Optimizing the Incremental Update Mechanism by Inlaying File Indexes on Flash Storage

With the aim of reducing network traffic and up-dating costs, emerging mobile systems generally use incremental update algorithms to upgrade the image of mobile applications. Existing incremental update algorithms reuse the data in the local old image of the application. In this paper, we analyze the file accesses patterns during incremental updates of applications and find that most data in the new image is from the old image. Thus, we propose a new file access mechanism to optimize the memory consumption of incremental update algorithms by inlaying the file indexes in the storage device. We implement the proposed mechanism in Linux kernel 5.19.1 and evaluate it with real-world images of applications. The experimental results show that the proposed mechanism can reduce 30% -80% memory footprints in comparison with HDiffPatch, the state-of-the-art incremental update algorithm.