Re: A BIT OT : PER PACKET LOAD BALANCING AND SLOW WAN LINKS

I think the point probably was meant to be that it's generally
speaking *not* appropriate for high-speed links carrying lots of
different traffic flows, where there are likely better ways to load
balance. That's not to say that it necessarily *is* appropriate for
lower-speed links, IMHO, but rather, under a limited set of
circumstances, it *might* be. In cases where I need to load balance a
single flow or a low number of flows over a couple of low-speed links,
I generally turn to MLPPP. That way you don't hose UDP (etc) with
unordered packets. Yes, you pay a small tax in overhead for this
benefit.

It's been a good while now, but I occasionally deploy multicast for a
client to support a video feed from a satellite downlink. In one case
I had to backdoor this feed offsite from where the satellite receiver
was located to a facility where I had two T-1s available. The video
feed was roughly 2.5 Mbps. I initially attempted per-packet load
balancing (I did mention this was some years ago didn't I?) but the
codecs in the Set Top Boxes would not lock up on the unordered packets
coming their way. MLPPP solved that little problem quite nicely.

Bottom line is that you need to know your traffic flows before
considering such an approach and there are viable alternatives out
there.

On Sep 7, 2009, at 4:18 , wdf wefwe wrote:

> Hi all ,
>
> can anyone pls tell me why per-packet load balancing is suitable for
> slow wan links (from page 339 of Diane Teare CCDA book)
>
> thanks
>
> CCNAx3 (Voice , Security , Wireless)
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Mon Sep 07 2009 - 16:41:19 ART