学术文献库

Distributed algorithms that operate in the fail-recovery model rely on the state stored in stable memory to guarantee the irreversibility of operations even in the presence of failures. The performance of these algorithms lean heavily on the performance of stable memory. Current storage technologies have a defined performance profile: data is accessed in blocks of hundreds or thousands of bytes, random access to these blocks is expensive and sequential access is somewhat better. File system implementations hide some of the performance limitations of the underlying storage devices using buffers and caches. However, fail-recovery distributed algorithms bypass some of these techniques and perform synchronous writes to be able to tolerate a failure during the write itself. Assuming the distributed system designer is able to buffer the algorithm's writes, we ask how buffer size and latency complement each other. In this paper we start to answer this question by characterizing the performance (throughput and latency) of typical stable memory devices using a representative set of current file systems.