Highlighted
Respected Contributor.
Respected Contributor.
846 views

Connections per second vs hits per second?

Jump to solution

All,

I need a little help interpreting my test output.  We're making web service calls using the web protocol in PC.  Typically we make 1 call per script with just a single iteration.  We have "simulate a new user on each iteration" unchecked.

Our app has run into a new scalability problem that we're trying to identify.  The symptom is that at high vuser counts, CPU goes really high.  One curious correlation is that vusers, connections per second, and CPU all increase in the same way.  Previously we've been able to handle this number of hits per second, so I'm wondering if a change in the test itself, specifically vuser count, could be the culprit.

At 1x load we're using 130 vusers and get about 55 hits per second with connections per second below 5.

As we ramp up to 2x (260 vusers,) the hits per second only increases to about 70 but connections per second jumps to about 45.

In all cases, connections per second and connection shutdowns are equal.

vusers.png

The app server is Tomcat 7.x and it's configured to allow 80 connections.  There are 4 Tomcat instances in the test, so in theory they should be able to accept 320 connections.

Any thoughts?  I'm suspicious about the relationship between vusers and connections per second.

 

thanks

 

0 Likes
1 Solution

Accepted Solutions
Highlighted
Respected Contributor.
Respected Contributor.

Re: Connections per second vs hits per second?

Jump to solution

This was a result of having too many vusers with slow pacing.  The http keep alive timeout was 60 seconds on the PC side and 10 seconds on the Tomcat side.  We had some scripts with pacings > 10 seconds, so every request required a new connection.

To make matters worse, PC seems to use 2 connections for mutual authentication.  It connects, starts the handshake, then disconnects with the message "Re-negotiating https connection [0] to foo.mycompany.com."  It then reconnects, completes a full handshake, and submits the request.  So, with slow pacing, there are really 2 connections per request.

 

thanks

View solution in original post

0 Likes
2 Replies
Highlighted
Respected Contributor.
Respected Contributor.

Re: Connections per second vs hits per second?

Jump to solution

This was a result of having too many vusers with slow pacing.  The http keep alive timeout was 60 seconds on the PC side and 10 seconds on the Tomcat side.  We had some scripts with pacings > 10 seconds, so every request required a new connection.

To make matters worse, PC seems to use 2 connections for mutual authentication.  It connects, starts the handshake, then disconnects with the message "Re-negotiating https connection [0] to foo.mycompany.com."  It then reconnects, completes a full handshake, and submits the request.  So, with slow pacing, there are really 2 connections per request.

 

thanks

View solution in original post

0 Likes
Highlighted
Acclaimed Contributor.
Acclaimed Contributor.

Re: Connections per second vs hits per second?

Jump to solution

Thanks for sharing the solution, even if it was solved by yourself!


br /ola
---------------------------------------------------------------------
Please mark post as solved if your problems or questions is/are resolved.
If this post was valuable to you, please consider kudo it.

.
0 Likes
The opinions expressed above are the personal opinions of the authors, not of Micro Focus. By using this site, you accept the Terms of Use and Rules of Participation. Certain versions of content ("Material") accessible here may contain branding from Hewlett-Packard Company (now HP Inc.) and Hewlett Packard Enterprise Company. As of September 1, 2017, the Material is now offered by Micro Focus, a separately owned and operated company. Any reference to the HP and Hewlett Packard Enterprise/HPE marks is historical in nature, and the HP and Hewlett Packard Enterprise/HPE marks are the property of their respective owners.