Mary (and Michelle), >>>"When does any of this stuff (Merced, McKinley, Everest et all) start taking away
business from the mainframes and if it does how does that effect you personally at
work?"
I should have saved it (at least twice in the past year or so, I have put in a post saying why Merced won't impact our, or IBM's S390 business, at least not for a long, long time). Oh well, here goes again.
1. Merced won't run MVS or VM, the main OS's of the S390 world. There is now over a trillion dollars worth of legacy software that runs on MVS or VM. The world's largest banks, insurance companies, car manufacturers, airlines, oil companies, EDS types, Merril Lynch types (good customer, in spite of TK), etc. don't want to convert these programs over to NT, so they'll stay with S390. There are some people moving over to NT from S390, but there are more going back to 390 because they like having one very powerful, single image, unbreakable machine doing their processing, instead of a fragile network of servers. That leads to the next key feature of S390:
2. RAS (reliability, availability and serviceability). S390's have extra hardware spent for these three cousins. For reliability, these machines have ECC on all memory, whether cache or main, mainstore patrol; data integrity testing for every instruction executed (excess three gray code); cyclic redundancy (Fire Code) testing for all I/O transactions; instruction retry, up to seven times, redundant power supplies and fans; and much, much more. For availability, hot swappable power supplies, hard drives and fans and dynamic CPU upgrade, where you can add CPU's to the system without powering down. For serviceability, well, the hot swappable stuff again, plus the system board (mobo in PC terms) is a FRU (field replaceable unit) so we don't mess around with troubleshooting a lot, just throw in a new one. Also for serviceability, AMDAC, by which a field specialist can take over any machine in the world (over modem), and run hand loops or diagnostics or invoke some articial intelligence software, which looks at the symptoms of a down machine, diagnoses probably 90% of possible bugs, and even orders the spare part to be sent out to the customer site. Intel is doing something like this (Landesk). We've had it for over 20 years (but not with the AI software for all that time). Oh, BTW, the MTBF of the new CMOS mainframes is running at between four and ten years. Ten years is longer than the lifetime of most machines. Mainframe customers upgrade just like we do. So, most machines will never fail, unless the get "recycled", or reused through third party people. Mainframe people used to smugly call PC's and Macs toys. Well they're a lot more than toys now, but still not up with big iron.
3. Security...hardware encryption.
4. Scalability...NT has a long ways to go on this one. People over on the SUNW thread say NT flat out sucks for scalability. OTOH, Under MVS, a twelve CPU system, with 80 MIPS per CPU, probably comes in at about 12X80X.78 = 748.8 MIPS. Before you say a Pentium Pro is more powerful, there are MIPS and there are MIPS. Mainframe MIPS represent far more processing power than PC MIPS. Back to scalability, NT 5.0 is supposed to "make it better", but there are skeptics (many). We'll see.
5. I/O bandwidth...up to 384 fiber optic (ESCON) channels, up to 192 parallel (OEMI) channels and up to 64 one Gigabit /sec. cluster links. I don't think even Merced will match this (McKinley?).
Mainframes sales, now based on CMOS technology vs. the obsolete ECL, are growing faster now in terms of MIPS shipped per month, or year, than ever.
Enough. Hope this helps. Don't sell IBM short. The only other S390 companies are where I sit (owned by Fujitsu now) and Hitachi Data Systems (Japan).
As far as yours truly at work, I feel more job security, thank you, than at any time since about 1991.
System 390 lives!!!
Tony |
