Analyzing I/O Request Characteristics of a Mobile Messenger and Benchmark Framework for Serviceable Cold Storage
Cloud computing systems require massive storage infrastructure, which has significant implications on power bills, carbon emissions, and the logistics of data centers. Various proprietary ‚??cold storage‚?? services, based on spun-down disks or tapes, offer reduced tariffs, but also lead to extended times to first access. One way to improve cold storage systems is to build them on a file system that allows for the I/O patterns of storage devices. We have developed a cold storage testbed for mobile messenger services which takes into account the power consumption of each hard disk in the system. We analyzed a trace of I/O requests from a messenger service, and found that they had a strongly skewed Zipfian distribution, and that most of the stored data is cold. Current cloud benchmarking tools cannot reproduce this pattern of I/O. Therefore, we have developed a tool for benchmarking cold-storage systems that emulates this type of long-tail distribution, and can contribute to reducing the power consumption of mobile messenger services.
Cold Storage, Large-scale Storage Systems, Energy-efficiency, Statistical Analysis, Benchmark