sna slow response time

phil.wydra · ‎04-29-2003

I have a CIP connected to a mainframe. I have CSNA and CLAW defined. I have an XCA defined for 20 AS400's to connect. Some AS400's are on interfaces on the same router as the CIP and some AS400's are remotely located and utilize DLSW. All AS400's connect fine.

Now the problem:

There are times during the day when SNA traffic moves slowly from an AS400 to the mainframe. Not all AS400's slow down at the same time. If I issue the SNA aping command

'd net,aping,id=AS400name' during periods when the transmissions are slow, I get response times of over 2000 ms. If at the same time, I issue a normal TCP/IP ping command between the IP stacks on both boxes, I get response times of 20 ms. The exact same path is taken by both devices.

There does not seem to be a utilization issue. When the slowdown occurs, there usually are several SNA connections from the AS400 to the mainframe. AS400 SNA APINGs from a box that is slow to the mainframe are fast to other boxes.

My question is:

Is there a default value that limits the amount of data that can be pushed through an LLC session? If so can the value be changed. Are the any suggestions for getting higher traffic flow?

dixho · ‎04-29-2003

There are a number of parameter related to the windows size. (i.e. the amount of data can be pushed through DLSw). Cisco routers use LLC2 window size 7. This cannot be changed. Cisco routers slow down LLC2 traffic if there is a congestion on TCP or DLSw routers receive DLSw flow control indicators.

1. tcp-queue-max in "dlsw remote-peer tcp" statement controls the output TCP queue size. This control

2. init-pacing-window and max-pacing window in "dlsw local-peer" controls the DLSw circuit window size.

Before changing any parameters, I suggest you to find out the bottle neck of the network. I understand that WAN is not the bottle neck. I suggest you to do the followings:

1. do a show tcp and find out if there is any retransmitted segment in the TCP queue. Please go to the following URL:

http://www-tac.cisco.com/Support_Library/field_alerts/fn20673.html

There are a number of bugs caused by TCP and affected DLSw performance.

2. do a show llc2. Find out the state of the LLC2 circuit. Make sure that the LLC2 circuit is in a normal state.

3. do a show dlsw circuit detail. Find out if there is any half window or reset window indicator sent or received.

4. do a show cls. Find out that state of the circuit.

You may need to identify the bottle neck by taking sniffer traces on the AS/400 side or take a CIP console trace.

ihutchinson · ‎04-12-2005

Hi Dixho, in your response above you refer to URL http://www-tac.cisco.com/Support_Library/field_alerts/fn20673.html

I have tried to view this URL but I only get a blank page. I have also searched CCO for Field Notice 20673 and come up with nothing. Can you assist us in locating this URL as we feel it may be of use to us.

Thanks, Ian.