Originally posted by partcyborg
View Post
No hardware, no kernel version, no details on the rules or the cmdline used to generate packets, and the author states at the top "This is a basic benchmark that is not close to the real world."
Beyond that, the entire result seems to be due to drivers, as per the author: That seems to be a mellanox driver queue balancing because it only uses 12 cores at 100% and the other cores are idle
In other words, the entire test seems to be due to driver / kernel differences, rather than anything to do with the merits of the packet filter. Treating this anything more than an interesting anecdote is incredibly disingenuous.
Leave a comment: