This paper presents the Hadoop Online Prototype(HOP), which is a pipelined version of MapReduce. It allows maps to operate on infinite data and reduces to export early answers. To support pipelining, HOP modified the upstream task to push data to the downstream task, as it is produced. Pipelining delivers data to downstream operators more promptly. Not only the immediate downstream worker can start before the previous worker completes its work on the entire data, several other downstream workers in the chain after the immediate downstream worker can also get started as well. To fully utilize the benefit of the combiner[1], HOP will wait for the buffer to grow to a threshold size, instead of sending the buffer contents to reducers directly. To simplify the fault tolerance of HOP, the reducer treats the output of a pipelined map task as tentative until the master informs the reducer. Thus, the reducer could just ignore any tentative spill[2] files produced by the failed map attempt.