Parallel Processing of HTTP Requests in E-Commerce: A Modeling Framework

Abstract

Since e-Commerce has become a prevalent form of doing business, the workload posed to e-Commerce websites, especially the global ones, increases continually on a daily basis. In order to maintain high QoS levels and keep pace with ever-increasing service demands, e-Commerce systems have to exhibit high performances. Within the paper, we propose a modeling framework for performance evaluation of generic e-Commerce systems based on the utilization of Client/Server Interaction Diagrams (CSIDs) and the class of Non-Markovian Stochastic Petri Nets (NMSPNs). CSIDs are used for capturing the interaction between the client and particular servers for specific HTTP requests, corresponding to specific e-Commerce functions that e-Customers invoke during online shopping sessions. As an example, we represent the CSID for the SEARCH function using NMSPNs, thus providing a performance model suitable for estimating the speedup gains of e-Commerce systems that utilize multicore CPUs and parallel processing of HTTP requests within their web servers.