USENIX warts

Neal Cardwell (cardwell@cs.washington.edu)
Mon, 27 Apr 1998 21:32:32 -0700 (PDT)

Some of the papers in this year's USENIX sound a *lot* like some of the
wart and traffic shaping ideas we've discussed. They even propose a name
that makes a good deal more sense than "wart": "transformer tunnel". Does
anyone know anything about these? Luckily none of them seem to touch on
the overlay network stuff...

"Transformer Tunnels: A Framework for Providing Route Specific
Adaptations"
http://www.cs.rutgers.edu/~sudame/abstracts.html
(no .ps up yet?)

Abstract: In a network using links with diverse properties, packet flow
that is fine tuned for some links (by selecting proper packet size,
transmission rate, encryption used, etc.) may be inappropriate for other
links. Ability to change the flow properties over segments of the network
allows flows with different characteristics to coexist; making it possible
to adapt to diverse link properties. Application specific adaptation
mechanisms (such as proxies), do not force adaptations on all the packets
flowing over the link, and are therefore insufficient for this purpose. We
propose transformer tunnels which force adaptations on all the packets
flowing through them. By attaching various transformation functions to
transformer tunnels, we can fine tune the flow properties with low
overhead. Transformer tunnels can coexist with proxies as the adaptations
provided by both are independent of each other. We also provide an API for
developing transformation functions. We have implemented transformer
tunnels in our wireless network, and present the effects of mobile hosts
using this mechanism to transform flows over the last hop link for
reducing losses during handoffs, and for improving the link utilization.

"Increasing Effective Link Bandwidth by Supressing Replicated Data":

http://www.sds.lcs.mit.edu/~jrsantos/usenix98.ps

"A Framework for Alternate Queueing: Towards Traffic Management by PC-UNIX
Based Routers":

ftp://ftp.csl.sony.co.jp/pub/kjc/papers/altq98.ps.gz