Ignoramus3187 wrote:

On 2009-02-05, Pete C. wrote:
I heard a lot of that "mission critical" talk. I do not think that it
is worth the exorbitant amount of money, given that Linux already
gives us the reliability. We have enough scripts written for quick
failover.

Mission critical service includes such things as on site spare parts
inventory, guaranteed on-site response and system repair times, etc.

We have spare parts on site too, and computers available to take over,
I do not see what is the big deal about "parts". We have several
people who are proficient with screwdrivers and can replace said
parts when necessary.

Google runs its service using very crappy computers that are
the polar opposite of "mission critical". It does just fine due to
software based failover.

This was my objective as well, to be relatively reliable, without
paying too much money to "enterprise vendors".

Different enterprises have different standards for reliability,
different costs associated with failures and different regulatory
requirements (or no regulatory requirements).


All critical systems already have multiple levels of redundancy to
ensure rapid service recovery in the event of a system loss or even
a site loss. Mission critical service isn't to ensure uninterupted
service, proper systems design ensures that, mission critical
service ensures a minimum exposure to reduced redundancy.

Yep.


Where I work we have many 10s of thousands of servers (no joke), and
among those we have like a dozen Linux systems that were an experiment
in their viability in the enterprise space. That experiment was some
years ago and there has been no expansion in the Linux space, even
though the company is very big on cutting costs.

Depending on what you do exactly, which you obviously do not have to
state here, you could save a "very large" amount of money. Google can
do it.

If 30+ servers have not crashed in 6 months, that tells me that their
reliability is decent. I reboot them every week, automatically of
course, just in case.

We work to "five nines" i.e. 99.999% availability standards, so not
crashing in six months isn't enough to meet that standard. My systems
regularly run years without a crash, and do not require weekly reboots
either. The reboots during routine maintenance every few months are
plenty. The bulk of failures are transparent such as a failure of a disk
that is mirrored and automatically replaced with a hot spare so the
replacement of the failed disk is at our leisure.

You must be paying a lot of money for this.

Yep. I hear the power bill alone for just one DC is around $250k/mo and
there are a lot more than one.


There is two approaches to having reliability, one with expensive
hardware, and the other with failovers. The latter is cheaper, and is
good enough for us.

The last big trouble we had, was from a failed circuit breaker. (one
lug loosened and heated)

Redundant power supplies in systems and storage, redundant PDUs,
redundant static transfer switches, redundant UPSes, redundant
generators. Redundant network connections, redundant storage
connections, redundant mirrored and hot spares disks, physically
separate fallback sites, etc.

On my personal stuff at home I have several UPSes, a couple generators,
a couple spare cable modems, routers, network switches, dialup for
additional backup, etc.


Perhaps one of these days it will win one on my shootouts.

If you want to try Ubuntu, I can help you set it up.

I'd need to find an application for it.

The last big shootout I did was with Linux/EMC vs. Windoze/Mach3 and
Mach3 won. That was before EMC2 was out, so perhaps EMC2 would have done
better in the shootout. In either case, I settled on Mach3 and it
continues to do everything I need and given my HSM needs, one control PC
can control multiple machines.

My home web/mail/storage server is also running Windoze, and given it's
4.5 years of continuous service without a crash and without being
compromised, it seems to be meeting the requirements just fine as well.
I am considering replacing the old Dell desktop it runs on with a little
Shuttle mini PC as I expect it would pay for itself in power savings in
about a year, as well as being a higher performance system.

--
Due to extreme spam originating from Google Groups, and their inattention
to spammers, I and many others block all articles originating
from Google Groups. If you want your postings to be seen by
more readers you will need to find a different means of
posting on Usenet.
http://improve-usenet.org/