Big Data

Accelerating Data Transfers In Iterative MapReduce Framework

Download Now Date Added: May 2012
Format: PDF

MapReduce has become popular in recent years due to its attractive programming interface with scalability and reliability in processing big data problems. Recently several iterative MapReduce frameworks including their Twister system have emerged to improve the performance on many important data mining applications. Utilizing local memory on each compute node to cache invariant data, the authors are able to accelerate MapReduce execution but they still find performance issues when transferring massive data between or during iterations.