资 源 简 介
Introduction
This project provides a highly optimized MapReduce runtime library for Multi-core architectures. The main features are as follows:
Modularized design, users can independently develop & execute map/reduce modules without recompiling the runtime.
Specialized in data-intensive computing. Typical applications are WordCount, inverted index, and so forth.
High performance, twilight-mapreduce can be 2x~3x faster than its counterparts.
This work is under the supervision of professor Wei Zhou.
How to use this library
A job can be submitted to the twilight runtime by providing three modules: (1) an application modules that implements the Map(), Reduce(), and Partition() functions, (2) an input data reader, and (3) an output data writer. These modules are compiled into dynamically linked libraries, which are loaded by the runtime library once the job is sumbitted.
Before we present a concret
