Re: SAN or NAS ?
From: Mike Naime (mnaime_at_kc.rr.com)
Date: 02/19/04
- Next message: Main, Kerry: "RE: OpenVMS thrashes slowaris and aix in cluster tco/virus study"
- Previous message: sms_at_antinode.org: "Re: DIR /PAGE output differs between SET HOST (RTA) and RSH (TNA)?"
- In reply to: John Brandon: "SAN or NAS ?"
- Next in thread: Rob Young: "Re: SAN or NAS ?"
- Reply: Rob Young: "Re: SAN or NAS ?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Thu, 19 Feb 2004 04:21:05 GMT
John Brandon <brandon@dalsemi.com> wrote in message
news:04021808084765@dscis6-0.dalsemi.com...
> Just wanted to see what the opinions of SAN or NAS were ...
>
> I have a SAN environment on my VMS servers.
>
> The UNIX group has their NAS.
>
> They (UNIX) run ORACLE and other database related applications on their
NAS.
>
> They (UNIX) think it a good idea - saves money - etc.
Saves money yes! but performance is limited to your already overburdened
network pipe.
> Even one (UNIX) states that there is no difference between SAN and NAS.
LOL
>
>
> Or is the joke on me?
The joke is on them definitely.
Andrew did a good summary, now to add my bit.
All that I can say is that the Unix group either does not require the
throughput to the database that you get from a SAN (Dedicated storage) or
they are happy with high disk queue wait states on a TCPIP network pipe.
:-)
We had a recent issue at work with the AIX systems on our SAN. After some
close scrutiny of SAN performance and what to tweek, it turns out that the
AIX Unix systems access the same storage controllers in a radically
different manner than VMS does. VMS is fine with mirror sets. AIX want 6
drive raidsets. I have the same Application running on both AIX and VMS
systems. The AIX system has a higher disk throughput requirement to perform
the same database queries. The VMS system is not complaining of throughput
problems where the AIX system says that my pipe is 100% full.
HSG vs EVA is night and day. I was transfering a complete production system
from HSG LUNS to EVA LUNS today. I monitored the disk que length while
doing this. (actually I ran it twice just to collect the stats the second
go-round) The HSG's hit 250 to 300 as a max value where the EVA was 1.
:-)
Using a 1GIG SAN fabric, we measured about 20M/sec LUN throughput on the
HSG's (Max for that LUN, not the controller) We got 80M/sec on the EVA. I
still need to test throughput on our 2Gig fabric.
Unless your network backbone is also 1GIG, they cannot compete with the
throughput of a SAN. Even if they are at 1GIG, you could be at 2 GIG today.
10 GIG in a year or two. (I forget the timeline)
> J*o*h*n B*r*a*n*d*o*n
> VMS Systems Administrator
> firstname.lastname.spam.me.not@dalsemi.com
- Next message: Main, Kerry: "RE: OpenVMS thrashes slowaris and aix in cluster tco/virus study"
- Previous message: sms_at_antinode.org: "Re: DIR /PAGE output differs between SET HOST (RTA) and RSH (TNA)?"
- In reply to: John Brandon: "SAN or NAS ?"
- Next in thread: Rob Young: "Re: SAN or NAS ?"
- Reply: Rob Young: "Re: SAN or NAS ?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|