Here you assume you are doing linear copy in a single copying thread. There are cases in which this assumption is untrue, for example, when copying large file from one storage cluster to another you may want to copy chunks in parallel between the chunk servers.