On 10 Sep 2012 08:11:00 -0700, in bit.listserv.ibm-main you wrote: >Clark Morris is entitled to have and to express his [not at all >disinterested] views. > >His question > ><begin snippet> >Does i-o on system z take less cpu and memory resource than i-o on >Intel servers?, p servers? ><end snippet/> > >is, however, challenging to this entitlement. > >The answer to this question is yes. z/Architecture channel-based I/O >is very different from that used by, for example, Intel servers. In >particular it uses many fewer cp[u] cycles, and its permits many >independent I/O operations to be executed concurrently and >asynchronously.
Server side has changed in the Intel world and given the uses to which the p series (and I believe that the i series has the same base hardware), the I-O may have much improved. In short what is the I-O mechanism on both these days (also the Sun, etc.)? It may not be channels but since both sides of Unisys tackled the same I-O question in different ways, there probably are other efficient ways of handling I-O and offload some of the work to controllers, etc.. > >Can it really be the case that Mr Morris is unaware of these >differences? Or was his question disingenuous? > >In passing, let me not that 'cpu' is/was an acronym for central >processing unit; and 'unit' implies singularity. Mainframes now >often, even usually, have more than one CP, (central processor) or, if >a three-character acronym is required, CPE (central processing >element). For purposes of the question, I was using CPU as an adjective so the number of CPUs is irrelevant since on the z series more than one cpu can be involved in handling I-O and I would assume the same is true on other platforms. Thus the question is how is the I-O processing distributed on the various architectures? If the Intel series still has to do much of the processing that is done in the controllers on the z series, that can explain why one would need more central processor power for the same work in an Intel environment. My interest is in understanding how much of the z advantage is in the architecture and how much is in existing applications that are hard to move. Having seen most of the mainframe work in the Maritime provinces of Canada move to other platforms over the past 15 years and with it the market for whatever limited expertise I have, I am interested in better evaluating the competing platform claims. Clark Morris > >John Gilmore, Ashland, MA 01721 - USA > >---------------------------------------------------------------------- >For IBM-MAIN subscribe / signoff / archive access instructions, >send email to [email protected] with the message: INFO IBM-MAIN ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
