Some profiling seems not well generated, here is the raw pictures: https://leaf.dragonflybsd.org/~sephe/perfcomp/1K.png https://leaf.dragonflybsd.org/~sephe/perfcomp/8K.png https://leaf.dragonflybsd.org/~sephe/perfcomp/16K.png https://leaf.dragonflybsd.org/~sephe/perfcomp/ipfwd-bi.png
On Fri, Mar 3, 2017 at 3:44 PM, Sepherosa Ziehau <[email protected]> wrote: > Hi all, > > Since so many folks are interested in the performance comparison, I > just did one network related comparison here: > https://leaf.dragonflybsd.org/~sephe/perf_cmp.pdf > > The intention is _not_ to troll, but to identify gaps, and what we can > do to keep improving DragonFlyBSD. > > According to the comparison, we _do_ find one area DragonFlyBSD's > network stack can be improved: > Utilize all available CPUs for network protocol processing. > > Currently we only use power-of-2 CPUs to handle network protocol > processing, e.g. on 24 CPUs system, only 16 CPUs will be used to > handle network protocol processing. It is fine for workload involving > userland applications, e.g. the HTTP server workload. But it seems > forwarding can enjoy all available CPUs. I will work on this. > > Thanks, > sephe > > -- > Tomorrow Will Never Die -- Tomorrow Will Never Die
