[e2e] Open the floodgate - back to 1st principles

Ghanwani, Anoop anoop.ghanwani at hp.com
Mon Apr 26 09:50:20 PDT 2004


> -----Original Message-----
> From: end2end-interest-bounces at postel.org 
> [mailto:end2end-interest-bounces at postel.org] On Behalf Of 
> Aroon Nataraj
> Sent: Sunday, April 25, 2004 8:15 PM
> To: Injong Rhee
> Cc: Guy T Almes; end2end-interest at postel.org
> Subject: Re: [e2e] Open the floodgate - back to 1st principles
> 
> 
> Injong Rhee wrote:
> 
> >Then again, use of multiple flows is sometimes the only 
> option with the
> >current NIC
> >performance. The current NIC (10Gbps) only gives around 
> 2-3Gbps (if you are
> >a nice
> >guy, you might get 4-6Gbps :-). With this NIC performance, acheiving
> >something like the
> >full bandwidth util. of  tens of Gbps is not possible. You 
> need to use
> >multiple NICs to achieve this. This end-host (NIC) 
> bottleneck forces you to
> >use multiple flows anyway...now the question is back to square one...
> >
> >  
> >
> Actually using multiple NICs doesnt require multiple TCP 
> level flows - 
> [bonding/teaming/aggregation of NICs] -- work at layer-2, to TCP  the 
> 'team' would appear as a single interface.

Even at L2, the algorithm used to do load balancing of 
traffic typically puts packets of a single flow on the 
same link within the aggregate in order to avoid reordering.
Packets within a single TCP connection would almost certainly
be treated as belonging to the same flow and would end up
being transmitted over only one of the links within the 
aggregate.  You would need multiple TCP-level flows
to see any benefit from link aggregation (at least the 
way it's defined in IEEE 802.3).

Anoop


More information about the end2end-interest mailing list