An analysis of trace data for predictive file caching in mobile computing

摘要: One way to provide mobile computers with access the resources of a network, even in absence communication, is predict which information will be used during disconnection and cache appropriate data while still connected. To determine feasibility this approach, traces file-access activity for three diverse application domains were collected periods over two months. Analysis these using traditional new measures reveals that user working sets tend small compared modern disk sizes, users reference same files several days or weeks at time, different do not write file except highly constrained circumstances. These factors encourage conclusion an automated caching system can built wide variety environments.