Description
Speed tests rely on algorithms to calculate the throughput of a connection. The tests are simulations of the throughput the connection should be capable of.
However, there are several problems in relying upon speed test results.
1) The results of speed tests vary widely not only from one site to the next but even from one attempt to the next on the same site.
For example, Fortinet has performed tests in a lab environment using SpeedTest software available at
http://www.speedtest.net. The results for a same test could vary by up to 40%. It appears that the flash application is very sensitive to TCP recovery and windowing mechanisms.
2) It is widely recognized that many of the algorithms used for speed tests do not take into account proxy servers, in part because the tests make incorrect assumptions about the relationship between latency and throughput.
When using UTM features on a FortiGate, all files must be buffered by application proxies in the FortiGate. It should be noted that packets can experience higher latency between server and client but the overall throughput can still be high. The latency introduced by the buffering does not necessarily affect overall throughput but will cause the traffic to be more bursty.
Solution
For the reasons stated above, Fortinet does not advise the use of speed tests for the testing of UTM firewalls. If there are concerns about the actual throughput and performance of a FortiGate, please contact Fortinet Technical Support.
