z/OS DB2 V10 NFM Runstats

Tim Wilkins

z/OS DB2 V10 NFM Runstats

In the DB2 V10 performance manual it mentions that the runstats utility is zIIP processing eligible but inline stats isn't.

 

Has anyone done any comparison runs of a reorg with inline stats and without inline stats and doing the stats seperately to see if there is any advantage of not doing inline stats but instead move the runstats to a seperate step? 

Steen Rasmussen

z/OS DB2 V10 NFM Runstats
(in response to Tim Wilkins)
Tim - besides the benchmarks and zIIP eligible, please remember that INLINE STATS doesn't support all the variances of RUNSTATS like COLGROUP and HISTOGRAM etc.
So if your applications are dependent on these statistics, you might not have a choice.

Steen Rasmussen
CA Technologies
Sr Engineering Services Architect

From: Tim Wilkins [mailto:[login to unmask email]
Sent: Friday, August 02, 2013 3:31 PM
To: [login to unmask email]
Subject: [DB2-L] - z/OS DB2 V10 NFM Runstats


In the DB2 V10 performance manual it mentions that the runstats utility is zIIP processing eligible but inline stats isn't.



Has anyone done any comparison runs of a reorg with inline stats and without inline stats and doing the stats seperately to see if there is any advantage of not doing inline stats but instead move the runstats to a seperate step?

-----End Original Message-----

Avram Friedman

RE: z/OS DB2 V10 NFM Runstats
(in response to Tim Wilkins)

The advantage of inline stats is the data for analysis is obtained while the data is being passed any way for a diffrent function like COPY or REORG.  Generally speaking DP things run at the speed of I/O.  This is consistan with Av's first law of data processing which state "The principal activity of data processing is processing data"

 

Now the cost of running standalone runstats is most certainly extended elapsed time for the sum of other utility + runstats due to the duplicate data passing.  Running out of night or generally running out of window is one of the more important performance / capacity / and operational issues many companies face.

 

The advantage of native runstats is to the extent the work runs on a zIIp the billing amount for CPU may be reduced.  As utilities are often run during non peak times this would not reduce the shops hardware bill, just the internal cost for the mesured function.

In Reply to Tim Wilkins:

In the DB2 V10 performance manual it mentions that the runstats utility is zIIP processing eligible but inline stats isn't.

 

Has anyone done any comparison runs of a reorg with inline stats and without inline stats and doing the stats seperately to see if there is any advantage of not doing inline stats but instead move the runstats to a seperate step? 




Avram Friedman

IBM-sys-Prog.com

Tim Wilkins

RE: z/OS DB2 V10 NFM Runstats
(in response to Avram Friedman)

I ran numerous test doing a stand alone simple runstats on the table space and indexes and I didn't see any zIIP usage, at least not in SDSF where it displays zIIP info.   I didn't see any sysout messages to indicate zIIP usage.   Oh well it was a nice idea so to speak.     

Yifat Oren

z/OS DB2 V10 NFM Runstats
(in response to Tim Wilkins)
Hey Tim,

Check SMF 30-4 (for job step) or SMF 101 (for threads).
Those SMF records indicate both the zIIP used time and the zIIP eligible
time that actually ran on a CP.

I have seen zIIP offload for RUNSTATS executions. It can be significant.

Best Regards,
Yifat

_____

From: Tim Wilkins [mailto:[login to unmask email]
Sent: Wednesday, August 07, 2013 10:35 PM
To: [login to unmask email]
Subject: [DB2-L] - RE: z/OS DB2 V10 NFM Runstats



I ran numerous test doing a stand alone simple runstats on the table space
and indexes and I didn't see any zIIP usage, at least not in SDSF where it
displays zIIP info. I didn't see any sysout messages to indicate zIIP
usage. Oh well it was a nice idea so to speak.


-----End Original Message-----