Eliminate TCP sessions with long open durations

TYLER WEST · ‎05-04-2016

Is there a way to tell WAAS to no longer optimize sessions that have an excessively long open duration? We are running 5.5.5 and often have sessions for either email or CIFS that may have been open for days on end. Is there a way to put a time limit on those sessions? Our client is bumping up against the connection limit occasionally and eliminating those long-lived sessions would reduce the "noise floor" if you will and reduce the number of overall connections. I realize there is also a user training aspect to this issue but we also realize that not everyone is going to shut down their email or open windows to file shares when they are done for the day. Thanks!

Aleksey Pan · ‎05-04-2016

Hi Tyler,

This doc probably would be helpful to customize the timeout values:

http://www.cisco.com/c/en/us/td/docs/app_ntwk_services/waas/waas/v551/reference/cmdr/glob_cfg.html

(config) tcp

To configure TCP parameters on a WAAS device, use the tcp global configuration command. To disable TCP parameters, use the no form of this command.

tcp {cwnd-base segments | ecn enable | increase-xmit-timer-value value |
init-ss-threshold value | keepalive-probe-cnt count | keepalive-probe-interval seconds |
keepalive-timeout seconds}

no tcp {cwnd-base segments | ecn enable | increase-xmit-timer-value value |
init-ss-threshold value | keepalive-probe-cnt count | keepalive-probe-interval seconds |
keepalive-timeout seconds}

Syntax Description

cwnd-base segments	Sets initial send congestion window in segments (1–10).
ecn enable	Enables TCP explicit congestion notification.
increase-xmit-timer-value value	Specifies the factor (1-3) used to modify the length of the retransmit timer by 1 to 3 times the base value determined by the TCP algorithm. Note Use this keyword with caution. The keyword can improve throughput when TCP is used over slow reliable connections but should never be changed in an unreliable packet delivery environment.
init-ss-threshold value	Sets initial slow-start threshold value (2-10).
keepalive-probe-cnt count	Specifies the length of time that the WAAS device keeps an idle connection open. The number of probe counts is 1–10.
keepalive-probe-interval seconds	Specifies the number of times that the WAAS device retries a connection. The keepalive probe interval is in seconds (1–300).
keepalive-timeout seconds	Specifies the length of time that the WAAS device keeps a connection open before disconnecting. The keepalive timeout is in seconds (1–3600).

Defaults

tcp cwnd-base : 2

tcp increase-xmit-timer-value: 1

tcp init-ss-threshold: 2 segments

tcp keepalive-probe-cnt : 4

tcp keepalive-probe-interval : 75 seconds

tcp keepalive-timeout : 90 seconds

Hope this helps.

Regards,

Alex

nathy1984 · ‎05-25-2016

Hello

Did you find a solution for this problem?

If yes, please provide the solution for me.

Regards,

TYLER WEST · ‎05-25-2016

Unfortunately, no. But to some extent it makes sense why you might not want WAAS to have that capability. It's a bit too intrusive. Since posting this we have determined that most of these really come down to poor user behavior which is manageable with training. Outside of that you just have to find ways of managing the total connections to prevent exceeding the max threshold. One of the things we did in this customer is flip the WCCP ACL on it's head. Instead of using the ACL to deny certain protocols and allow all other TCP traffic we spent time identifying what really needed to be optimized or what truly got the benefit from it and only allowed those TCP ports and denied all others. That has brought the total connections under better control.