The LU deal was certainly a big win for SUN. That market is UNIX, big systems. CPQ has a play in that space via Alpha but the delay in getting Wildfire out the door probably cost them that deal as the Alpha TurboLasers are getting a little long in the tooth.
As I pointed out to Jim Kelley, Gartner just released a study showing that the big SUN systems have poor reliability in the field, less than 98% and in some installations less than 94%. By contrast, CPQ's intel-based lines routinely get 99.5%, Alpha systems are 99.99%, and Himilaya systems (the gold standard for reliability) are 99.9999%.
Putting that in perspective, a Himilaya user can expect only a few seconds a year of downtime for any reason, scheduled or unscheduled. An Alpha user may see 45 minutes or so over the course of a year. A ProLiant user in a clustered configuration may see an hour or two, unclustered perhaps as much as 2 days a year total downtime (much of that is due to the need to cycle NT servers for configuration changes...).
A SUN UE10000 user, on the other hand, will experience between 7 and 21 DAYS a year of downtime. This is really a terrible number. It points to a big reliability weakness of the big SMP machines - the more stuff you have the more stuff there is to go wrong.
That is just simple math - if a single system has an hour a year of downtime, a 32-way system will have 32 hours a year, a 64 way system will have 64 hours a year. Unless, like the Tandem systems, steps are taken to assure that failure of a component has no effect on system operation. |
