The bandwidth demands on the (WORLD-WIDE) WEB continue to grow at an expone
ntial rate. It is thus becoming crucial to provide solutions improving the
WEB latency. In that framework, the most promising low cost solution lies i
n the use of caches at the level of the clients, network and servers. Cachi
ng effectiveness then relies upon adequate cache management so as to keep i
n the cache the WEB objects that are the most likely to be re-accessed. How
ever, the effectiveness of a single cache remains poor as it is in general
no higher than 40%. One way to further improve caching effectiveness is thu
s to make caches cooperate so as to increase the probability of retrieving
an object at the caching level. The cache cooperation protocol must then be
such that it induces a negligible load for the network and cooperating cac
hes.
This paper presents our solutions to improving the effectiveness of WEB cac
hing concerning both cache management and cache cooperation. Regarding cach
e management, we propose two novel algorithms that exploit the latest resul
ts about WEB usage, enabling us to undertake replacement decisions that are
more accurate than the one taken by existing algorithms. From the standpoi
nt of cooperating caches, we propose a cooperation protocol, which minimize
s the associated network bandwidth, processing load, and storage consumptio
n among caches.