CernVM-FS Profiling

CernVM-FS is essential for distributing high-energy physics software stacks to distributed computing infrastructures. Even if CernVM-FS is highly optim- ized, there are still situations where its performance can be improved. This project has two main goals. First, we tried to create a general set of tools and procedures that can be used to analyze any situation where CernVM-FS becomes a bottleneck. Second, we applied these tools on an already known multiple process - multiple data benchmark and spot a bottleneck in CernVM- FS, caused by serialized data decompression in a cold cache scenario.