On 4 May 2006 06:53:11 -0700, in bit.listserv.ibm-main [EMAIL PROTECTED] (Tim Hare) wrote:
>Maybe this belongs on the Linux-390 list - but since the thread started >here, let's continue it here. > >If I were IBM, I would gather some of the smart Z/VM/Linux folk, and I >would go to places like Google, Yahoo!, or Six Apar with some mainframes i >in a trailer already running a hundred-or-so virtual Linux instances, for >a week or three. I'd let the company build their software on those Linux >images and test it out. > >If they want to test for longer, something like six months, I'd loan them >the machines FOR FREE, along with priority access to smart folks back at >the ranch (I'm not saying Google or Six Apart are guilty of this - but >many people just don't think about the incredible amount of knowledge and >intelligence within IBM). > >Treat it like a Redbook Residency for "extremely-large-scale Web >applications on z/Linux". Add one unusual contractual requirement: >everyone on the project blogs it all - with no sales department >censorship. > >I think the companies will be surprised that their work can run on >mainframes without major headaches; and I believe that they will also see >a potential to improve the cost of operations with the z/Series better >environmentals compared to a server farm. > >At the end of the process, you'd have web gurus singing the praises of the >hardware (in my opinion) and both IBM and the large-scale web folks would >have a great understanding of what it would take to sell zSeries boxes to >the next generation of large businesses, what it takes to support it, and >how the hardware handles it. Via the blogging component, you will have >exposed the hardware to a large number of web gurus that we don't normally >reach through SHARE and other venues, and will help overcome the image of >mainframes as ancient technology. I would be very careful if I were in IBM. The p series may be better for this type of work. Google has done a lot of optimization that is targeted to the Intel box and may be counterproductive on a z series. The p series is a faster box and it doesn't have some of the mainframe limitations on dasd device size. If the p series can handle more real memory it may be better suited. The search engine work is compute intensive and probably memory intensive which may not make it that great an application on a z box. In general, the newer the application the greater the data storage requirements and raw computing speed requirements. A picture can be larger than the largest record allowed by z/OS access methods thus requiring a data base to handle it. Parsing, text manipulation and similar activities can eat cycles. It would be interesting to see which computing environment that runs on z/OS would be best for Google: the algorithms rewritten for z/OS and application run on z/OS, the algorithms rewritten for TPF (Transaction Processing Facility, formerly Airline Control Program) or z/Linux under z/VM. If z/Linux, the optimum number of images on a 16 way (32?) would be of interest. > >Tim Hare >Senior Systems Programmer >Florida Department of Transportation >(850) 414-4209 > >---------------------------------------------------------------------- >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 ---------------------------------------------------------------------- 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
