building a RAID system

Robert G. Brown rgb at phy.duke.edu
Thu Oct 9 17:17:34 EDT 2003


On Thu, 9 Oct 2003, Daniel Fernandez wrote:

> > that's unfortunate, since the main way HW raid saves CPU usage is 
> > by running slower ;)
> > 
> I cannot get the point here, the dedicated processor should take all
> transfer commands and offload the CPU why it would run slower ? In some
> tests a raid system for a single workstation ( no networking ) it's a
> bit useless (slower) unless you want to transfer really big files. In a
> networked environment there could be a massive number of I/O commands so
> should be critical.

Key word:  "should"

Benchmark results: "often does not"

Your best bet is to try both and run your own benchmarks and do your own
cost/benefit analysis.  When you say things like "better response time"
one is fairly naturally driven to ask "does the difference matter", for
example.  Given that we run over 100 workstations from a SW RAID with
nearly instantaneous (entirely satisfactory) performance, you'd have to
really be hammering it to perceive a difference.

> In fact we have two choices:
> 
> 	- Use an spare existing ( relatively obsolete ) computer and couple it
> with a HW RAID card.
> 
>         - Spend on a fast CPU computer and a good but cheap Parallel ATA
> controller.

Or a cheap computer + PATA or SATA controller.  Even a cheap computer
has 2+ GHz CPUs and hundreds of MB of RAM these days. Spend more on
what you put the disks in, power, cooling.

If it is an old/obsolete computer, will it have enough power, enough
cooling?  Regardless, the disk cost itself will dominate your costs.

   rgb

> 
> 
> > _______________________________________________
> > Beowulf mailing list, Beowulf at beowulf.org
> > To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> > 
> 
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
> 

-- 
Robert G. Brown	                       http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567  Fax: 919-660-2525     email:rgb at phy.duke.edu



_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list