To minimize write times before inverters reduce the data, each parser writes
its segment files to its local disk. In the reduce phase, the master communicates
to an inverter the locations of the relevant segment files (e.g., of the r
segment files of the a–f partition). Each segment file only requires one sequential
read because all data relevant to a particular inverter were written
to a single segment file by the parser. This setup minimizes the amount of
network traffic needed during indexing.