Graph-based techniques and analysis have been used for IP network traffic analysis. The objective of this paper is to study the hosts’ interaction behavior and use graph clustering algorithm, the Markov clustering algorithm, to group (cluster) hosts which have interaction using the HTTP protocol. Using real network traces, the clustering results show that MCL algorithm successfully group the hosts to their corresponding clusters. Analyzing the clustering results, it is showed that communications between one source IP address to one destination IP address, one source IP address to several (different) destination IP addresses, and several (different) source IP addresses to one destination IP address, are grouped to their own clusters. General Terms Algorithms, Measurement