Clarification on results returned by polling 'slbVServerNumberOfConnections' snmp oid

Unanswered Question
Feb 26th, 2010


Our testing team are currently load-testing a pair of Apache webservers sitting behind an ACE module (running A2.1.4) which is performing round-robin load balancing to the two boxes. Connection timeout values are default and server-side connection re-use is not enabled

Using our network monitoring tool (SolarWinds Orion) I've configured a custom poller to poll the 'slbVServerNumberOfConnections' OID associated with the virtual server / VIP serverfarm.

The test they are running is to introduce 10 virtual users every 10 seconds until 1200 users is reached with each user making a page request every 120 seconds in the 1st test, 150 in the second and 180 in the third.

I've attached the graph of the results of the 3 tests below.

The testing team expected each run to use progressivly less connections (as the time between page requests was increased) but each run appears almost identical from an ACE perspective.

Also, even though the maximum number of virtual users was 1200 the maximum number of connections tops out at just over 2000?

I'm being asked to explain the results but I not too sure how to interpret them so any advice or insight would be appreciated.

thanks in advance


I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Gilles Dufour Fri, 03/05/2010 - 00:38

First, verify your result with show command like 'show service-policy' and 'show conn', 'show resource usage'.

Then, you may also need to capture sniffer traces to understand the traffic behavior.

It's sometimes interesting to see those traffic generator in action's not always what we expect.

The sniffer tool will also perform some analysis and you can compare the sniffer results with ACE and the load generator.



This Discussion

Related Content