Periodic keep-alive messages exchanged between clients and servers are common in IP networks and let servers determine whether clients are still active or reachable. The management console generally ignores TCP keep-alives if they adhere to the standards in RFC 1122 and excludes any associated statistics from byte counts and observation totals.
However, some applications are designed to use custom keep-alive mechanisms. If the response from the client is an acknowledgment that contains a payload, the management console treats the client response as a request for data, and starts the Server Response Time (SRT) timer. This results in an inaccurate SRT measurement and SRT observation count once the keep-alive timer expires, when the server typically sends out another packet.
Popular applications that use keep-alives include Citrix and Microsoft Exchange. If you suspect that another application is sending keep-alives, look for the inverse relationship between observations and SRT and for SRT averages in the second range instead of the millisecond range.
A CA Standard Monitor or CA Multi-Port Monitor can be configured to filter application keep-alive messages by Server Response Time to avoid skewing server metrics.
The management console uses NRTT observations to filter applications that use application keep-alives. If necessary, you can adjust the threshold for the minimum number of NRTT observations during a 5-minute interval.
|
Copyright © 2014 CA Technologies.
All rights reserved.
|
|