Next: Implementation Issue: TCP Handoff
Up: Forwarding Requests to other
Previous: Which Requests
A simple solution to the proxy
selection question is to pick a proxy that is nearest to the proxy
where cache miss happens. Since this is a ``local'' decision within
the ISP's backbone, it is easy to implement in practice. An
alternative is to choose those proxies that are nearer to the client
such that user perceived latency could also be reduced. However,
depending on the environment, it is likely difficult to obtain
clients' geographical location information.
In summary, forwarding requests can reduce backbone bandwidth
consumption, but it may possibily increase client latency as well.
The tradeoff between bandwidth and latency is the major factor affecting
Although our mechanism is targeted at reducing backbone bandwidth for
populating the caches, it can also be applied in other situations, such
as proxy overload or network congestion.
However, how to coordinate proxies to obtain other
benefits is not covered in this paper.