Generating Block IO Trace Data from a Cloud Site Using Packet Capture and Analyzing the IO Trace Data
Kazutaka Ogihara · 2019
To acquire storage workload, I captured whole packets between IA servers and network storage systems in one cloud site, and obtained block IO trace data by analyzing these captured packets. The cloud site consisted of about 100 physical servers and each physical server had some virtual machines (VMs). There were 2,000-3,000 virtual disks attached to the VMs. In this paper, I describe the following: 1. A method for generating block IO trace data from the captured packets from the cloud site using plural packet capture servers. 2. An analysis report from whole IO trace data, comparing this trace data and other trace data in public repositories.