What kind of [RAID] system works best?

[David Collier-Brown]

Bill Eldridge <bill at rfa.org> wrote: 
> I'm currently testing out some SCSI->IDE RAID 5 systems
> (not banging on them too rigorously at the moment, but
> no obvious flaws so far).
> 
> Basically, you get a SCSI interface that the computer sees,
> but the enclosure holds 3 IDE drives, and the whole thing
> snugs into a tower chassis.  RAID 5 at about $2200 for
> 12 Gigabytes (using 6 Gig Seagates).

	Not too shabby!

	A caveat, though, about speed vs capacity:  RAID
	5, when writing, is roughly as fast as a single disk.
	Reading, it's a factor of how many disks you have.
	I usually estimate read speeds of 7200 rpm disks
	as (number of disks * 560 KB/S), so with three
	drives in raid 5, read speed is 2*560 or 1120 KB/S
	(I don't count the parity drive, of course).

	For example, a simple stripe of 6 7200 rpm disks
	implies 560 * 6 = 3,360 KB/S reads of random data.
	Raid 5 will give you 2,800 KB/S reads and ~360 KB/S 
	writes (modulo buffering & write-behind, which trade off
	reliability for speed)

	This is slower than the transfer speeds you usually see quoted,
	because it's based on observed start-i/o rates on real
	equipment, and 8 KB buffering.  In turn, these are limited
	by physics: the number of bits the head flies over per second,
	which are strictly limited by RPM and bit density.
	Vendors may disagree (:-))

--dave
-- 
David Collier-Brown,  | Always do right. This will gratify some people
185 Ellerslie Ave.,   | and astonish the rest.        -- Mark Twain
Willowdale, Ontario   | http://java.science.yorku.ca/~davecb
Work: (905) 477-0437 Home: (416) 223-8968 Email: davecb at canada.sun.com