Benchmarking - Large file transfers

Benchmark our transfer speed for large files. We mostly care about relative slow-down vs running the 'cp' command - so it can be on a local system.

We should benchmark:

• Client standalone
• Client + controller with backends
• Client + controller + worker

We should be able to explain the source of slowdown - for example for the standalone client there should be a file read and write with POSIX apis - we should try to explain why it might be slower than 'cp'. For the web versions we should try to determine the individual 'costs' - file read/write + network/socket transfer.