Customers results…
by Stephen Wilde
Recently I was asked by a customer for an R16 evaluation unit. After dispatching the unit it appeared the R16 would be in good company, namely; Citrix, F5, Jetnexus and CIA. The customer was evaluating a unit from all the above manufactures. Win or lose I felt this would be an interesting independent test of the R16 against some well know names………
F5 BigIP 3400 - {SSL TPS 4,500+} [Price £24,000+] {Very configurable, V.Fast, + API, SSL licence model} = Result 5/5
Citrix 12000 – {SSL TPS 2,000+} [Price £16,000+] { Very configurable, Fast, + API, SSL licence model} = Result 4/5
Jetnexus – {SSL TPS 300+} [Price £8,000+] {Slow software SSL & API} = Result 3/5
CAI Webmux – {SSL TPS ???} [Price £3,000+] {Unable to test SSL throughput. Poor Linux OS, not good for multiple sites due to network limitations} = Result 0.5/5
Loadbalancer.org Enterprise R16 – {SSL TPS 850+} [Price £1,495] {Configurable, fast, had to create loop backs (as CIA)} = Result 3.5/5
Addition test comment; based on; Performance V Price, the Loadbalancer.org R16 stands out from the crowd.
Based on the above test results you can see the R16 delivers and at an amazing price point. Whilst the R16 will not suite all customers it will certainly offer true hardware load balancing at a price point that is very affordable.
September 22nd, 2008 at 12:17 pm
Hi, I know what customer prospect performed this testing and as such I know the testing method used. Testing Enterprise network appliances can be difficult. Many end user struggle to create realistic environments. In this case the end user used 1 web server and 1 injector (FYI we have 30+ servers in our stress test environment) In my opinion that test method was not valid. I feel that it would be more appropriate to publish some information on the testing method as anyone can make up a number. I am also interest in seeing how they tested 4500 SSL TPS on 1 server!
September 22nd, 2008 at 3:26 pm
I guess one of the points is that test results are completely arbitrary :-). Customers always have their own set of requirements which may or may not correlate with what the vendor is trying to sell.
You may also be interested to know that the customer was given a £20,000 discount on the units they finally chose…
As for one test back end, I don’t see the problem with that as the SSL is terminated on the load balancer and therefore a single web server could easily handle 4500 requests in HTTP mode. If you wanted to do a large HTTP test then you’d probably want an avalanche/reflector set up.
Mind you how many customers do you know who are really doing more than 300 TPS :-).
I also quite like the CAI Webmux and they got slated… Alas the customer is always right, even if they are wrong.