Re: SAN Array Setup for Various Boxes
- From: "Hass, Harold" <HHass@xxxxxxxxxxxxxxx>
- Date: Mon, 19 Dec 2005 15:05:47 -0800
I doubt if you're going to find a book that will answer the question to
your satisfaction for your environment. The variables are very wide.
For example, the number and types of switches and the ports on your SAN
fabric; the number of front-end (host side) adapters; the number and
type of back-end (disk side) adapters; the disk drive types and
arrangement; the disk device (LUN) makeup and, ultimately, the number of
spindles. On the host side, the type and number of HBA's and how they
access (load balancing etc.) are important as is the type of data being
accessed.
My experience has been that the performance of a good SAN will be
markedly better than any locally attached disk being migrated from.
Again, it's a hard thing to predict. We've seen measurable through put
increases of 100-200% over SSA drives when migrating medium sized Oracle
database applications to an EMC Symmectrix. Hard to measure as to how
much is application, how much is raw feeds & speeds, but it's "lots"
better.
Disk space volume isn't the whole story. Reduncancy, performance, and
most importantly, manageability, are all a part of it. Is a SAN fast,
reliable, manageable? Yes. Is a SAN expensive and a lot of work to
integrate? Yes. Is it worth it in your environment? Who knows!
Research it and keep an open mind.
Harold Hass
Senior Network Systems Engineer
County of Fresno, CA
-----Original Message-----
From: IBM AIX Discussion List [mailto:aix-l@xxxxxxxxxxxxx] On Behalf Of
Mark Monroe
Sent: Monday, December 19, 2005 12:55 PM
To: aix-l@xxxxxxxxxxxxx
Subject: SAN Array Setup for Various Boxes
To All,
Can someone recommend a good book or books that gives recommendations on
array layouts and SAN setup etc?
Few questions if you don't mind.. <smile>
The main purpose for a SAN is to take a large repository of disk drives,
create redundant arrays and then make this disk space available to
several machines. As such, the arrays and drives will be simultaneously
providing data for multiple boxes and applications.
Here is my quandary, and perhaps some good books would clear this up for
me, but, I am concerned about any one particular box, such as a heavy
production SQL or DB@ database box hogging the array and creating
performance problems for other boxes. Sure, I could create multiple
arrays for all of the various boxes, but heck if you are going to that,
what's the point of having a SAN?
Is the performance of something like a DS4300 with the turbo option and
fiber channel drive so good that it can service 6-8 production systems
and give them all excellent performance? or should this type of
environment just stick with local boxes for each machine?
At what point do the number of machines, in general, begin to saturate
the fiber connections to the SAN? 10 machines? 20 machines? 30
machines? etc?
Thoughts/Ideas?
Mark
- Prev by Date: Re: migratepv pp
- Next by Date: one smb question
- Previous by thread: Re: SAN Array Setup for Various Boxes
- Next by thread: one smb question
- Index(es):
Relevant Pages
|
