On Fri, May 13, 2011 at 7:35 AM, Rick Jones <span dir="ltr"><<a href="mailto:rick.jones2@hp.com" target="_blank">rick.jones2@hp.com</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>> For a short time, servers with gigabit NICs suffered but smarter NICs</div><div>
> were developed (TSO, LRO, other TLAs) and OSs upgraded to support them<br>
> and I believe it is no longer a significant issue.<br>
<br>
</div>Are TSO and LRO going to be sufficient at 40 and 100 GbE? Cores aren't<br>
getting any faster. Only more plentiful.</blockquote><div><br></div><div> NICs seem to be responding by hashing incoming 5-tuples to distribute flows across cores.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
And while it isn't the<br>
strongest point in the world, one might even argue that the need to use<br>
TSO/LRO to achieve performance hinders new transport protocol adoption -<br>
the presence of NIC offloads for only TCP (or UDP) leaves a new<br>
transport protocol (perhaps SCTP) at a disadvantage.</blockquote><div><br></div><div> True, and even UDP seems to be often blocked for anything other than DNS.</div></div>