Technical folks often times are perfectionists, which means
they want to build reliable, fault tolerant systems. This is a good thing, but
you need to be wary of falling into the trap of overbuilding redundancy.
A wise man once reminded me that a car only has one spare
tire. This came up because we were building out our disaster recovery site and
the team thought it would be a good idea to have a clustered email environment,
on new hardware of course, at the recovery site.
Don’t get me wrong clusters are always good, but they also
cost more. Instead we decided to save the 25k that would have been used on the
servers to fund a real test of our recovery plan. In hindsight it was a much
better choice. A recovery site ideally is never used so why over spend on it.
We are actually debating on whether we should spend for the
redundant servers we do for clusters. I mean if it’s part of a cluster already
do we really need RAID drives and redundant power, or is the fact that it is
already redundant good enough. The verdict is still out, and the cost to add
RAID and a redundant power supply isn't that much when you only buy a few
servers at a time, but for someone with a huge data center, like Google or
Microsoft, the cost savings could be dramatic.
You should do the analysis on your environment and see where
technology went just a little too far, and see if you can either reduce new
purchases, or at least reuse some that you already made.
No comments:
Post a Comment