<<Comments interspersed below.>>
Timothy Sipples <[EMAIL PROTECTED]> Sent by: IBM Mainframe Discussion List <IBM-MAIN@BAMA.UA.EDU> 03/29/2006 07:21 AM Please respond to IBM Mainframe Discussion List <IBM-MAIN@BAMA.UA.EDU> To IBM-MAIN@BAMA.UA.EDU cc Subject Re: Bringing the fun back to z/OS - new course With respect to Patrick's comments, WebSphere Application Server/Java is certainly not IDMS/ADO (for example) from a resource utilization point of view. A "modest" two-way CP-only 31-bit-only system is simply not going to be delivering very high WebSphere volumes, I'm afraid. Unless your WebSphere Application Server workload is trivial, please do one of two things: (1) get a zAAP (for WAS z/OS); (2) get an IFL (for WAS Linux). It's frankly bad *finance* to run (much) WAS without either of these two options. Spend money to save a lot more money. <<Not to be flip but you're preaching to the choir here. I totally agree. We've had an upgrade on the books for a while and I've been waiting out these last several weeks to find out *if* I'm still employed. Our parent company is doing some extreme cost cutting measures, not that I don't agree with that, and I don't foresee an upgrade anytime soon. Luckily I *was not* one of the individuals tapped on the shoulder the other day, so I feel very fortunate. And I'm now tasked with keeping these workloads afloat and responsive, with an expected increase to Users to our current production WAS environment. We're currently doing about 150K transactions/day to a couple production WAS's.>> With either one of these two approaches mainframe WAS becomes not just affordable but, in numerous situations, the *most* cost-effective J2EE platform. My personal favorite is zAAP, but please choose at least one of these two avenues. <<Again, I agree. I know that we could take the several Dell plexes that are set up at our corporate center and fold them in very nicely on z architecture but getting the necessary ear has been very frustrating. Maybe at some point we'll get that work and I'm confident that we'll support and tune it effectively. Hopefully when the cost cutting is done and the dust settles we get a fair shot at making the necessary proposals>> Lastly, I think there's an implication that workloads in USS cannot fit into WLM service classing, goals, etc. in order to manage together with batch and other classic workloads. I hope nobody is saying that, because it's certainly not true. z/OS and WLM will manage all work, including USS-based work, as you tell it. If your system is too small to meet or exceed all goals at peak, that'll still be true regardless of the *type* of work you throw onto the system. WebSphere z/OS is spectacularly plugged into WLM -- it works really, really well, at least for the past three versions that I'm more familiar with (5.0+). But if I'm trying to suck an elephant through a straw and want the elephant to more or less retain its shape, well... :-) <<I've had firsthand experience, without zAAP, trying to WLM manage just the startup of a WAS V5 cell group. The support tasks are recommended in SYSSTC. The lights dim here in Pa. when the startup occurs. Again we are on RB6 at this time in 31 with a very busy LPAR but the burn in time required for startup of the cell group with tasks defined to SYSSTC would have this effect on small n-way cec's of any low end z series box wo/zAAP as well. I was tasked with trying to get WAS V5 running but based on the hardware/software setup it just wasn't possible without disruption to our othe workloads. I think most of us that have played with WAS understand the reason for zAAP's and that is the take the burden off the general purpose CP's in support of Java. Java just doesn't seem to play nice with the traditional workloads. I've been onboard with our developers and techs since the free offering of WAS, V3.02, and I can tell you that WAS anomalies react different than any other workload I've ever dealt with. This and the fact that WAS multi-tasks can wreak havoc on a small n-way environment when these unforeseen anomalous events present themselves. All in all I like WAS and Unix workloads but if I had my choice I would run them on dedicated LPARS where possible.>> - - - - - Timothy F. Sipples Consulting Enterprise Software Architect, z9/zSeries IBM Japan, Ltd. E-Mail: [EMAIL PROTECTED] ---------------------------------------------------------------------- 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
