Naiad: Incremental and Iterative Data-Parallel Computation

Naiad is a distributed system that supports efficient incremental and iterative data-parallel computation. It extends standard batch data-parallel processing models like MapReduce, Hadoop, and Dryad/DryadLINQ to support efficient incremental updates in the manner of a stream processing system, while at the same time enabling arbitrarily nested fixed-point iteration. This talk describes the design and implementation of Naiad, and shows that complex analyses like strongly connected components, requiring multiple nested loops, run on streams of input with subsecond update times.

Speaker Bios

Rebecca Isaacs is a researcher at MSR Cambridge in the systems and networking group, where she works on performance analysis and debugging of operating systems and networks. She is currently working on the Barrelfish operating system, previous projects include Constellation and Magpie. She has been at MSRC for 8 years, and has a PhD from Cambridge.

Date:
Haut-parleurs:
Rebecca Isaacs
Affiliation:
MSR Cambridge