Prefetching Mechanism for On-demand Software Streaming

In recent years, the “software as a service”, largely enabled by the Internet, has become an innovative software delivery model to provide network and service accessing of software. Dynamic on-demand deployment of software is a key method to achieve the above delivery model. Software streaming delivering is needed to support this deployment manner. During the streaming delivering of software, the execution waits until the missing data block is downloaded, which greatly influences the execution performance and user experience. A prefetching mechanism is presented for software streaming delivering based on N-Gram prediction model and an incremental data mining algorithm. By using historical access logs for data mining, then dynamically updating and polishing the prefetching rules, the proposed prefetching framework supports both file-level prefetching and block-level prefetching. The experimental results show that this prefetch-enable filesystem achieves a launch time reduced by 10% to 50%, as well as hit rate between 81% and 97%.