I realized I didn’t really understand how to interpret the test results that Mikrotik publishes for each router. It seems that people think the “25 IP filter rules” 512 byte packet result is the closest to real world performance. So, for the hEX refresh router, that shows 498.1 Mbps. Does that mean that you couldn’t use it for, say, a gigabit connection? It seems odd that a company would release a router that couldn’t do gigabit in this day and age.
What’s the best way to interpret these results, if not that? Thanks
The rule you’re mentioning is “rule of thumb”. But it comes with disclaimer: actual performance very much depends on actual confuguration and it’s possible that actual performance will be a lot higher than the rule of thumb estimation.
I believe that MT considers hEX refresh a true gigabit router (test result for full-size IP packets exceeds 1Gbps), so it’s not that they designed device for sub-Gbps speeds … it’s just that many MT device owners are “tinkerers” but many fail to produce most optimal configuration. And those configs are part of “real life” as well
25 IP filter rules 512 byte packet result is the closest to real world performance
i think published test are not meant to directly predict real world performance because the ammount of variables in real world is huge and unpredictable
but
i think the true utility of published tests are establish some foundation to compare the performance between different devices
in that context, 25 IP filter rules 512 byte packet result is useful to compare performance between devices in scenarios with configurations that use a lot of cpu processing
maybe if you have a consistent and repeatable scenario you can establish the ratio between this results and your actual performance, but is very specific of each implementation
For example In each network role a router can have a different configuration so it will have a different performance
That rule of thumb is quite accurate if you take the default configuration (with the firewall rules from MIkroTik) and disable the fasttrack rule (or before 7.18 when you use the default config with IPv6, because fasttrack was not available). If your configuration can actively make use of fasttrack, then the achievable throughput will be much higher in most cases, because for most of the packets of connections that could be fasttracked, performance would be nearer to the values in the “routing fast path” row.
For example, the old hEX RB750Gr3 using defconf but without fasttrack (or by using IPv6 with the default configuration) can achieve about 250-280Mbps (speedtest.net result) with RouterOS 7, but with fasttrack active has no problem reaching 930Mbps (speedtest.net result) or more, even with PPPoE overhead. With RouterOS 7.18, the hEX refresh (with nearly twice the routing performance) would have no issue with gigabit connections in the default configuration, for both IPv4 and IPv6. But if your configuration makes use of queues or mangle rules then fasttrack will no longer be usable and the throughput will probably drop to below 500Mbps.
