Fast file input / output is less relevant to the specific API calls you make, but rather about how you archive your I / O application.
If you perform all the I / O on a single thread in a sequential manner, for example
- Reading a block into memory
- The processing unit in memory is somehow
- Write to file
- Repeat until the end ...
You are experiencing system I / O throughput in a single thread processing cycle. An alternative but more sophisticated design is a multi-threaded application to maximize throughput and avoid latency. This allows the system to simultaneously use both processor bandwidth and I / O controller bandwidth. A typical design for this would look something like this:
, -. , , /. , , - , . , , .
, / - API - .
.