Date: Thu, 23 Dec 2010 10:30:22 +0100 From: Fabien Thomas <fabien.thomas@netasq.com> To: Eugene Grosbein <egrosbein@rdtc.ru> Cc: Jack F Vogel <jfv@freebsd.org>, net@freebsd.org Subject: Re: lagg/lacp poor traffic distribution Message-ID: <AE4B4F6B-D0EF-4714-8FC7-54DCD6EE0331@netasq.com> In-Reply-To: <4D123B7C.1070104@rdtc.ru> References: <4D0CFEFF.3000902@rdtc.ru> <1292844095.1917.136.camel@stormi> <4D1083D6.6010707@rdtc.ru> <84530C06-AC2E-4E2B-BFD4-693902BB0FA6@netasq.com> <4D10B2F4.1060404@rdtc.ru> <90F64F73-F9A8-4FD6-9303-2FC0D3424751@netasq.com> <4D10BE3E.6030506@rdtc.ru> <BDB1F984-9F2A-4746-B74E-4CABC952D3E2@netasq.com> <4D123B7C.1070104@rdtc.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
On Dec 22, 2010, at 6:55 PM, Eugene Grosbein wrote: > On 21.12.2010 21:57, Fabien Thomas wrote: >=20 >>> irq262: igb0:que 0 157354922 7927 >>> irq263: igb0:que 1 577369 29 >>> irq264: igb0:que 2 280207 14 >>> irq265: igb0:que 3 241826 12 >>> irq266: igb0:link 2 0 >>> irq267: igb1:que 0 164620363 8293 >>> irq268: igb1:que 1 238678 12 >>> irq269: igb1:que 2 248478 12 >>> irq270: igb1:que 3 762453 38 >>> irq271: igb1:link 3 0 >>> cpu2: timer 39576052 1993 >>> cpu3: timer 39576095 1993 >>> cpu1: timer 39575913 1993 >>> Total 989503327 49849 >>>=20 >>> It seems I have four queues per igb card but only one of them works? >>=20 >> Yes. >>=20 >> Jack will certainly confirm but it seems that RSS hash does not seems = to take vlan in account and default to queue0 ? >=20 > I've just read "Microsoft Receive-Side Scaling" documentation, > = http://download.microsoft.com/download/5/d/6/5d6eaf2b-7ddf-476b-93dc-7cf00= 72878e6/ndis_rss.doc >=20 > RSS defines that hash function may take IP and optionally port numbers = only, not vlan tags. > In case of PPPoE-only traffic this card's ability to classify traffic = voids. > Then, unpatched lagg fails to share load over outgoing interface = ports. >=20 > It seems, we really need sysctl disabling lagg's use of flows, don't = we? Yes I think that it is necessary to be able to disable it because he = cant be always optimal. One improvement to the queue count would be to hash the queue id before = the modulo. >=20 > Eugene Grosbein
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AE4B4F6B-D0EF-4714-8FC7-54DCD6EE0331>