I've noticed a discrepancy between the way the F5 cluster and Exchange measures CAS connections. When checking the number of current connections in the F5 virtual server stats for Exchange OWA (HTTP/HTTPS), I was very surprised to see that the number of connections was far higher (as much as 5 times higher!) than the number being reported by Exchange. I was wondering if this is due to a difference in how "connections" are counted. If so, how are connections counted by the F5 virtual server stats?

Hi Stuart, are you using OneConnect on the VIP? If so, BIG-IP will reuse existing server-side connections. You won't see a 1-to-1 ratio of client/server connections. You can see OneConnect profile reuse stats in the GUI or by running "show ltm profile one-connect " This is on top of the total number of connections to the VIP being divided (hopefully) evenly among your pool members. Mike

Hi, If you have a proxy in the middle of the connection, you maybe exists duplicated in the connections. A. Diezma

As far as I can tell, I'm not using OneConnect. Good to know though. Thanks!

Are you sure you're not comparing OWA user sessions to TCP connections? There will typically be up to 8 times as many TCP connections as there are user sessions (due to the 8-connections-per-hostname limit in most browsers). 5 sounds like an expected multiple. Otherwise I'd say wherever you're monitoring connections in Exchange must be counting things differently - perhaps not counting Keep-alive (but idle) TCP connections.

Discrepancy: Current Virtual Server Connections

11 Replies

mikeshimkus_111
Historic F5 Account
Aug 21, 2013
Hi Stuart, are you using OneConnect on the VIP? If so, BIG-IP will reuse existing server-side connections. You won't see a 1-to-1 ratio of client/server connections. You can see OneConnect profile reuse stats in the GUI or by running "show ltm profile one-connect "

This is on top of the total number of connections to the VIP being divided (hopefully) evenly among your pool members.

Mike
- Stuart_Page_131
  Nimbostratus
  Aug 21, 2013
  As far as I can tell, I'm not using OneConnect. Good to know though. Thanks!
adiezma_1656
Nimbostratus
Aug 21, 2013
Hi,

If you have a proxy in the middle of the connection, you maybe exists duplicated in the connections.

A. Diezma
- Stuart_Page_131
  Nimbostratus
  Aug 21, 2013
  No proxy in use. Thanks.
eey0re
Cirrostratus
Aug 21, 2013
Are you sure you're not comparing OWA user sessions to TCP connections? There will typically be up to 8 times as many TCP connections as there are user sessions (due to the 8-connections-per-hostname limit in most browsers). 5 sounds like an expected multiple.

Otherwise I'd say wherever you're monitoring connections in Exchange must be counting things differently - perhaps not counting Keep-alive (but idle) TCP connections.
nitass
Employee
Aug 22, 2013
I was wondering if this is due to a difference in how "connections" are counted. If so, how are connections counted by the F5 virtual server stats?

i understand F5 connection is tcp connection which is in F5 connection table. is it same as in Exchange?
- Stuart_Page_131
  Nimbostratus
  Aug 22, 2013
  I'm performing session based queries on Exchange.
- nitass
  Employee
  Aug 22, 2013
  i do not think number of tcp connections is going to be equal to number of user sessions.
nitass_89166
Noctilucent
Aug 22, 2013
I was wondering if this is due to a difference in how "connections" are counted. If so, how are connections counted by the F5 virtual server stats?

i understand F5 connection is tcp connection which is in F5 connection table. is it same as in Exchange?
- Stuart_Page_131
  Nimbostratus
  Aug 22, 2013
  I'm performing session based queries on Exchange.
- nitass_89166
  Noctilucent
  Aug 22, 2013
  i do not think number of tcp connections is going to be equal to number of user sessions.