3

I am working on to benchmark two different load balancers (pound and haproxy), but are unsure how this can be done. Of couse httperf can be used, but it may be technically difficult to actually make a load balancer crash due to a very high rate. How can I consider other variables that can be observed which may make indications of which one would be the more powerful under more stressful conditions

user3185936
  • 151
  • 4

1 Answers1

1

You really need to set up a benchmark that reflects your actual use case. Are you terminating TLS (that's very CPU intensive)?

Googling 'haproxy benchmark' and 'pound benchmark' yield lots of results using various methods to simulate clients.

Here's an example: http://www.haproxy.org/10g.html - Willy explains their test setup including the software used to simulate the clients.

Jim G.
  • 2,607
  • 1
  • 18
  • 19
  • Here's another helpful link where they actually used pound in front of haproxy (which is no longer necessary - haproxy does native TLS termination since v.1.5) - http://blog.loadbalancer.org/load-balancer-performance-benchmarking-haproxy-on-ec2-quick-and-dirty-style/ – Jim G. Apr 10 '15 at 17:51
  • Thanks for advice. Do you have any suggestions for what to use to make benchmark diagrams ? – user3185936 Apr 11 '15 at 09:46
  • @user3185936 , maybe this [online draw tool](https://www.draw.io/) can help you to make your diagrams... – Dalton Cézane Jun 21 '17 at 04:45