Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 16 Jan 2014 11:06:47 -0800
From:      Adrian Chadd <adrian@freebsd.org>
To:        Nikolay Denev <nike_d@cytexbg.com>
Cc:        "Alexander V. Chernikov" <melifaro@freebsd.org>, "freebsd-net@freebsd.org" <freebsd-net@freebsd.org>
Subject:   Re: ECMP hash keys?
Message-ID:  <CAJ-VmokKTSgY_3VHHdbG0BiteMp33BeyXvKPdDW6SMUD8SOksA@mail.gmail.com>
In-Reply-To: <CA%2BP_MZGZtvqiAQEayz3nh1kN=RYHfoU%2BfsqCCrjqGikH1kMZWw@mail.gmail.com>
References:  <52D5138B.8050100@fsn.hu> <CA%2BP_MZFQU4%2B05Pk5cZ4NMZujD9vXDrV=mehN7_vz1OZ6r2-f1Q@mail.gmail.com> <52D6525D.50102@FreeBSD.org> <CAJ-VmomP-JaVopS0aneeV82OFtM1Pvb=qKn__mn=ooDXOdgmQw@mail.gmail.com> <CA%2BP_MZGZtvqiAQEayz3nh1kN=RYHfoU%2BfsqCCrjqGikH1kMZWw@mail.gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On 16 January 2014 08:02, Nikolay Denev <nike_d@cytexbg.com> wrote:

> Hi Adrian,
>
> Probably a stupid question, but I'm trying to understand more about this,
> so basically the benefit of using essentially an additive hash function would be
> that both directions of the same flow/connection would end up for
> processing on the same core?

Yes. For TCP workloads where you have a lot of shared state being
changed in the transmit and receive direction, there's a _lot_ of lock
contention going on all over the place.



-a



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-VmokKTSgY_3VHHdbG0BiteMp33BeyXvKPdDW6SMUD8SOksA>