Access Characteristic Guided Remote Swapping for User Experience Optimization on Mobile Devices

Abstract

With the rapid development of mobile devices, remote swapping has been widely studied across mobile devices. However, one challenge for remote swapping is its unsatisfying user experience. This is because remote swapping always requires a large amount of data swapping across devices. In this work, an access characteristic guided remote swapping scheme, ACR-Swap, is proposed to optimize user experience. This work is motivated by observations from our comprehensive studies on the access characteristics of existing remote swapping. First, the swap-in operations of system service processes are more frequent than that of the application-specific processes. Second, apps have a different amount of swap-in operations in different running periods. Based on the observations, ACR-Swap is designed with two schemes to optimize the remote swapping. First, a process-aware page sifting (PPS) scheme is designed to identify processes and determine data placement across devices. Second, an adaptive-granularity prefetching (AGP) scheme is proposed to prefetch data across devices based on the running period of apps. ACR-Swap is demonstrated on real mobile devices. Experimental results show that ACR-Swap can significantly reduce the app switching latency compared with the state-of-the-arts and improves the app caching capability, compared to no swapping.