[prev in list] [next in list] [prev in thread] [next in thread] 

List:       toasters
Subject:    Re: RAC install on NetApp
From:       tmacmd () gmail ! com
Date:       2008-03-21 16:30:53
Message-ID: 179510826-1206117051-cardhu_decombobulator_blackberry.rim.net-639211888- () bxe005 ! bisx ! prod ! on ! blackberry
[Download RAW message or body]

http://www.specbench.org

Remember to look at the "real" numbers. 
In other words, how many filesystems are used
(1 in the case of the 24 node GX cluster, 2 in the case of most if not all netapp \
clusters, 1 in the case of nearly all the rest of the netapp cases, everybody else \
usually use an obscene/unrealistic number of filesystems)

The one real flaw I think is in this benchmark is that it pretty much evenly \
distributes I/O evenly across all filesystems.  Sent from my Verizon Wireless \
BlackBerry

-----Original Message-----
From: "Robert K. Borowicz" <rob-7704@austin.rr.com>

Date: Fri, 21 Mar 2008 10:32:05 
To:toasters@mathworks.com
Subject: Re: RAC install on NetApp


Toast{ed} types,

What is the benchmark site or reference paper I should look at for NFS 
perf numbers for filers?

The below scenario is getting interesting in that the NFS clients (RAC 
hosts) are Sun T2000's with known on-board GigE driver issues. We are 
patching as I type...

As a brainless test this morning I took *off* the "noac" mount setting 
and get 4-5X the perf then with it.

I see myself going down that fabled path of "well if its a 3020 with 
GigE and CAT 6509 infra no jumbo frames you *should* get X perf"  It 
would be good to compare with some conservative and reasonable numbers 
somebody else already wrote down.

And no EMC guys allowed to play here.

;-)

TIA as always,

-Rob

> 
> 
> -----Original Message-----
> From:   Robert K. Borowicz [mailto:rob-7704@austin.rr.com]
> Sent:   Wednesday, March 19, 2008 10:32 AM Pacific Standard Time
> To:     toasters@mathworks.com
> Subject:        RAC install on NetApp
> 
> Toasters,
> 
> My DBA just came to me and complained that a Ora RAC 10g build took 18
> hours when using a Filer NFS mount. Seems I've heard of this issue
> before but forget the fix. Has someone seen this before ?
> 
> TIA
> 
> -Rob
> 


[prev in list] [next in list] [prev in thread] [next in thread]