Telnet session swamping the network?

[thewho?]

Hello,
HP-UX 11,N4000, 100Mbit card, Reflections
I don't know if this is a problem or normal operation.
We have situations where a programmer scans a large database and, to make sure his/her program is moving along, he/she sends some information to his/her telnet screen. This causes a flood of IP packets and everybody in this LAN that has a telnet session with the server notices it.
The N4000 is connected directly to a Cisco Catalyst 2900 switch and the users in our LAN go to several HP Procurves 4000M and from there to another Cisco.
According to glance, our normal traffic is around 800 ps, I sometimes see peaks of 3,200 ps but the response time is still good in our LAN. Only when the peak is caused by somebody in our LAN we see degrading response time.
By 'response time' I mean the time delay when you type in your telnet session.
I ran another large query in another N server with the same results: the response time increased for the telnet sessions connected to that server without affecting the sessions in the original server.
Is this normal? Is there something I can do to prevent this (other than ask the programmers to reduce the data sent to their screen) ?
Thanks in advance,
Luis

[John Bolene]

You mention some and a flood in the same statement.

A telnet session should not be capable of that much traffic as to slow down lan traffic for others.

What might be happening is that the massive query itself is slowing down the machine and causing response to be slower.

Is the data sent to his screen constantly streaming by or just an ocassional line?

Have you looked at glance during the time in question?

[Roger Baptiste]

Luis,

What sort of 'information' is being sent by the user sessions? I don't think, it
would be that much to cause
a network congestion, unless
they are spewing out the
database operations result
at frequent result, which
could mean spewing thousands
of records in one burst.
If all they went is to
make sure there connection/program is alive and running, they can echo
a simple status command when
the operation starts and
let it remain that way until
the program completes. In the
background they can have
the output being written
to a log file which can be
"tailed'' by another telnet
session.

I hope i understood your
question correctly and this
is only a network congestion
issue and not CPU or other
resource bottleneck.

-raj

[A. Clay Stephenson]

Hi Luis:

I don't think the messages have anything to do with this. It's simply the query itself. You said that you ran another large query but there are queries and there are queries. You could absolutely nail this down by getting the database out of this. Write a very tight script
or better a small C program or even simply cat a large text file to this guy's terminal and see if you see the same slowdown. Things that
really slow a network down are SqlNet (or Net8) traffic between a server and a slug of clients.

What you may have to do is ask the DBA to run this stuff during off-peak times or (God forbid) tighten up his code to run more efficiently.

Regards, Clay

[thewho?]

Thanks for your prompt responses.
The database is not using a lot of resources (except for the network when running this query to the screen).
On my sand box I ran the same query twice, one with the outpout sent to file and another one to the screen.
The query sent to a file didn't use a lot of resources (obviuosly the network resource wasn't used).
The query sent to the screen did create a lot of IP packets and caused a slower response time ( I logged in from a different PC to verify this ).

As Clay suggested, I 'cat' a large text file and this didn't consume a lot of network resources, which leads me to believe that the 'display' command for the database creates a lot of network overhead.

By the way, Clay, I don't think the DBA is at fault this time, I've worked with him for the longest time and he's one of the smartest ones I've ever met ; )

Thanks,
Luis
Database & Unix Administrator.