to reduce the intra cluster communication traffic and the DMA latency. The inclusive
scheme is particularly effective when the size of Web content is large and the popularity
skewness is high.
Future Research Directions
As the current trend of merging heterogeneous services into Web services contin
ues, the requirements of Web servers would become increasingly more diverse compared
to traditional Web servers. Additionally, security would become a highly critical issue
in most enterprise Web servers. Thus, we plan to extend our research to meet the future
demand imposed on Web servers.
First, our proposed scheme for SSL connections improves performance in a Web
server cluster. Based on these results, we will expand our work to a real deployment.
We will implement SSL enabled content aware and content oblivious webswitches. Since
there is no publicly available distributed Web server that provides SSL connections, we
plan to implement SSL functions in a distributed Web server.
Second, we will focus on Quality of Service (QoS) in cluster based Web servers
with respect to required contents, clients groups and service availability. Providing QoS
not only in network transmission, but also in the processing capacity of Web servers
has become highly desirable. The previous studies on QoS in cluster based Web servers
suggested that the nodes in a cluster are divided according to the required service load for
each class. But this coarse grain partitioning may cause low utilization of Web servers
and may not provide flexible adjustment on the current service requirements. Thus,
we plan to propose a more adaptive and fine grain QoS scheme based on current local