ReStore: Reusing Results of MapReduce Jobs



Analyzing large scale data has emerged as an important activity for many organizations in the past few years. This large scale data analysis is facilitated by the MapReduce programming and execution model and its implementations, most notably Hadoop. Users of MapReduce often have analysis tasks that are too complex to express as individual MapReduce jobs. Instead, they use highlevel query languages such as Pig, Hive, or Jaql to express their complex tasks. The compilers of these languages translate queries into workflows of MapReduce jobs. Each job in these workflows reads its input from the distributed file system used by the MapReduce system and produces output that is stored in this distributed file system and read as input by the next job in the workflow. The current practice is to delete these intermediate results from the distributed file system at the end of executing the workflow. One way to improve the performance of workflows of MapReduce jobs is to keep these intermediate results and reuse them for future workflows submitted to the system. In this project we build ReStore, a system that manages the storage and reuse of such intermediate results by employing execution coat and monetary cost optimization techniques. ReStore optimizes these workflows of MapReduce jobs by exploiting the following: (1) store the outputs of jobs and sub-jobs in the workflow to reuse them to rewrite future queries, (2) use a machine learning model to predict the cost of alternate query rewritings and choose the the plan with best expected performance, and (3) use estimated output sizes of jobs and sub-jobs in the workflow, their predicted execution time, and monetary cost of executing jobs in the cluster and storing their output to decide on which plans to materialize. We have implemented ReStore as an extension to the Pig dataflow system on top of Hadoop, and have conducted experiments on a private cloud and on EC2 and we uses data sets and queries from benchmarks such as PigMix and a Pig Latin representation of TPC-H.

Project Members

Iman Elghandour, Assistant Professor, Alexandria University, Egypt.

Ahmed E. Khalifa, Research Assistant, Alexandria University, Egypt.

Past Collaborators

Ashraf Aboulnaga, Associate Professor, University of Waterloo, Canada (Supervisor at UW).

Publications

Ahmed Khalifa, Iman Elghandour, and Nagwa El-Makky. IncReStore: Incremental Computation of MapReduce Workows . In Proc. IEEE Int. Conf. on Data Engineering Workshops (ICDEW) , Helsinki, Finland, 2016.

Iman Elghandour and Ashraf Aboulnaga. “ReStore: Reusing results of MapReduce jobs,” In Proceedings of the VLDB Endowment (PVLDB), 5(6):586–597, 2012. Presented at the International Conference on Very Large Data Bases (VLDB '12), Istanbul, Turkey, August 2012.

Iman Elghandour and Ashraf Aboulnaga. “ReStore: Reusing results of MapReduce jobs in Pig,” Demonstration at the ACM SIGMOD International Conference on Management of Data (SIGMOD ‘12), Scottsdale, USA, May 2012.