I do not think network performance may be beyond the scope of this forum.
The network is such a critical part of TSM performance that network tuning
for TSM operations should be a greater part of the discussions here.  I,
for one, am very glad to share the hard won knowledge of other network
tuners.

Orville L. Lantto
Datatrend Technologies, Inc.  (http://www.datatrend.com)
IBM Premier Business Partner
121 Cheshire Lane, Suite 700
Minnetonka, MN 55305
Email: Orville.Lantto AT datatrend DOT com


CONFIDENTIALITY NOTICE: This e-mail message, including any attachments, is
for the sole use of the intended recipient(s) and may contain confidential
and privileged information. Any  unauthorized review, use, disclosure or
distribution is prohibited. If you are not the intended recipient, please
contact the sender by reply e-mail and destroy all copies of the original
message.



Richard Sims <rbs AT BU DOT EDU>
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
03/23/2005 11:28 AM
Please respond to
"ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>


To
ADSM-L AT VM.MARIST DOT EDU
cc

Subject
Re: TCP settings for Jumbo frames






Network performance issues are rather beyond the scope of this forum:
you'd be better off conferring with your networking people, who can
evaluate what you're attempting and run traces to see what's actually
happening.

When approaching network performance test like this, the big affector
is network topology. In this case, I would expect that the two systems
being tested were in different subnets, going across one or more
routers, where the jumbo frames may have had to be disassembled into
smaller pieces, which incurs overhead. Again, this is something your
site network people have to look at.

    Richard Sims

On Mar 23, 2005, at 12:04 PM, Miller, Ryan wrote:

> Here's a question from one of my cohorts, since this is our first try
> at this, I thought I would see if I could get any input from anyone
> else doing this.
>
> We have been testing jumbo frames(8990) using TSM on an AIX client.
> The base line test with MTU of 1500 moved 22G in 44 minutes. Changing
> the MTU to 8990 and the TCP receive and send buffer sizes on both the
> mainframe and AIX client to 256K, the same 22G of data takes 1 hour
> and 40 minutes.
>
> Also the largest input packet size is 8192, not 8940. Are their any
> tcp parameters to se on the client that would help with the jumbo
> frame transfer time?