log in  |  register  |  feedback?  |  help  |  web accessibility
Logo
Phoenix++: Modular MapReduce for Shared-Memory Systems
Ashwin K Kayyoor - University of Maryland, College Park
Wednesday, September 12, 2012, 1:00-2:00 pm Calendar
  • You are subscribed to this talk through .
  • You are watching this talk through .
  • You are subscribed to this talk. (unsubscribe, watch)
  • You are watching this talk. (unwatch, subscribe)
  • You are not subscribed to this talk. (watch, subscribe)
Abstract

Authors: Justin Talbot, Richard M. Yoo, Kristos Kozyrakis  (Stanford University).

Venue: Proceedings of the second international workshop on MapReduce and its applications, (MAPREDUCE) 2011.

 

Abstract:
 
This paper describes our rewrite of Phoenix, a MapReduce framework for shared-memory CMPs and SMPs. Despite successfully demonstrating the applicability of a MapReduce-style pipeline to shared-memory machines, Phoenix has a number of limitations; its uniform intermediate storage of key-value pairs, ineffi cient combiner implementation, and poor task overhead amortization fail to e fficiently support a wide range of MapReduce applications, encouraging users to manually circumvent the framework. We describe an alternative implementation, Phoenix++, that provides a modular, flexible pipeline that can be easily adapted by the user to the characteristics of a particular workload. Compared to
Phoenix, this new approach achieves a 4.7-fold performance improvement and increased scalability, while allowing users to write simple, strict MapReduce code.

 

This talk is organized by Abdul Quamar