The unpredictable workload generated by surfers on an Electronic commerce (EC) site poses challenges to web sites maintenance. Proactive maintenance on web site is important. This depends much on how to best characterize its workload profile. An EC site would consist of a list of core EC processes, the demand on these EC functions forecast is critical for sustaining the site performance within the Service Level Agreements (SLA). The service demand of on-line functions determines the resources utilization that supports the operations. In e-market, online companies must meet customers? expectation on Quality of Services (QoS) to sustain its profit. This paper attempts to propose a model to proactively anticipating performance problems with the waves of changing customers? demand. A web mining approach is applied on the server?s log file of an empirical web site. The resulting knowledge would foster the development of better management on web resources of EC sites.