On Sat, 23 Sep 2006 02:33:05 +0000, Ted MacNEIL <[EMAIL PROTECTED]> wrote:
>>WLM can do an ecxellent job of maintaining good response time for TSO and CICS environments as well as for batch production and test workloads. And it can do it with the processors running very near >capacity > >Please pass around whatever you're smoking; I need a toke. There is no need for this kind of personal attack. Just because you have not had the positive results that I have does not mean that I am wrong or using drugs. If you need help getting your systems to perform better let me know. > >Our Batch suffers for multiple reasons. >The main being the lack MTTW. > >I have worked with WLM GOAL MODE since 1999, and I have presented to >CMG Canada on my experience. Good for you. I started about the same time. I haven't presented, but I have posted here. I learned about making MVS perform in the mid to late 1970s from a man who was running 100 active TSO users on a 4 MB 370-158 with 90th percentile respomse times under one second. IBM had told him that he'd never be able to run more than 60, and that only if he didn't run anything else. He did run other work. There were test IMS regions and considerable development batch work. I implemented goal mode in production in 1999 also when I put OS/390 release 2.4 into production. I implemented WLM managed initiators the day we went production. WLM made everything perform very well. I used response time goals for CICS transactions, TSO and batch. The only velocity goal I had (aside from CICS startup) was for a group of started tasks that consumed a lot of resources and were considered less critical than CICS. These included ADSM and WebFocus. I put as much as I could into Discretionary, including third period TSO. When we began work on OS/390 2.4, we were runnung close to 90% CPU. Just before we went production, we were consistently running between 98% and 100% all day long. There was a lot of concern that performance would suffer when we made the big jump to the higher release because of increased resource requirements by the operating system. Our main production LPAR ran TSO, CICS, ADABAS and batch, oth test and production. The result was that CICS response times were slightly improved, but the perception was that they were improved even more than they were because they were more consistent. TSO response times were much better for trivial actions and about the same for almost all the rest. Only the longest running transactions ran a little longer. Batch turnaround time during the day remained good most of the time. In short, in this highly constrained environment, WLM helped us get the most out of our systems while providing good response times > >While Goal mode is better that COMPAT, it has its warts. >Especially, when you are at capacity. Running at capacity is where WLM provides the greatest benefit. > >I DO believe that, if you have more than one Image, TSO should be separated from ONLINE. Your earlier post said that it should be seperated from batch and you didn't mention online. Again, based upon my experience, I disagree. How much CPU does TSO consume on your systems? Is it enough to be a significant problem? What else do you run on your TSO LPAR and how high does CPU utilization get? > >If you don't, then agree to disagree. It is obvious that we disagree. As long as you continue to state your *opinions* as *facts* and "Best practices," I will post my contrary opinions. You and I can disagree, but that doesn't mean that I will defer to your expertise. Tom Marchant ---------------------------------------------------------------------- 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
