ADSM-L

Re: [ADSM-L] cloud pools, directory containers, dedup etc.

2016-03-23 10:29:20
Subject: Re: [ADSM-L] cloud pools, directory containers, dedup etc.
From: Stefan Folkerts <stefan.folkerts AT GMAIL DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 23 Mar 2016 15:27:56 +0100
I would advise you to read the blueprint documents, they can be found here :

https://www.ibm.com/developerworks/community/wikis/home?lang=en#!/wiki/Tivoli%20Storage%20Manager/page/IBM%20Spectrum%20Protect%20Blueprints

They give some good sizing guidelines.
TSM active log, database, memory and cpu impact on dedup is very
significant.
I would not advise any form of dedup without the TSM database and activelog
on SSD's, i've seen too many real world problems with that.
But with the proper CPU, memory and db/activelog performance it will work
fine.
The new containerpools are a huge improvement and you no longer need 10K
disks for very large environments.

Inline deduplication takes 10Gb/s (true 1GB/s) and reduces it inline with
the M spec system at around 20-25% CPU util on my 20 session test runs.
With inline compression this can rise towards 50%

The cloud proposition is one I see as a replacement for deep tape storage
solutions, it is not a performance solution.
So if you are currently using tape for large database backups and they
require high restore  performance I would think twice about moving a the
cloud solution since you will be placing the primary pool in the cloud,
there is no copypool option as of yet.

A good alternative can be a replication solution in the cloud that
companies offer, it can often be used for DR as well.
This replaces a copypool on tape solution and you get good speed from the
disk solution on site.
With good I mean good for dedup on disk with all it's limitations.




On Tue, Mar 22, 2016 at 7:35 PM, Lee, Gary <glee AT bsu DOT edu> wrote:

> Just jumping into the cloud storage arena with tsm.
> Just installed server 7.1.5 to experiment with cloud storage.
> We have been mandated to replace our largely tape based environment with
> amazon cloud storage.
>
> I know very little of tsm's implementation of this, and how it may be used.
>
> I see lots of mention of dedup, of which I am somewhat scared. I have
> always preferred such things as dedup and compression be performed at the
> storage hardware level.
> So,this all prompts several questions.
>
>
> 1.       Are there use case docs or best practice docs for implementing
> cloud storage in tsm?
>
> 2.       What is the impact of dedup on server performance?
>
> 3.       What impact does dedup have on database and logging?
>
> 4.       Impact on server recovery and disaster protection?
>
> 5.       Can we start off without doing dedup, then when comfortable with
> it implement it?
> That will be enough to start with.
>
> Also, has the administrator's guide returned for 7.1.5?
> I didn't see it anywhere.
>
> Thanks for any and all help.
>

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