LOTS of people in the HPC community have done work towards tracing I/O patterns.
Tools for generating traces
- Darshan DXT is a tracer built into Darshan that can be enabled.
Analyzing traces
- Xiaosong Ma’s team developed a wavelet-based approach to pulling individual checkpoint bursts out of noisy server-side logs in 2016. (Server-Side Log Data Analytics for I/O Workload Characterization and Coordination on Large Shared Storage Systems