HP3000-L Archives

October 2000, Week 2

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:
Reply To:
Date:
Wed, 11 Oct 2000 14:31:59 -0500
Content-Type:
text/plain
Parts/Attachments:
text/plain (41 lines)
Steve Belkacem wrote:

> Hi people,
>
> 1. We currently use mirroring on out production
> machines.

I don't know how you are using mirroring but it should work better than
your experience. Of course it doesn't work on system volume sets. Is it
configured correctly, have you reviewed you configuration with the
response center..

>
>
> 2. 12H Autoraid.

In response to some of the comments about AutoRaid. It will not perform
as well as a XP256 but it cost a lot less(30K vs 300K).  We are working
on allowing the second controller to be used. The issue with disc space
on a AutoRaid is the same as a MODEL 20 except the AutoRaid doesn't make
you manage the device directly. If all available space is configured
than AutoRaid uses mostly RAID 5 which would require 25% of all space to
be allocated to parity data(the same as the model 20 in Raid 5). If you
want the AutoRaid to operate in RAID 1 mode(hardware mirroring) then you
only allocate about 55 percent of the available space and the Auto Raid
uses RAID 1 for all data(the same as the model 20 in Raid 1 mode).

You should always assume that more spindles get better performance. The
memory in the AutoRaid is used for holding the mapping tables and to
allow some write optimizations. You should not assume the AutoRaid is a
cache device like a XP256. A six spindle AutoRaid would perform worse
than a twelve spindle AutoRaid if the I/O load was "high". We have seen
that in some configurations, restores performs slower on AutoRaid than
on JBOD and we have documented this with a workaround which is to use
Performance Mode during restores...

Of course with all performance data "IT DEPENDS"

Steve Macsisak
MPE/iX Performance Consultant

ATOM RSS1 RSS2