ADSM-L

Re: Token Ring Question

1998-05-11 03:58:11
Subject: Re: Token Ring Question
From: "Lambelet,Rene,VEVEY,FC-SIL/INF." <Rene.Lambelet AT NESTLE DOT COM>
Date: Mon, 11 May 1998 09:58:11 +0200
Alan,
we also got results like yours.

But, in order to better manage nodes and filespaces on the ADSM server,
we decided to install the ADSM agent on every Netware server.

Exception: we launch all restores from one Netware server and use the
"-fromnode" option. This to simplify security at the operation level.

Regards,

René Lambelet - (3543 - *A581 
Nestec SA - 55, Av. Nestlé - CH-1800 Vevey
Tel: ++41/21/924'35'43 / Fax: ++41/21/924'45'89
E-Mail: rene.lambelet AT nestle DOT com



>-----Original Message-----
>From:  WINSOR.A@BCBSRI [SMTP:winsor.a AT BCBSRI.E-MAIL DOT COM]
>Sent:  Saturday, May 09, 1998 9:06 PM
>To:    ADSM-L AT VM.MARIST DOT EDU
>Subject:       Token Ring Question
>
>Server: ADSM V2.1 on OS/390
>Comm: TCP/IP 3.2; 16Mb T/R
>Client: Primarily Netware 4.10
>
>A Netware client can backup its own local drives and can also backup a remote
>drive on another Netware machine; i.e.  NetwareA can attach NetwareB's drives
>as if they belonged to NetwareA.  We use this method when we don't want to
>"pollute a Netware server with another destabilizing process" (not a decision
>that I always agree with).  I've run benchmarks using a local drive and a
>remote drive with a "LOAD DSMC SEL ..." of the same 15MB file on each drive.
>As expected, the local drive backups are quicker: an average elapsed time of
>34.9 seconds compared to 44.3 seconds on a remote drive.  But the network
>throughput for a remote drive is much better than for the local drive: 1100
>KB/S for the remote and only 640 KB/S for the local.
>
>My guess is that the with the local drive, the token is released after every
>TCP/IP block is sent upstream and ADSM must wait for it to make the rounds
>before it can send another block.  But with remote drive access, the ADSM
>station already has the token at the completion of the remote file I/O and
>takes advantage of the possession by sending the TCP/IP block up immediately
>(if one is ready).  Can anybody confirm/deny this or provide any other
>explanation for the difference in throughput?
>
>Thanks.
>Alan Winsor
>Blue Cross Blue Shield of RI
>winsor.a AT bcbsri.e-mail DOT com
<Prev in Thread] Current Thread [Next in Thread>
  • Token Ring Question, WINSOR.A@BCBSRI
    • Re: Token Ring Question, Lambelet,Rene,VEVEY,FC-SIL/INF. <=