Robert McCarthy wrote:
Networker says my max parallelism on my server is 32, it won't allow
anything higher.
This depends on what edition of NetWorker enabler you have purchased.
Basically you pay more for a higher server parallelism. The parallelism
also increases when you add storage nodes.
Another thought.. Quantum's website says the SDLT320 will write 16
mb/sec uncompressed, thats about 1gb per minute roughly. That is almost
10 hours to write 500gb of raw data, maybe my backups will just be that
slow.. I don't know the inner workings of Networker with compression and
such though.
You have two choices. Client side parallelism compresses the data before
it hits the network, but can cost you in CPU cycles on the client.
Compression on the tape device is better in most cases IMHO. If you use
both then you won't get any benefit from the tape drive's compression
circuitry, and you will appear to write data at your tape drive's native
speed and capacity. Don't forget that the data has already been
compressed at that point, so your 600GB volume might only show up in
NetWorker as 300GB. YMMV.
The suggestions of parallelism values are largely irrelevant here, since
that is not what is causing you a problem. You are having a problem
because your volumes are unworkably large, plain and simple. NetWorker
only takes one stream from a volume. This is the way it has always been.
It has been discussed many times, but so far Legato have not come up
with a feature to solve this. It was originally that way because there
was no point in taking two streams from a single spindle, however with
volume manager software this is rarely a problem these days. The usual
workaround is to define smaller save sets at directory level, but the
problem is that is restricts the way you arrange your data on the
volume, and if someone adds a directory you need to update the backup
configuration or the data won't get backed up. The workaround to the
workaround is to define a directive that excludes the defined save sets
and picks up any stray data. Check the archives for details.
At 08:09 AM 4/1/2004, you wrote:
Client Parallelism is a hidden option in your client resource. So if
you want to use 3 devices parallel with that client, set it to a value
of 12. The other question is, can your server create enough IO to
serve the possible native writespeed of your DLTs.
Device parallelism is not a hard limit, it is just used for
loadbalancing.
You should set your server parallelism general to a higher value.
Better set it to 256
regards
Klaus
-----Ursprüngliche Nachricht-----
Von: Legato NetWorker discussion
[mailto:NETWORKER AT LISTMAIL.TEMPLE DOT EDU] Im Auftrag von Robert McCarthy
Gesendet: Donnerstag, 1. April 2004 14:53
An: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Betreff: Re: [Networker] AW: [Networker] Large Volumes & Speed..
Client Parallelism? Is there an option I don't see it in the client
record.
Device parallelism is 4 per drive. (this was the default)
Server parallelism is 24 (I changed it by guessing 6 drives at 4 makes
24)
Thanks, Rob.
At 07:39 AM 4/1/2004, Grohal Klaus wrote:
>What's the value of your client parallelism, device parallelism and
server
>parallelism?
>
>Regards
>Klaus
>
>-----Ursprüngliche Nachricht-----
>Von: Legato NetWorker discussion
[mailto:NETWORKER AT LISTMAIL.TEMPLE DOT EDU] Im
>Auftrag von Robert McCarthy
>Gesendet: Donnerstag, 1. April 2004 14:15
>An: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
>Betreff: [Networker] Large Volumes & Speed..
>
>
>I have some really large volumes that I need backed up on a regular
basis,
>there are 6 volumes around 600Gb each. I added them to my normal
cycle but
>when it hits a full volumes it takes days to backup. I currently run a
>Windows 2000 server with two Dell PV-136T (Dell branded Adic Scalar 100)
>with 3 SDLT320 drives per box. It seems Legato only writes to one
tape at
>a time per volume, is this by design? It has six drives available to
use,
>but only seems to write different clients to the six drives in
>parrallel. Any suggestions on speeding up large volume backups? I run
>Gigabit copper to my backup server, I feel the bottleneck is with Legato
>and possibly the SDLT drive write speed may be a factor. Any
thoughts will
>be appriciated.
>
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
