* [Cerowrt-devel] peeling harder with cake @ 2015-07-03 1:27 Dave Taht 2015-07-03 3:15 ` [Cerowrt-devel] [Cake] " Jonathan Morton 0 siblings, 1 reply; 3+ messages in thread From: Dave Taht @ 2015-07-03 1:27 UTC (permalink / raw) To: cake, cerowrt-devel I compared one of mikael's datasets with the post get-rid-of-atomic AND peel harder version of cake... Look like peeling harder is better. Also got more throughput for some reason. http://snapon.lab.bufferbloat.net/~d/saving300usecpeelingharderwithcake.png -- Dave Täht worldwide bufferbloat report: http://www.dslreports.com/speedtest/results/bufferbloat And: What will it take to vastly improve wifi for everyone? https://plus.google.com/u/0/explore/makewififast ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Cerowrt-devel] [Cake] peeling harder with cake 2015-07-03 1:27 [Cerowrt-devel] peeling harder with cake Dave Taht @ 2015-07-03 3:15 ` Jonathan Morton 2015-07-03 16:01 ` Dave Taht 0 siblings, 1 reply; 3+ messages in thread From: Jonathan Morton @ 2015-07-03 3:15 UTC (permalink / raw) To: Dave Taht; +Cc: cake, cerowrt-devel > On 3 Jul, 2015, at 04:27, Dave Taht <dave.taht@gmail.com> wrote: > > Also got more throughput for some reason. Is the NIC doing software GSO or does it have hardware support? If the former, it would suggest that software GSO is a universally bad idea and should be excised. If the latter, GSO should be disabled full stop for this hardware, so we can stop fannying about with peeling. - Jonathan Morton ^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Cerowrt-devel] [Cake] peeling harder with cake 2015-07-03 3:15 ` [Cerowrt-devel] [Cake] " Jonathan Morton @ 2015-07-03 16:01 ` Dave Taht 0 siblings, 0 replies; 3+ messages in thread From: Dave Taht @ 2015-07-03 16:01 UTC (permalink / raw) To: Jonathan Morton; +Cc: cake, cerowrt-devel On Thu, Jul 2, 2015 at 8:15 PM, Jonathan Morton <chromatix99@gmail.com> wrote: > >> On 3 Jul, 2015, at 04:27, Dave Taht <dave.taht@gmail.com> wrote: >> >> Also got more throughput for some reason. > > Is the NIC doing software GSO or does it have hardware support? If the former, it would suggest that software GSO is a universally bad idea and should be excised. If the latter, GSO should be disabled full stop for this hardware, so we can stop fannying about with peeling. :) It is software GSO, and yes, given the 4.0 results for forwarding on this platform, I would consider seriously cutting it down in size. (say, 2 MTU, max, down from 64k, or less - should give better cache behavior, too), That said, TSO/GSO/GRO is everywhere, and peeling, needed. Take the intel ethernet chips, for starters. But it would be best to experiment and benchmark, further. > > - Jonathan Morton > -- Dave Täht worldwide bufferbloat report: http://www.dslreports.com/speedtest/results/bufferbloat And: What will it take to vastly improve wifi for everyone? https://plus.google.com/u/0/explore/makewififast ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2015-07-03 16:01 UTC | newest] Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2015-07-03 1:27 [Cerowrt-devel] peeling harder with cake Dave Taht 2015-07-03 3:15 ` [Cerowrt-devel] [Cake] " Jonathan Morton 2015-07-03 16:01 ` Dave Taht
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox