HP3000-L Archives

December 2000, Week 1

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Dennis Heidner <[log in to unmask]>
Reply To:
Dennis Heidner <[log in to unmask]>
Date:
Sun, 3 Dec 2000 02:31:31 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (23 lines)
Raid 1 can indeed reduce the "apparent" read access time.  But the
"apparent" write time also increase - but on most HP e3000, the write time
isn't a problem.  The writes are handled in the background by XM, so you
are a little removed from the actual impact.

The problem with read requests on Raid-5 is that it actually can't return
the data until it has gotten the data off the corresponding data stripes
and verified the parity.  Which means the drive seek and rotation times
become much more noticeable.  This can be fixed by using special drives
that can have the spindles synchronized so that the sector 0 is under the
heads for all the drives at the same time.    Most Raid-5 controllers have
a pretty good size cache,  which means writes can appear to complete very
quickly --- while the controller actually writes the stripes.

In the PC world, there are several pretty good disk controllers that allow
you to choose Raid-1 mirroring or Raid-5 striping.  I believe that both the
Promise and Adaptec, when operated in a mirrored disk mode will actually
return the data from which ever drive responds first -- similar to Mirrored
disk/iX.  Most of the PC raid adapters also provide utilities so that you
can be alerted when a drive fails.

A good reference site is www.raid5.com,  look for the "raid.edu" tab.

ATOM RSS1 RSS2