PARTITION”, rather than PARTITION, for example, then there will be no problems. If that is not done, then there would be some SQL errors when those SQL statements are precompiled, prepared or bound. If the application is dynamic SQL, then you would get the errors at run time. I’d suggest that you contact the vendors to see if their application is ready for V8, and what release level and or service is needed.
|
| 7 | Angela Arettines | Keyspan | (Q1) Using a simple environment as an example with only one test subsystem, TST1, and one prod subsystem, PRD1, would the following states be the recommended method of migrating from V7 to V8:
(1) TST1 at V7&PRD1 at V7
(2) TST1 V8 CM & PRD1 V7
(3) TST1 V8 CM & PRD1 V8 CM
(4) TST1 V8 ENFM & PRD1 V8 CM
(5) TST1 V8 NFM & PRD1 V8 CM
(6) TST1 V8 NFM & PRD1 V8 ENFM
(7) TST1 V8 NFM & PRD1 V8 NFM
with the assumption that states 4 and 6 would be of short duration (same night planned in our case).
(Q2) Are there any precompile/bind issues when TST1 is in a different state than PRD1 and an application program in TST1 has to move to PRD1 before both subsystems are in the same mode? (i.e. in any of states 2,4,5 or 6 above) For example, if a new program is precompiled,compiled, linkedited and bound in TST1 when TST1 is running in V8 CM and PRD1 is still in V7, will there be a problem when the V7 code in the PRD1 ssid is used to do the bind on that program when it needs to be moved to production? Similarly, any problems in any of those other mixed states?
(Q3) Will DDF between subsystems in different modes be successful? Or do all subsystems with applications that use DDF between subsystems be required to always be in the same mode? e.g. Can a program that is run on PRD1 which uses an alias to TST1 (or some other subsystem not in the same mode) be successful in using DDF to get data from the other subsystem which is running in a different mode (e.g. PRD1 on V7 while TST1 on V8 CM or PRD1 on V8 CM while TST1 on V8 NFM)?
ROGER'S ANSWER -----
That scenario seems to be a good one, but we’ve seen many choices that work for customers. If you stick with the NEWFUN(NO) option or the V7 precompiler until some time after moving to NFM, you avoid the problems. DDF works across versions and releases.
|
| 8 | John Piccoli | cds | How can a terminal emulator's code page(CCSID) be identified?
ROGER'S ANSWER -----
See APARs PQ89018 and PQ97373. Check for DSNE345I message. We use an ISPF function to determine the terminal emulator CCSID.
|
| 9 | Neal Lozins | SoftBase | Is DRDA block fetch forced when DB2 V8 block fetch is used? If not, wouldn't that be a good idea?
ROGER'S ANSWER -----
Check SG24-6465 V8 Performance Topics, section 7.2 for a discussion of the improvements for multirow fetch in the DRDA environment.
|
| 10 | Dallas Rushing | Citigroup | Any SQL available for identifing user defined indexes on the catalog?
ROGER'S ANSWER -----
There is a query in DSNTIJPM or DSNTIJP8 for V7 that you can use and modify as needed. It looks like REPORT24 in the level I just looked at.
|
| 11 | Michele Mitchell | Mellon Trust | Would migrating all subsystems (test and production) to CM fulfill the licensing requirement of being off V7 and thus, not having to pay for two versions of DB2? Or do you have to have all subsystems in NFM?
Thanks
Michele
ROGER'S ANSWER -----
V8 includes CM, ENFM and NFM. That’s fine
|
| 12 | Michele Mitchell | Mellon Trust | Is the V8 catalog still user managed? Can't find a simple answer anywhere.
Thanks
Michele
ROGER'S ANSWER -----
Yes, it’s still user-managed. Is that simple enough, even if it’s not the answer you wanted?
|
| 13 | Michele Mitchell | Mellon Trust | We're seeing conflicting things about Dynamic Caching in V8. Some things we've read indicate you don't have a choice, it's on by default. But then we saw something that says it has to be allocated at 5M by default because the YES/NO parm is dynamic and there has to be something there just in case. This seemed to indicate having NO is still an option. Is this true? We're very nervous about turning this on because we have a VERY HEAVY dynamic workload, and none of it seems to be reusable (host variables that change). We tried turning this on in test years ago and crashed the system.
ROGER'S ANSWER -----
You can specify YES or NO for CACHE DYNAMIC SQL. You can specify 0 for MAX KEPT DYN STMTS. While we expect most customers will get a significant improvement from using the newer default, we can’t be sure that all will improve, so you can turn this off. The cost of the full prepare is generally many times (say 10 to 100 X) the cost of an SQL statement. There are options to change behavior with REOPT(ONCE) and REOPT(ALWAYS) if you expect reoptimization to be preferred. This could be a significant improvement for your shop, as it is for others.
|
| 14 | Michele Mitchell | Mellon Trust | It was mentioned in the presentation that some places saw up to 40% increase in CPU usage. Do you know what type of workload it was that saw this increase? We have heavy online, dynamic and batch, and normally run at 90% CPU usage so we're a bit nervous about this.
Thanks
Michele
ROGER'S ANSWER -----
I saw a mention of this on the web, but no details of the description there. There are many detailed measurements in the V8 Performance Topics Redbook, SG24-6465. I’d suggest starting there, as there is a detailed discussion across many measurements and variations. Use the Acrobat 7.0 reader and search for the word regression.
DB2 V8 Performance Topics Redbook
|
About Us | Solutions | Products | News | Partners | Support | Contact
Downloads | Customer Login | Web Site Map
|
|
|
 
|
