ADSM-L

Re: Improving NT client throughput

2001-06-22 11:57:21
Subject: Re: Improving NT client throughput
From: "<James> <healy>" <James.Healy2 AT AXACS DOT COM>
Date: Fri, 22 Jun 2001 11:58:22 -0400
Vint,
     Haev you tried running and "incremental by date only"?




"Vint Maggs" <vint.maggs AT SRS DOT GOV>@VM.MARIST.EDU> on 06/21/2001 10:29:03 
AM

Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>

Sent by:  "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>


To:   ADSM-L AT VM.MARIST DOT EDU
cc:

Subject:  Improving NT client throughput


I have a bunch of Windows 2000 clients (file servers) whose backups are
really dragging.  My server is TSM 3.7.3 on Solaris 2.7. The clients are
running TSM v 4.1.2.12.  They have 2 Xeon 700Mhz CPUs and 512 MB RAM,
100Mb NICs on a gigabit backbone.  Each one of these clients has several
hundred thousand files, one is in excess of 1 million files.  TSM
initiates the backup at midnight but it is not until 5 hours later that
these clients really begin pumping data.  I assume this is because it is
taking so long to analyze the files.  If I had my way they would repartion
the drives on these clients so that all these files are not contained in a
single filespace, but that isn't going to happen.

Is there any way to improve performance here?  We are considering the
following options, please give me your opinions.

On Clients:
set changinretries to 1
set tcpbuffersize from 62 to 512
we have already confirmed the tcpnodelay, largecommbuffers, and
memoryefficientbackup options are in our favor.

On Server:
resetting idletimeout from 60 to 15 (default)
resetting commtimeout from 360 to 60 (default)

Thanks,
Vint
<Prev in Thread] Current Thread [Next in Thread>