Date: Sat, 17 May 2014 07:52:28 -0700 From: Adrian Chadd <adrian@freebsd.org> To: "Bentkofsky, Michael" <MBentkofsky@verisign.com> Cc: "freebsd-net@freebsd.org" <freebsd-net@freebsd.org> Subject: Re: [rfc] tcp timer update for RSS Message-ID: <CAJ-Vmo=fqW72o44p7PEpoJZQvn9mHZimypFvT15zmsXhFUW1fQ@mail.gmail.com> In-Reply-To: <CAJ-Vmok-QFMHURShKbZBatw1md55Uc_5rMte3rCVZz9B2Z-X_g@mail.gmail.com> References: <080FBD5B7A09F845842100A6DE79623346F2EB44@BRN1WNEXMBX01.vcorp.ad.vrsn.com> <CAJ-Vmok-QFMHURShKbZBatw1md55Uc_5rMte3rCVZz9B2Z-X_g@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
.. and as a note - it'll all be behind #ifdef RSS. -a On 17 May 2014 07:49, Adrian Chadd <adrian@freebsd.org> wrote: > On 17 May 2014 07:44, Bentkofsky, Michael <MBentkofsky@verisign.com> wrot= e: >> Hi Adrian, >> >> >> >> I haven=E2=80=99t had the chance to look this over carefully yet as we= =E2=80=99re at BSDCan. >> I think I understand what you=E2=80=99re trying to achieve by aligning t= he per-CPU >> timer processing per core. In principal that sounds reasonable, although= I >> am unsure if you were trying to solve a particular performance issue wit= h >> this particular change. My sense is this is all preparatory with the goa= l of >> all inp processing to become per core. Could you comment on the general >> evolution you=E2=80=99re considering? Do most of the PCB structures beco= me per-core, >> as in PCB groups? >> >> >> >> If you=E2=80=99d like us to test this change, I=E2=80=99m happy to do so= . At the moment I >> don=E2=80=99t know if we=E2=80=99d expect to see any benefit =E2=80=93 d= o you have any traffic >> conditions for which this showed any difference? But we can certainly dr= ive >> many hundreds of thousands of connections at reasonably high connection >> rates if that will help. > > Hi! > > Yes, the aim is to provide RSS support in the RSS model of "align > everything to a specific core." The goal for RSS is to remove both > lock contention between cores and keep data CPU/cache local to improve > efficiency there. > > There's nothing obvious that'll be beneficial right now. The items at > https://wiki.freebsd.org/NetworkRSS have to occur before it is > beneficial to anyone. > > > > -a
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmo=fqW72o44p7PEpoJZQvn9mHZimypFvT15zmsXhFUW1fQ>