Retransmission timeout (RTO) causes network latency on SSL connections through NetScaler.

The following are the symptoms of this issue:

• Load Balancing issues with Microsoft Exchange.
• Delay of over 1 minute seen when connecting to HTTPS resource through NetScaler.
• Bypassing the NetScaler improves speed by 10x.
• In NetScaler traces it is observed that there are many TCP Zero Window events from the SNIP to the backend resources.
• When this occurs there is no Zero window or small window seen on the client.
• Setting TCP flavor to BIC and increasing TCP send/receive buffer alleviates the issue somewhat.

To resolve this issue upgrade to NetScaler 10.5 61.x+ and 11.0 64.34+.

Problem Cause

This issue known issue which is tracked with issue ID #565938.

RTO occurs when the sender misses too many acknowledgments. A calculation is then made to wait before resuming transmissions. After some amount of time, usually at least one second, the sender slowly starts transmitting again. RTO causes delays that can be many seconds to several minutes.

• http://docs.citrix.com/en-us/netscaler/11/release-notes/main-releases/fixed-issues.html
• http://docs.citrix.com/en-us/netscaler/10-5/ns-rn-main-wrapper-10-5-con/main-releases/known-issues.html
• CTX219900 - Configure Time Delay for GSLB Services to be Marked as DOWN When MEP Connection goes DOWN on NetScaler 11.1.