[KLUG Hardware] Re: IDE RAID

[Bryan J. Smith]

On Fri, 2002-05-24 at 10:47, Richard Zimmerman wrote:
> I've NEVER dealt with RAID before so if I were to setup a server based on
> RAID, would I place /boot /swap and /tmp on Raid 1 and the rest on Raid 5,
> right?

Well, since I usually have 40-80GB usable in RAID-1 (or RAID-10), I
setup all my system partitions on that disk -- /, swap, /tmp, /var,
/usr, /var/spool, /var/lib, /var/www, /usr/local, etc... (those are the
typical filesystems I make on a server).  The RAID-5 volume is then
/home/(servername), /home/cdimages, /home/porn^H^H^H^Hmp3, etc... as I
like.  ;-P

Since 3Ware != ATA from the standpoint of the PC BIOS Int13h services,
you there is _no_need_ for a separate /boot as long as / is in the first
1TiB (1TiB = 1,024 GiB ;-).

> Also, I bought a Promise Fasttrak 100 TX2 RAID card (used). From what
> I've been reading here, am I correct it is NOT a True RAID card (as I
> originally thought)?

It is the _exact_same_product_ as the Promise Ultra 100, except you paid
$50 more for it.  ;-P

I'll send you my Draft article "Dissecting ATA RAID Options" off-list. 
Anyone else who wants a copy, just let me know (current Draft is 1.9 --
still a little ways to go until completion). 

> P.S. Thanks for mentioning www.package2you.com. Pretty interesting stuff on
> that website... I've always wanted a duel processor system on the cheap...
> Looks like I've found several options for it! :-)*
> *I'm not looking for the speed, just gaining the experience of using a duel
> processor system.

Dual processor is bliss, especially with Linux 2.4.  Ignore benchmarks,
the "responsiveness" is the key.  You can use the "pre-empt" patch to
get the same out of uni-processor systems, but it's still not as good. 
For desktops at least.

When it comes to servers, I/O _throughput_ is _just_as_important_!  I
love seeing people buy a "cheap" dual-processor with a single memory and
PCI channel and wonder why their server "hangs" when they are doing a
lot of disk I/O.  Put dat NIC and disk controller on _separate_ PCI
busses!

-- Bryan

-- 
The US government could be 100x more effective, and 1/100th the
Constitutional worry, if it dictated its policy to Microsoft as
THE MAJOR CUSTOMER it is, and not THE REGULATOR it fails to be.
---------------------------------------------------------------
Bryan J. Smith, SmithConcepts, Inc.   mailto:b.j.smith@ieee.org
Engineers and IT Professionals     http://www.SmithConcepts.com