|
PureBytes Links
Trading Reference Links
|
There are two distinct issues to consider. Hardware redundancy, and backup.
Most of the comments posted that are against a simple RAID 1 implementation
use backup as a reason not to use it and there is clearly a misgiving about
cost and reliability.
Given that one’s PCs are used for a mission-critical application, hardware
redundancy should be a given, in my opinion. However, most of the (hardware)
discussion centers around disk redundancy. To deny that RAID is “THE”
preferred solution is to disagree with probably all mission-critical
practices. I've swapped about 5 failed disks in RAID 1 configs at one time
or another. There are some hardware prerequisites for this to work exactly
as intended, but nothing unordinary. It’s always worked for me.
Focusing on PC redundancy on the one hand and backup on the other should
sought the wheat from the chaff. Google’ing around will locate clustered
solutions for less than $5k. Clustered meaning if one computer fails the
other takes over, in real-time. If that seems expensive to ensure around 99%
uptime, you really shouldn’t be running anything mission-critical (such as
trading for a living). Suffering disaster is always substantially more
expensive than appropriate planning and deployment.
Backup (on the other hand) is an entirely different function to a non-stop
paradigm. Using a tool like Casper has its shortcomings. For example, you
can't image on the fly to another computer. The image must be to a slave
disk in the same PC, and Casper cannot change image properties--image size
for example. Imaging should ONLY be done when an OS isn't running--at the
lowest level if you will. Casper cannot safely/totally/100% image. I have a
copy and have tried lots of scenarios and it doesn’t pass muster in my
opinion. The easiest to use imaging software I've found (and use) is
Partition Commander (PM, www.v-com.com). For those who want to use imaging
for backup, PM can image a RAID 1 to a single removable/portable slave disk
in the same PC. You can pull-out the image-disk and take it offsite if you
wish.
The backup solution I like the most is "Backup for Workgroups" (BFW). It
also has a mirroring function, a disaster recovery from CD tool, and a
roll-back function. The latter should appeal to anyone running
mission-critical apps who feel they can’t recover damaged data and/or
programs on the fly. I maintain 25 versions of changed files for a period of
(up to) 180 days. If I’m hit by a virus, which hasn’t happened since
installing a “hardware” firewall a while back, I can selectively choose
which files to rollback and to what date. It only takes a few minutes.
Rather than writing about BFW, take a look at www.lockstep.com. A well
thought out backup and disaster recovery policy does wonders for those
dreaded real-world situations (that do occur).
An interesting exercise (I think), would for those interested to post their
ideal requirements in terms of redundancy, backup, performance, with
examples of components, and see what a consensus ballpark cost might be. My
guess is $4-6K. Whoever posts first, if at all, perhaps use this subject”
“Backup, Redundancy, Performance and Cost”
Colin West
|