ADSM-L

Re: Slow restore for large NT client outcome.. appeal to Tivoli

2000-09-21 13:09:31
Subject: Re: Slow restore for large NT client outcome.. appeal to Tivoli
From: John Wiersma <John_Wiersma AT RGE DOT COM>
Date: Thu, 21 Sep 2000 12:54:10 -0400
My experience here has been that Backups fly, Restores snore.
My best rate for a restore of files of mixed length, mainly
small, has been 1.5 GB /hr from tape. (We are using IBM
3590B drives, soon to become 3590E.)
In a test, a restore from the disk pool brought
me way up to 1.66 GB/ hour. Yee ha!
Another test I did was to FTP the same data
from my client to the server where ADSM lives and
bring it back again. That got me about 6 GB/hour,
both directions.
We are using ADSM on AIX, with NT clients.
Since the backup rate is acceptable, I feel I can
rule out network and  hardware issues to a
large degree. (I suppose I could try tweaking
the read/write ratio on the RAID controller.)
We have found that ADSM is a decent solution
for 'ad-hoc' restores, (the everyday confusion users
suffer over the delete key) but large disaster
recovery type restores scare the heck outta us.
At one point when we thought we may have lost
a logical drive (don't ask!) we started a restore and
at the rates we were getting from a highly fragmented,
non-co-located library, we estimated that it was going
to take 3.4 weeks to restore 45 GB.  Not the kind of
news mgmt likes to hear.  We are now working on
turning co-location on. We hope that will bring it down
to a mere 30 hours. Yippee!

John Wiersma
Network Analyst
Rochester Gas & Electric Corporation
john_wiersma AT rge DOT com
(716) 724-8053

         RRRRRRRRRRR   !!!!!!!!!!!!!!  !!!!!!!!!!
       RRRRRRRRRRRRR !!!!!!!!!!!!!!  !!!!!!!!!!
      RRR       RRR !!!             !!!
     RRR       RRR !!!     A N D   !!!
    RRRRRRRRRRRRR !!!    !!!!!!!! !!!!!!!
   RRRRRRRRRRRR  !!!    !!!!!!!! !!!!!!!
  RRR    RRR    !!!        !!!  !!!
 RRR      RRR  !!!!!!!!!!!!!!  !!!!!!!!!!!!!
RRR        RRR !!!!!!!!!!!!!  !!!!!!!!!!!!!
<Prev in Thread] Current Thread [Next in Thread>