Cache Predicting Algorithm Based on Context-Aware in Pervasive Computing

In pervasive computing, mobile device needs to make data access continuously, but in the influence of network and other factors, it could be disconnected. In order to support the continuous data access in case of disconnection, it needs to predict the possible access of data made by users and cache these data on the mobile client. Through data cache, it could store the data that the user could access in future on the client in advance, therefore the effects of disconnected mobile device on data access could be avoided and the quality of service could be accordingly raised. Among the present buffer management algorithms, the computing process either takes no account of context, or divides the data in groups to make a forecast separately according to the context information. These two methods could not fully consider the trend of the whole user access and the impact of user context on the predicted consequences. This paper puts forward a buffer management algorithm, which establishes the association between various data on the basis of data access records and makes data group; after computing the accessing frequency of data sets over the current contexts, it makes cache replacement of the results in terms of cache residence time and accessing frequency. The results of simulation tests show that this kind of algorithm could effectively improve the cache hit rate in the case of disconnected operation for handheld mobile devices, and better support the disconnected operation of mobile devices.