Hi,
we had also performance problems, especially with NT servers. I found out,
that there were two bottlenecks:
NetWork: TIP/IP and IPX was installed. After we had disabled IPX, the
network performance has increased (factor 2).
Better client hardware = better performance. It is very timeconsuming, to
scan large filesystems for files and compare them to the ADSM database.
Regards,
Jörg
Michael Bartl <michael.bartl AT ZENTRALE.ADAC DOT DE> on 21.09.99 09:25:33
Please respond to "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
cc: (bcc: Jörg Nouvertné)
Subject: AW: ADSM on MVS Network Throughput
Kyle,
we're in the process of investigation ADSM performance, too.
We do nightly incrementals of about 50 servers (all NT).
The servers are connected to an ATM backbone using dedicated 16mbit TR
segments (madge TR switch), the ADSM server (3.1.2.20 on OS/390) has access
to the backbone using two 2216 machines (each using one ESCON link).
All the NT clients give us an aggregate data transfer rate between 100 and
1100 kb/s. The lower rates represent servers that have large filesystems
and
only have to send few files to ADSM, the highest rates come from servers
that have a higher daily change rate or backup only a few very large files.
The peak data rates measured at the 2216 are between 30 and 40 mbit/s.
For a few months now we're trying to find out where out bottlenecks are.
The
NT guys think their machines to be much faster than they behave, the
network
guys are absolutely sure that the network is not limiting anything (and
they
don't move a finger until we use 100% bandwidth over the whole night), and
last not least our ADSM server and the S390 are of course much faster than
light can be... Very dangerous territory!
I'll keep you informed when we get some results.
Best regards,
Michael
--
Michael H. Bartl
Michael H. Bartl
ADAC e.V. (Rechenzentrum, Organisation und Planung)
Tel.: (089) 7676-4063, Fax: (089) 7676-8161
mailto:michael.bartl AT zentrale.adac DOT de
http://www.adac.de
|
