Nexus5 사용자들의 실시간 IO trace 분석

Abstract

As the number of smartphone users increase, the number of devices with high computation power and applications that exploits network and storage are also increasing; however, as multiples of applications are running concurrently fighting for limited resources such as CPU and storage, it is becoming more difficult to satisfy the needs of everyday user’s needs. Although manufacturers and application developers are striving to improve the efficiency and the performance of applications, it seems like their effort is far behind in giving resolutions to low IO performance problem. In this paper, we capture and analyze IO traces of two users, which exhibit different IO patterns; user1 mostly used smartphone to listen to music or read stories on Facebook app and user2 used DMB or IPTV to watch videos. As a result, we find that about 59% of all traffics are generated during 10:00 to 11:00 and 19:00 to 22:00; /data partition received 37% of all IO traffic; write IOs are 12 times more than read IOs; about 33% of all IOs are accessing miscellaneous files; and, SQLite and SQLite journal accesses constitutes about 20% and 26% of total IOs, respectively.