Martin, I honestly don't understand your question. The DSNB07 files are for temp tables (used to be DSNDB04) and workspace for ORDER BY and GROUP BY in SQL, and probably other things that I don't know enough about in DB2 to understand. The 32K tables are to support records greater than 4K.
The Long Service time comes from Seq pre-fetch IO which is pretty much standard fair for these tables. It looks like the same chaining technique is used as normal 4KB tables, so that each IO request is 32x32KB, or 1MB in size. Adding buffers won't help because it is DB2 Sequential pre-fetch, which simply wraps around the same set of 64x32K buffers no matter how big the pool is. There's also some sibling pend on some of the volumes that increases Disconnect Time. If this was SORTWK you can add all the BUFNO you won't but it won't make a scrap of difference. SYNCSORT and DFSORT both use their own buffering and chaining techniques for SORTWK. SORTWK is used when memory, Dataspaces and Hyperspaces are not adequate for sorting the records, so reducing SORTWK IO means increasing the sorts access to memory. SORTWKs typically can have some seriously long chained IO (I've seen 14 CYLS for DFSORT in one SSCH and have reports of over 80 CYLS for SYNCSORT). From observations of DFSORT I have noticed that a large MAINSIZE and CFW on will tend to favor longer and fewer channel programs, but I wouldn't dream of trying to get SORT to do smaller chaining and more IO. So again I don't follow your question Martin because you can't simply add buffers to these databases or SORTWKs in a way that will reduce the IO arrival rate. If we're talking about other sequential processing then I would think you would suggest to detune the sequential IO by reducing buffers and therefore the connect time. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf Of > Martin Packer > Sent: Saturday, September 05, 2009 1:54 AM > To: [email protected] > Subject: Re: [IBM-MAIN] EMC DASD and Hyper-PAVs > > > Not disagreeing at all but did adding extra buffers to the Sort Work > Buffer Pool (if dedicated) not help? > > Martin Packer > Performance Consultant > IBM United Kingdom Ltd > +44-7802-245-584 > > email: [email protected] > > Twitter ID: MartinPacker > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
