tyvm, Timothy, for your expert analysis. Please pardon my ignorance, but what is a PCI?
On Fri, Feb 17, 2012 at 2:27 AM, Timothy Sipples <[email protected] > wrote: > George Henke observes: > >The feedback I have gotten so far, based on a few private replies, is > about > >8 - 12 people per CEC. > > Maybe, but that doesn't mean when you double the number of CECs you would > double the number of people, or vice versa. That is, you can't extrapolate > linearly in either direction, as our meat computers subconsciously often > do. > > For example, let's suppose you were a "big" shop in the year 2002 and you > were running 10 z900 machines, each configured as 213 models with a PCI of > 2888 each. So you had 28880 PCIs total, plus some coupling facility > engines. > > Then assume you experienced 8% per year compound growth in capacity (with > transaction volume growth, etc. -- holding the application set constant for > this example) so that after a decade you'd end up with approximately 62350 > PCIs (28880*1.08^10). Well, that capacity would fit on a mere two CECs > today: a pair of z196s, perhaps at capacity setting 742 each (31675 PCIs > each). An ~80% reduction in floor space, which unfortunately probably got > more than filled with more expensive and less reliable infrastructure. And > actually, in practice, when you take 10 footprints down to 2 you tend to > pick up some nice virtualization benefits, so that's probably too many > PCIs, never mind possible zIIP and other benefits. > > So in that decade would you have also taken a staff of 100 people (10 per > CEC) and reduced it to 20 people? That would be an order of magnitude jump > in staff productivity per PCI over 10 years. That seems extreme. Perhaps > you wouldn't have 100 people (if you started with 100), but I don't think > you'd have as few as 20 either, ceteris paribus. > > I don't think there's any serious disagreement that the mainframe has led > the way in providing huge productivity improvements just about any way you > measure it. As a generalization, you mainframers are extraordinarily > productive, both in comparison to your predecessors and in comparison to > your non-mainframe peers. (Keep up the good work -- and more, please.) > > There are some analysts who have looked at this stuff and who concur with > the sort of trends and characteristics I describe above. Mainframes are > characterized by very strong scale economies. There are at least two ways > to take advantage of that: be big(ger) -- more transactions, more volume, > more batch with the same or similar application set -- and be broader -- > more applications sharing the same mainframe infrastructure. That doesn't > mean you can't do fine financially and otherwise running a single > application at low volumes, but you can do even better bigger and/or > broader. > > > -------------------------------------------------------------------------------------------------------- > Timothy Sipples > Resident Enterprise Architect (Based in Singapore) > E-Mail: [email protected] > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to [email protected] with the message: INFO IBM-MAIN > -- George Henke (C) 845 401 5614 ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
