On Fri, 14 Dec 2007 19:49:28 -0500, Craddock, Chris <[EMAIL PROTECTED]> wrote:
>... >I don't mind that you disagree with me Rick, but if you're expecting >vendors (or anyone else) to provide reliable and detailed storage >estimates you're going to be permanently disappointed. I've been disappointed about this for some time and I guess it's not likely to change. I'll retire in a few years so I guess that's ok. >... The real world just isn't like that anymore. Maybe you're right, but to me that means a lot of vendors just don't give a d**n any more. Some products (like the Java and Websphere stuff mentioned earlier) probably can't give good guidelines, but lots of others could. Hint: If I can find a non-zero REGION value that works, there is probably a way I could have been assisted in finding it. >... I'm not asserting that the new way is >better than the old way; just that it is fundamentally different. And some us ARE asserting. It's worse. >... As Ed Jaffe observed, a paradigm shift is needed. I absolutely agree, but the shift has to result in a change in the operating system before it can be applied by customers. The system has to provide protection that REGION currently does before we abandon REGION. > >The amount of storage that is actually used by -any- job is really the >result of running the "experiment" of running that job with a given set >of inputs, resources and contention. In other words it can only be >described by a probability density function for each job, and certainly >not by a single number for the whole system! True. That has always been the case. That's part of the old pardigm; that will be part of a new paradigm. (Did someone say anything about a single number for the whole system??? I missed that.) >... >Also, it is a colossal waste of time and effort (and system resources >when you guess wrong) to place the burden of guessing that number on the >system programmer (via IEFUSI) or the application programmer (via JCL.) >The system ought to manage contention for its resources based on >something quantifiable like the (business) importance of the associated >service class. >... Sheesh. I want to say 3 different things first in response to that. It is NOT a waste of time. It *should be*, but it's not. Not until something in storage management changes based on that paradigm shift that hasn't happened. IEFUSI comes into play only if others have not done their job (in spite of the lack of help from vendors). Application programmers don't figure into my particular rant. I'm complaining about the products installed by system programmers. (In my case, things like VTAM, NetView, and the ton of TCP/IP- related address spaces.) >The one thing so many people seem to think that IEFUSI does protect them >against (a runaway storage user) is precisely the thing it does not do. >There is literally nothing (out of the box anyway) that does that today. >IEFUSI and REGION are just useless anachronisms. ??? Depending on the pool, REGION can protect from a runaway storage user. In what way does it not provide protection? (Ok. It won't protect from runaway ECSA allocation, etc. Is that what you mean?) Pat O'Keefe ---------------------------------------------------------------------- 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
