Cluster-based input/output trace synthesis

I/O traces are crucial for understanding the performance of new storage architectures. Unfortunately, traces are extremely bursty and difficult to characterize. They are large, difficult to obtain, and unwieldy. In this paper, we examine a method of trace synthesis based on cluster analysis of the time-varying characteristics of the trace. Representative trace segments are selected, and a synthesized trace is recon-structed from the segments. We show that we can achieve a 5–10 % demerit factor for I/O response times with a reduction of data volume of 75–90%.