Yifat, Yes, I did see that description in VSAM Demystified, but it still does not make a lot of sense to me. What kind of "weighting factor"? How many buffers go to hiperspace when SMBHWT=10 or 30 or 80? There is no clue in any of the documentation I have seen so far, and no way for an ordinary programmer like me to find out how many were allocated after the fact because message IEC161I 001(xx)-255 does not always produce the second line as documented (with buffer and hiperspace buffer counts).
IMHO, IBM needs to document the formula used by SMBHWT. "Trust us, we know what we're doing" is not going to fly. We did see a significant decrease of 25-30% in total I/O's when using ACCBIAS=DO (but NOT using SMBHWT). CPU time did increase some but the I/O reduction gave better (lower) elapsed time and so would make the CPU cost justifiable. However, using SUBSYS=BLSR also produced approximately the same level (25-30%) of reduced total I/O counts and elapsed time with some CPU increase, and BLSR gives much more precise control over where and how many buffers actually get allocated. It's a tossup which to use without some serious standalone benchmark time on an idle machine (not an easy thing to come by these days). I do agree with some other posts I have seen re: hiperspace vs. dataspace that using hiperspace with today's z hardware hardly makes sense any more. Maybe it would be better for IBM to permit RMODE64=BUFF as an AMP option to move buffers into 64-bit space. Peter > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:[email protected]] On > Behalf Of Yifat Oren > Sent: Tuesday, November 15, 2011 6:47 AM > To: [email protected] > Subject: Re: What exactly does the SMBHWT subparameter do? > > Hi Peter, > > Have you seen the VSAM Demystified definition for SMBHWT? > > SMBHWT: Used to allocate hiperspace buffers based on a multiple of the > number of address space virtual buffers that have been allocated. It can > be an integer from 0 to 99. The value specified is not a direct multiple > of the number of virtual buffers that are allocated to the resource pool, > but act as a weighting factor for the number of hiperspace buffers to be > established. The hiperspace size buffer will be a multiple of 4K. These > buffers may be allocated for the base data component of the sphere. If the > CI size of the data component is not a multiple of 4K, both virtual > space and hiperspace is wasted. The default is 0 and means that hyperspace > is not used. > > It is not a direct multiple of the buffer count, but a "weighting factor" > .. > > > Very cryptic. > > In any case, did you see much of an improvement when adding hiperspace > buffers? Our experience shows that the I/O reduction (and elapsed) usually > achieved when adding the hiperspace buffers does not always justify the > CPU increase. -- This message and any attachments are intended only for the use of the addressee and may contain information that is privileged and confidential. If the reader of the message is not the intended recipient or an authorized representative of the intended recipient, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the message and any attachments from your system. ---------------------------------------------------------------------- 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
