Review: XCP

Controlling congestion in future high bandwidth-delay networks using TCP
is focus of many researchers, Different techniquies have been purposed to
overcome the congestion or to avoid it before it occurs, These approaches
can be categorized into two area, protocols purposed for end host and
protocols purposed for routers. There are many purposed solutions to be
implemented at routers to control congestion. Amongst them are FQ, CSFQ,
FRED but they mainly focus to work on existing network protocols with
slight changes at the routers. This paper presents another congestion
control mechanism to be implemented that is different in the sense that it
has purposed to reimplement the whole congestion control mechanism from
scratch.

The author has purposed a eXplicit Control Protocol, that is envisioned to
replace TCP, that has the build in mechanims to explicitly communicate
congestion to end hosts. Using this protocol intermediate routers
incorporate the congestion information in packet header to signal the
congestion to end host. By using this approach there is no need to
maintian flow states at the router becasue each packet carries the
congestion information with it that the end hosts could use to respond to
the network congestion. Author has used simulations to show that the XCP
outperforms TCP not only in high bandwidth-delay scenarios but also in
simple scenarios.

Though the author has shown that the XCP performs well as compared to TCP
but the main concern is that it requires the change in the whole internet
architecture, that includes the replacement of millions of routers within
current infrastructure. My concern is why the companies will do that and
also it seems unrealistic to change the whole internet model that has
millions of end host and routers running. So though the algorithms work
perfectly in theory but due to its nature I believe it will never be
implemeted in practice.
Received on Tue Oct 03 2006 - 00:38:59 EDT