IBM 610 workstation computer 3459

You have fallen behind the times. Hardware is extremely more caused by other systematic failures like bane overheat; power supply failures etc.

To the "newbies" around these newsgroups, I remember making a log book entry on a Prime that had been up for half a year, 183 days. That was exceptional. We had changed a mirrored disk on it and a few comms cards in that timeframe. It took a solid dose of sound systems design to get even that far.

To the "Oldtimers"; Hardware is just like cars these days. It just works, until some software writer make them crash. Nowadays a plain 'ruptime' with lowly non-server stuff gives stellar uptimes. Around here it looks like this :

acer up 348+16:08, 0 users, load 0.00, 0.00, 0.00 eden up 443+01:19, 1 user, load 0.06, 0.12, 0.20 wifes up 6+00:38, 2 users, load 0.17, 0.40, 0.35 husbands up 1:33, 5 users, load 0.05, 0.17, 0.20 pilt up 160+20:06, 8 users, load 1.85, 1.95, 1.89 pompel up 48+15:14, 0 users, load 0.04, 0.02, 0.00 puff up 42+02:47, 1 user, load 0.07, 0.08, 0.10 thalbutta down 318+18:08 via up 140+05:47, 0 users, load 0.00, 0.00, 0.00

This is not a particularly impressive uptime list. All of these are standard issue PCs with some Open Source OS, with a small ups to help. The main reason to reboot is a switch of kernel software.

No, failovers must work even if you lose the building you reside in. They must be automatic. They must present an ongoing service within a few seconds (around 1-99, depending on application) if you are not to lose any business.

That is why the SMP service Barb describes har become irrelevant. Systems provide excellent uptimes if they are handled right, and if they are handled wrong no SMP system will help you anyway.

SMP is there for quality and performance reasons. It does actually detract a tiny bit from reliability.

The computing hardware is not the limiting factor anymore. The software is, as well as the care and feeding of those computing platforms.

-- mrr