Improving backup stg performance of DEDUPE-DISK to OFFSITE-TAPE

[melcho]

Hello,

I have recently setup a dedup pool and I'm concerned about the throughput when running backup stg to my LTO3 tape pool. Currently there is not much data but when I start sending 1TB + to dedupe pool then how will that affect my backup stg. Are there any techniques I can use to improve the throughput.

Current env is;
TSM server 6.2.3 on AIX
16 gig RAM
dedup pool 5TB on XIV, 512 x 10gig volumes
using maxpr=3 on backup stg job
Mount Limit: 50 on devclass

Thanks in Advance,
Andrew


Server Process Start Stop Total time Wait time Active Data amount Agg. rate
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 26/05/2012 23:00 27/05/2012 0:03 1:02:49 0:03:47 0:59:02 49.1 GB 46.9 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 27/05/2012 23:00 27/05/2012 23:10 0:09:42 0:00:50 0:08:52 11.0 GB 67.9 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 28/05/2012 23:00 28/05/2012 23:39 0:38:58 0:02:33 0:36:25 27.8 GB 42.9 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 29/05/2012 23:00 29/05/2012 23:39 0:38:51 0:02:02 0:36:49 20.8 GB 32.1 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 30/05/2012 23:00 31/05/2012 0:20 1:19:50 0:01:37 1:18:13 42.3 GB 31.8 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 31/05/2012 23:00 31/05/2012 23:48 0:48:11 0:01:04 0:47:07 65.3 GB 81.3 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 1/06/2012 23:00 2/06/2012 1:05 2:05:53 0:02:12 2:03:41 100.9 GB 48.1 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 3/06/2012 17:54 3/06/2012 21:42 3:48:07 0:04:15 3:43:52 182.4 GB 48.0 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 3/06/2012 23:00 3/06/2012 23:16 0:16:01 0:09:26 0:06:35 9.8 GB 36.8 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 4/06/2012 23:00 4/06/2012 23:09 0:09:31 0:00:50 0:08:41 19.8 GB 124.8 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 5/06/2012 23:00 5/06/2012 23:42 0:42:24 0:00:54 0:41:30 21.6 GB 30.6 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 6/06/2012 23:00 6/06/2012 23:26 0:25:39 0:01:03 0:24:36 22.1 GB 51.7 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 7/06/2012 23:00 7/06/2012 23:18 0:17:56 0:01:21 0:16:35 25.7 GB 85.9 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 8/06/2012 23:00 8/06/2012 23:43 0:42:55 0:01:22 0:41:33 8.7 GB 12.1 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 9/06/2012 23:00 10/06/2012 0:29 1:29:26 0:03:57 1:25:29 75.0 GB 50.3 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 10/06/2012 23:00 10/06/2012 23:15 0:15:24 0:03:03 0:12:21 10.2 GB 39.7 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 11/06/2012 23:00 11/06/2012 23:13 0:12:58 0:01:04 0:11:54 10.5 GB 48.6 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 12/06/2012 23:00 12/06/2012 23:12 0:12:52 0:00:55 0:11:57 20.9 GB 97.4 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 13/06/2012 23:00 13/06/2012 23:12 0:11:57 0:00:59 0:10:58 21.5 GB 107.7 GB/hour
ADSM STGPOOL BACKUP - DEDUPE-DISK -> OFFSITE-TAPE 14/06/2012 23:00 14/06/2012 23:50 0:50:12 0:01:30 0:48:42 71.4 GB 85.3 GB/hour

[drivechip]

Create a instrumentation output and analyze it. As alternative you can open a call at IBM. They will do it for you. Even for a XiV it seems to be a bit slow.

[moon-buddy]

Am I reading this right?

The backup from Nodes to disk is dedup'ed and then the data on dedup'ed disk is moved to tape.

My understanding, and I hope I am wrong, is that once you do a TSM backup of dedup'ed data from disk to tape, TSM undedupes the data before sending to tape. This is where the 'slowness' comes in.

Anyone care to jump in?

[melcho]

Yes that is my understanding also. The data is deduped as a post process on the server and when you copy data from this pool to a copy pool then it is re-hydrated. I'm concerned that if this is "as good as it gets" then it will take twice as long to copy a nights worth of data (approx 2.5TB).