Azarias Reda, Edward Cutrell, Brian Noble
Web acceleration mechanisms play an important role in challenged network environments where connectivity is limited or expensive. However, as web usage gets increasingly personal and fragmented, traditional web acceleration systems that leverage redundancy in user requests to optimize performance fi.nd it difi.cult to perform well. This is unfortunate because personalization is an otherwise important trend that allows users to focus on content that is relevant to them. To start tackling this growing problem, this paper makes three contributions. First, we provide the .rst personalized, large scale web usage data in a developing country context. This allows researchers to get a nuanced understanding of access behavior that is not o.ered by aggregate data. Second, we present some analysis on this dataset, which provides tangible evidence for describing the increasingly fragmented and personal nature of web access even in developing countries. Finally, based on lessons learned from the analysis, we provide some recommendations for building eff.ective web acceleration mechanisms in the face of an increasingly personal web. We believe the next generation of web acceleration systems for challenged networks need to have a strong personal component.