ADSM-L

[ADSM-L] AW: TSM only reads from COPY1 during DB backup

2007-04-26 04:26:47
Subject: [ADSM-L] AW: TSM only reads from COPY1 during DB backup
From: Salak Juraj <J.Salak AT ASAMER DOT AT>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 26 Apr 2007 10:25:33 +0200
I don´t think one can generally state that.

If it were generally true than raid-1 controllers would have generally 
to be configured to read from one disk only 
in order , to paraphrase your words, 
not to cancell all gains obtained with disk-embedded read-ahead algorhytmus....

I believe such test results strongly depends on many different configuration 
issues,
like raid implemented in one or two controllers,  channel architecture used &
channel amount, read block size, raid buffer size, disk buffer size, read-ahead 
alghorythmus
both on controller and on the disk, parameter setting etc..

Undoubtly, 
theoretical high performance limit reading from one disk is the disk´s 
streaming speed,
theoretical high performance limit reading from two disks is twice as high.

As always, in reality you only can converge to that limits but not reach them 
fully,
and the more complicated the solution is (e.g. 2 disks against 1 disk) 
the larger the gap between real performance to the theoretical limit will be.

The test mentioned by you apparently falls 50% under the theoretical 
performance limit,
whis is rather bad, so I believe there either must have been a configuration 
issue, 
or one of the used components was of insufficient quality (maybe the raid 
firmware).

best

Juraj








> -----Ursprüngliche Nachricht-----
> Von: ADSM: Dist Stor Manager [mailto:ADSM-L AT VM.MARIST DOT EDU] Im 
> Auftrag von Paul van Dongen
> Gesendet: Mittwoch, 25. April 2007 21:55
> An: ADSM-L AT VM.MARIST DOT EDU
> Betreff: Re: TSM only reads from COPY1 during DB backup
> 
>    If it was't a dream, I remember reading something about 
> this issue. It stated that this behaviour was "by design", 
> because reading blocks from both copies of a mirrored volume 
> cancelled all gains obtained with controllers and disk 
> subsistems read-ahead algorhythms.
>    I even remember reading it was actually tested.
>    
> Paul
> 
> -----Original Message-----
> From: ADSM: Dist Stor Manager on behalf of Orville Lantto
> Sent: Wed 4/25/07 16:22
> To:   ADSM-L AT VM.MARIST DOT EDU
> Cc:   
> Subject:      Re: TSM only reads from COPY1 during DB backup
> 
> Performance is the issue.  As tapes get faster and faster, 
> trying to get a db backup without "shoe-shining" the tape 
> drive gets harder.  Using storage sub-system mirroring is an 
> option, but not the recommended one for TSM.   Perhaps there 
> is a sound technical reason that db reads cannot be made from 
> both sides of the mirror, but it could be that it was just 
> programmer "convenience".  Either way, I will have to design 
> around this "feature" to squeeze a bit more performance out 
> of my storage.
>  
> Orville L. Lantto
> Glasshouse Technologies, Inc.
>  
> 
> ________________________________
> 
> From: ADSM: Dist Stor Manager on behalf of Richard Sims
> Sent: Wed 4/25/2007 10:49
> To: ADSM-L AT VM.MARIST DOT EDU
> Subject: Re: [ADSM-L] TSM only reads from COPY1 during DB backup
> 
> 
> 
> On Apr 25, 2007, at 10:55 AM, Orville Lantto wrote:
> 
> > Reads should be safe from mirrored volumes and are commonly done in 
> > operating systems to load  balance.  Not taking advantage of the 
> > available IO resource is wasteful and puts an unnecessarily 
> unbalanced 
> > load on an already IO stressed system.  It slows down db 
> backups too.
> 
> Then your issue is performance, rather than database voracity.
> This is addressed by the disk architecturing chosen for the 
> TSM database, where raw logical volumes and RAID on top of 
> high performance disks accomplishes that.  Complementary 
> volume striping largely addresses TSM's symmetrical mirror 
> writing and singular reading.  Whereas TSM's mirroring is an 
> integrity measure rather then performance measure, you won't 
> get full equivalence from that.
> Another approach, as seen in various customer postings, is to 
> employ disk subsystem mirroring rather than TSM's application 
> mirroring.  In that way you get full duality, but sacrifice 
> the protections and recoverability which TSM offers.
> 
>     Richard Sims
> 
Diese E-Mail enthält vertrauliche und/oder rechtlich geschützte Informationen. 
Wenn Sie nicht der richtige Adressat sind oder diese E-Mail irrtümlich erhalten 
haben, informieren Sie bitte sofort den Absender und vernichten Sie diese Mail. 
Das unerlaubte Kopieren sowie die unbefugte Weitergabe dieser Mail ist nicht 
gestattet.

This e-mail may contain confidential and/or privileged information. If you are 
not the intended recipient (or have received this e-mail in error) please 
notify the sender immediately and destroy this e-mail. Any unauthorised 
copying, disclosure or distribution of the material in this e-mail is strictly 
forbidden.

<Prev in Thread] Current Thread [Next in Thread>