>>What metrics/reports/support documents can you cite? > well- let's just say that offloading 80% of 2% on a GP may not get you much.
Who said anything about 2%? We're talking numbers big enough to save you money. >>Remember there is also overhead just for having more than on CP? >>How much more overhead is there to choose a zIIP (or a zAAp, for that matter)? >depending on your hardware (number of books), there can be a 2-11% performance >hit >because you memory and/or your caches are in another core or book >(both zIIP and zAAP). And, how much different is that from GP to GP processing switches? >>And, what is that compared ti software cost savings? >>Enquiring minds want to know! >Offloading 10% will pay for the hardware in some number of months (8-10) And, doing more? >My recommendation is 20-30% white space on the GPs to get your monies worth on >the exercise Where did you get this metric? >and getting your software costs reduced (or rather, not go up). That sounds pessimistic. I'm still waiting for you to cite something concrete. I've been involved in IFL/zIIP/zAAP evaluations. Remember that these specialty engines run at full-speed, rather than the knee-capped speeds of the GP's. So, while overhead exists, is it a a problem? And, is the performance hit an issue? - Too busy driving to stop for gas! ---------------------------------------------------------------------- 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
