Cisco's AIP vs HSSI

Kent W. England kwe at 6SigmaNets.com
Tue Oct 15 17:06:01 UTC 1996


At 09:46 AM 15-10-96 -0400, Padma Krishnaswamy wrote:
>One other point about the  AIPs relative to the  HSSI/DSU ; you may not be
>interested in this feature , but the AIPs implement traffic shaping- ie
You can
>set Peak and Average rates and burst size; you could use this to subrate your
>traffic -should you be in a situation where this is desirable.
>
>
While this is true, I would caution the novice that TCP sources don't really
benefit from rate shaping and the sort of burst capability you get with VBR.
TCP appears to work with the explicit rate control of ABR, but there are some
concerns in the research community that it might not work because both TCP
and ABR are working the congestion management problem simultaneously and there
may be pathological or even common situations where these congestion
management
schemes conflict or degrade each other. Time will tell.

To date, we know that TCP works well with UBR and large buffers. Large is
defined by TCP and the path length and works out roughly to twice the
bandwidth
delay product on the longest path.

--Kent






More information about the NANOG mailing list