> I don't see how it's possible to get better performance by
> aggregating all of the disks and then partitioning them out.
It's not that you always get better performance by aggregating, it's
that you _don't_ always get better performance the other way, while
always adding complexity.
If 'Application 1' and 'Application 2' have exactly identical disk I/O
demands, creating identical dedicated arrays 'RAID 1 Array 1' and
'RAID 1 Array 2' can be considered to have basically the same
performance as one RAID 10 array, albeit with surplus complexity. If
'Application 1' and 'Application 2' have asymmetrical demands,
however, that setup rewards one unnecessarily and punishes the other.
In another message, I noted situations in which separate arrays would
be preferable.
> I always thought the rule was to separate logical processes across
> separate spindles whenever possible and practical.
I'd phrase the rule "Give as many spindles as possible to each
independent function," so in some cases the spindles may be shared
between functions, but each function can use more of them, which both
limits contention by concurrent processes and also enables a function
to perform optimally when there is no contention (i.e. asymmetry).
But I think the rule is really less a hardware requirement than an
_awareness_ requirement--what you'll do when you have the need and/or
money. It's essential to realize what the current/future bottlenecks
might be on a server, even if you don't take action right now, which
means dissecting the applications you currently run and those you
might pop in.
> Kind of like the *nix people bashing the M$ people.
Kind of like the term 'M$'...:o)
--Sandy
------------------------------------
Sanford Whiteman, Chief Technologist
Broadleaf Systems, a division of
Cypress Integrated Systems, Inc.
e-mail: [EMAIL PROTECTED]
SpamAssassin plugs into Declude!
http://www.mailmage.com/download/software/freeutils/SPAMC32/Release/
---
[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]
---
This E-mail came from the Declude.JunkMail mailing list. To
unsubscribe, just send an E-mail to [EMAIL PROTECTED], and
type "unsubscribe Declude.JunkMail". The archives can be found
at http://www.mail-archive.com.
