John Thinnes writes: >We are a small shop looking to migrate from a single LPAR >2066-001 to a z9. On a normal day we hit 100% CPU utilization >for several hours during Prime Shift (8AM-4PM). The Prime >shift workload consists of production online work (CICS, IMS >and DB2) from our business community, development work (TSO, >batch and online) from our programming and QA groups and a >small amount of production batch.
Slight digression, but do you plan to configure a couple or three LPARs (even if they're not actually doing much at any moment)? I ask because part of what you're paying for is support for LPARs, and typically most businesses like having good service these days. Having "a couple" LPARs, even one on quiet standby, gives you some operational flexibility that can be very nice for the business users. And it's not like there's really a cost to that: basic mode is gone, and PR/SM is always doing its thing. No, it's not Parallel Sysplex, but it can be quite valuable to have a second production LPAR in order to minimize disruptions to the users. It's like having a spare tire for your car. It's included with the car, so you might as well keep it pumped up with air and ready to use if you blow a tire (or schedule blowing a tire :-)). >Our batch window is about 10 hours. We run at 100% utilization >during the first 8 hours, then drop to 50% for about 1 hour and >20% during the last hour. Assuming you're meeting your WLM goals, and the WLM goals are satisfying to the users, it sounds like you're getting your money's worth. Good job. >Our IBM reseller is suggesting we purchase/lease(36 mon) a 2096-Q02 >and run it as a downgraded N02 for approx 6 -12 months (or as long >as we can) and gradually add capacity as we grow. This is to save >on our IBM and ISV software costs. Part of the cost justification >for the purchase assumes 6-12 months running as a N02. I guess you could buy a -Q02, downgrade it to an -N02, and run that way. The other choice is to buy what you need (such as the -N02) and execute a CIU (Customer Initiated Upgrade) contract so that you can turn on whatever you need when you need it. Or some of both. Here's a guide to CIU and the other, related features (watch the wrap): http://www-1.ibm.com/support/docview.wss?uid=isg2f9dfc895a8a02302852570130054a44e For reference, the -N02 is 30 MSUs and the -Q02 is a whopping 47 MSUs. (Your current 2066-001 is 32 MSUs.) You have -O02 and -P02 steps between those two configurations, and that's just the 2-way configurations. Counting all the n-way configurations in between you have an amazing 11 intermediate steps between the two configurations. Said another way, the -Q02 is almost 57% more capacity than your starting configuration. How fast are you growing? Do you have a lot of latent demand? When you run the financial comparison you really should be comparing to the hypothetical of getting the next step up in the z800. (Well, you can't actually order that, but theoretically you might find one on the used market.) I think that's a 2066-0A2, and you'd take a big jump up to 44 MSUs for not too many more MIPS. On top of that you'd look at your growth over the next, say, 3 or 4 years and project forward. The z9 BC has 19% fewer MSUs for the same LSPR (average) performance, so it's more bang for the software buck. Also comes with a year of hardware maintenance. And the smooth z9 BC capacity increments are very nice indeed. Now, the hardware capacity isn't necessarily going to influence your *IBM* software charges. Assuming you can go onto VWLC (subcapacity pricing), and control that with softcaps, you can have as much activated hardware capacity as you want and still control your IBM software charges (with very few exceptions). You can run with a 3 MSU softcap on a System z9 Enterprise Class with 54 engines -- and pay for 3 MSUs. ISVs may be another thing. :-) Also, softcaps may be useful to you since you might be able to ride above the softcap line and get extra, "free" capacity. So arguably having a "little" extra hardware capacity could be useful, balanced against any ISVs who don't think you should enjoy that benefit. >They are also suggesting a zIIP. About 10-15% of the Prime >Shift utilization is DB2 DDF work. Any additional tools to >model the zIIP effect on the GP utilization? Unless you need this information to make the z9 BC decision itself, you can ask for a "loaner zIIP" on the BC and measure the impact with your real live workload before your employer buys one. That would be quite a definitive test. I cannot guarantee that you would qualify for a loaner since I don't know your situation, but it doesn't hurt to ask. For that matter you might want to sit down with an architect-type and see if you'd benefit from putting DB2 Connect on an IFL (still also with the zIIP). That configuration often has benefits. Hope all this rambling helps. It sounds like a good time for you to jump to the BC. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Specializing in Software Architectures Related to System z Based in Tokyo, Serving IBM Japan and IBM Asia-Pacific 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
