In systems performance I would advise to never think of any workload as unidimensional (ie: Any file system optimization can either improve IO latency or be useless)
Issuing individual truncates of 1B files can be just as much of a CPU problem then an IO one for example.
Comments
In systems performance I would advise to never think of any workload as unidimensional (ie: Any file system optimization can either improve IO latency or be useless)
Issuing individual truncates of 1B files can be just as much of a CPU problem then an IO one for example.
But why wouldn't using one of many CPU cores be sufficient?