Rick, We were talking about getting vendors to model your performance on their respective offerings, not an existing IO problem. It's the sort of thing that was done by people at BGS that lived in small cells underground and they were occasionally rewarded with raw meat.
The proposition is that a vendor can take 24 hour of data an accurately predict your performance with no other input than some SMF records. These DASD controllers can do half a million IOPS and an accurate forecast of performance based on 900 or 1800 second averages over 24 hours is required. And then you have to guarantee that performance for 3 years no matter how the customer grows, shrinks or changes the workload. Have a look at the source code for FlashDA and tell me that it's going to do that :-( While Shane did not go easy on vendors, I thought that one of his points was that this sort of warranty protects the performance guy at a customer site from actually having to understand IO performance. Yes I work for a vendor, but I've spent more time working for customers. No need for me to put shields up, I have a thick skin and head to match. Yes there are many, many brilliant IO performance gals and guys out there working on MVS IO performance. I don't think Shane was referring to anyone in that category. Ron > -----Original Message----- > From: IBM Mainframe Discussion List [mailto:ibm-m...@bama.ua.edu] On Behalf Of > Rick Fochtman > Sent: Friday, August 27, 2010 1:35 PM > To: IBM-MAIN@bama.ua.edu > Subject: Re: [IBM-MAIN] Test DASD performance tools > > ------------------------------------<snip>---------------------------------- - > > >>That's a good idea, providing the customer is willing to provide accurate > >>description of what they want to have modeled. SMF has a lot of really good > >>averages, but very little in the way of detailed access patterns. > >> > >>Can you tell skip sequential access from random with a Type 74 record? Can a > >>DASD vendor factor your CPU upgrade into a storage model? Will a customer > >>agree not to change their IO patterns for the life of the performance > >>warranty? > >> > >> > > > >In less polite company all this would be known as arse covering. > >The customer wants the vendor to indemnify their (the customers) lack of > >knowledge (due in no small degree to vendor obfuscation), and the > >vendors want to extract as much as possible for as long as possible > >without enlightening said customer. > > > >Business as usual to this jaundiced eye. > > > > > --------------------------------<unsnip>------------------------------------ -- > -------- > I agree with Shane. > > You need to explain to this vendor, in no uncertain terms, that YOU are > the arbiter of whether this equipment is performing or not, based on the > regular performance of YOUR workload. If he/she/they want to cover their > collective rear ends, then the equipment better live up to the specs > they offered for it. PERIOD. Be nasty and unpleasant if you have to; > it's your shop, not theirs. And it's your money! > > Don't be afraid to demand detailed information and try to get some > references. As many as you can get~Preferably > > Rick > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO > Search the archives at http://bama.ua.edu/archives/ibm-main.html ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@bama.ua.edu with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
