"ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU> wrote on 12/16/2005
11:07:42 AM:
> I'd like to ask those who are using the LTO2 drives with TSM 5.3 a few
> questions and some advice. I'm in the process of building a system to
> replace the one I have. It will run AIX5.3 and TSM 5.3. My plan is to
keep
> the old system around and let the data expire off but lets talk about the
> new system. Because I have little experience with 5.3 maybe those that
are
> using it can give me some advice.
>
>
>
> Because these tapes hold so much data, using collocation as it is on TSM
5.2
> seems a bit much, one tape per node. I have heard a little about the way
> collocation has changed but for those using it on 5.3 would you relay
your
> experiences and recommendations.
Collocation by group is new with TSM 5.3. It works great. I've typically
done my collocation groups based on recovery levels, so like Tier1, Tier2,
and Tier3 groups, but it really depends on how many nodes. There are
probably a 100 different ways to divide up your groups. You also have to
think about the nodes you will be restoring first and how many at once you
will be doing and think about tape contention. I've also done just 2
groups which works well. One group for the 20 or so very important high
priority servers, and the other group for the rest - I've also done
collocation by group for offsite storage using this method too.
>
>
>
> Disk pools and tape drives will not be an issue but with almost 300 nodes
> the system will be busy at times. Please relay your experience as to how
you
> run schedules. Do you have any nodes/schedules that run direct to tape?
I typically only do large TDP backups direct to tape. Say SQL, Exchange,
SAP, etc. larger than 40-80GB or so, but it depends on how many of these
large ones you have, how many tape drives you have, and how large your
diskpool is. I prefer to send everything to disk unless they are so large
it doesn't make sense or I want to take advantage of backing up to multiple
drives at once when possible.
As far as scheduling goes, I usually create one schedule for every hour or
two throughout the night and then try to spread the nodes out across all
the schedules the best I can. You always have some that need to backup at
certain times so there are exceptions. I have had schedules before with
more than 50 nodes assigned to it and everything works just fine. The
network gets bombarded at backup time, but TSM handles the load easily if
sized appropriately.
>
>
>
> Archives. Do you keep a second copy offsite or are they just sent offsite
> and recalled if needed for reclamation or retrieval of data?
Depends on how important the archives are and how often you expect to do
retrieves. In my experience archives are done mostly to satisfy some
regulatory requirement or to make some PHB feel warm and fuzzy and they are
never used for retrieves. In this case, I don't want to waste any more
tapes than I have to and send archives to a dedicated tape archivepool,
checkout the archive tapes, don't do reclamations, and only bring back
those archive tapes when they are empty. Once the archives are done those
tapes basically sit on the shelf for 5 years or whatever.
>
>
>
> Anything you would do differently if you had the chance to redo your
system?
I think the biggest issue is sizing appropriately. Most TSM headaches come
from undersized servers, library, disk, tape, network, etc.
>
>
>
> Any other recommendations with TSM 5.3?
Use the new migrate stgpool and reclaim stgpool commands in your daily
scripts/schedules. They are awesome.
>
>
>
> I have a windows server already running TSM 5.3 just so I can take a look
at
> it. No tape drives just a bit of disk to run a couple of desktop
computers
> to as a test. It is also running the admin console and my plan is to
leave
> it there and not install it on the TSM server.
>
>
>
> Thanks for the advice and help,
>
>
>
> Geoff Gill
>
> TSM Administrator
>
> SAIC M/S-G1b
>
> (858)826-4062
>
> Email: <mailto:geoffrey.l.gill AT saic DOT com> geoffrey.l.gill AT saic DOT
> com
|
