[ILUG] Have you had your daily dose of reality?

[John P. Looney]

On Fri, Feb 18, 2000 at 11:46:42AM +0000, Brian Scanlan  mentioned:
> "Sun claims to be a leader in system reliability and more reliable than
> Windows.
>                             The Proof:
>                                    Despite Sun's claim that their high-end
> servers are highly reliable and built with redundant components, customers
> report that failures in service processors, controllers, processor cards,
> and other components have caused entire production systems to fail.
>                                    (Source: Gartner Group, press reports)"

 True. However, on a high end system, if a processor is found to be
faulty, the machine will panic, reboot, and isolate the damaged processor,
and the system will keep going. Which rocks. Any high-end intel box will
sit stupidly, beeping when it detects a severe CPU failure.

 Sun do have some "stupid" things. For instance, a Sun E10000 has pair of
"Service processors". Effectively, low-end sparc workstations, with
virtual console software on them. I don't mean "virtual console" in the
Linux term. Each part of the E10000 (16 motherboards, disk controllers,
backplane) has it's own diagnostic CPU on them, linked by a 300 baud
connection. So, you can connect into the backplane, and power up the
motherboards (each with 4 CPUs) individually. Rather cool.

 The idea is you have a backup SSP if one fails, for hardware reasons.
And, if you boot the two at the same time, the both try and control the
internal "control" bus of the E10000, which panics the E10000 & it
reboots. Fun, huh ? I wonder how many sysadmins have done that on a live
system ;)

 But, bar silly things like that, the E10000 is very well designed, for a
system that's not fault tolerant. Which it's sucessor is supposed to be
(think RAID 1 for motherboards).

>                             The Reality: 
> Sun's Live Upgrade feature still requires the
> operator
> to reboot the system. Some reboots, according to
> Sun, can take up to 5 hours. "
> 
> JEsus Christ, when you upgrade an OS, and change boot devices, MS expect a
> machine to not be rebooted? What a joke. Though, considering Sun can
> live-patch their kernel as of S8 with no problems, I'm sure that's not far
> behind. ;) There was still nothing untruthful about Sun's statement.

 And the "five hours" is if it crashed, and has 1000's of GB of disks to
fsck...

Kate