Slow down on 11.0?

[Adi Kremer]

Hi,

I have just finished cold install 11.0 upgrading from 10.20 on a H40 box. The upgrade process went smoothly but there seems to be one issue which started immediately on Monday morning after the upgrade.

The telnet connection through an ISDN line has slowed down significantly. I have a menu based program with dos like graphics which now "takes its time" when responds to user key strokes as well as repainting the menu screens. The load on the system is very low (up to 6 users) and the system is highly idle. I can't duplicate the problem when I telnet from our LAN.

Any ideas?

[Denver Osborn]

any patches loaded?

[Antoanetta Naghiu]

I'm not familiar with H40.Check to play with kernel params. What are the specs of your box? (memory, cpu, disks)?
how much swap? Come back with details.

[Cheryl Griffin]

Adi,

You will want to make sure that you are at the proper patch level for the Mux/pty, streams, lan, and arpa transport filesets. You can do this by searching the ITRC patch catalog to see what is HP recommended.

(PS. This section of the ITRC will change within the next few days so that it is easier to find which patches are the recommended ones.)

You may also want to monitor the system to see if there is some other problems going on with the system as far as performance is considered. You might need to do some tuning for 11.0, like the kernel parameters that affect paging.

Lastly, you might want to run a trace using nettl and duplicate the slow telnet problem to see if you can capture the problem in the trace file. Then have that file analyzed by HP Support.

Best Wishes,
Cheryl

[Adi Kremer]

Thanks for the responses.

I have 512MG on this system and I am patched to the latest PTY/Streams, Arpa, lan patches. Now, I have another machine (k370)running 10.20 and which has the same application and is running fine through the ISDN line. Kernel paramaters (h40 and k370)are very similar.

Again, I can't duplicate it over the LAN so it seems what ever "data storm" is going on, my 10mb/s LAN can still handle it v.s a 128kb ISDN. Any ideas on specific kernel parameters? I guess my other option is to start network tracing. Anything in particular to look for?

Adi

[Cheryl Griffin]

Adi,
I know enough networking to be considered dangerous. When it comes to nettl tracing and decoding, the networking experts will have to help you. I just know that nettl is the tool to debug this.

Cheryl

[Anthony deRito]

Adi,

The kernel tunable parameter you might want to look at is the "netmemmax" parameter. Compare these on your two systems. The value of this parameter is the maximum memory size that will be used during IP fragmentation re-assembly.

When you are transmitting IP packets, they are (or can be) broken down into "fragments" based on different MTU sizes over its travel. Memory is reserved for these "fragments" to be re-assembled into the fully assembled packet.

It is very possible, if this parameter is not tuned correctly, for memory to be used in such a way that fragmentation re-assembly is causing memory pressure on your system this causing a significant response time issue.

Hope this helps,

Tony

[Adi Kremer]

Tony,

Unfortunately, this parameter doesn't exist on 11.00.