Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
Cluster I/O with River: Making the Fast Case Common Remzi H. ArpaciDusseau, Eric Anderson, Noah Treuhaft,
 

Summary: Cluster I/O with River: Making the Fast Case Common
Remzi H. Arpaci­Dusseau, Eric Anderson, Noah Treuhaft,
David E. Culler, Joseph M. Hellerstein, David Patterson, Kathy Yelick
Computer Science Division
University of California, Berkeley
Abstract
We introduce River, a data­flow programming environment and I/O
substrate for clusters of computers. River is designed to provide max­
imum performance in the common case --- even in the face of non­
uniformities in hardware, software, and workload. River is based on
two simple design features: a high­performance distributed queue, and
a storage redundancy mechanism called graduated declustering. We
have implemented a number of data­intensive applications on River,
which validate our design with near­ideal performance in a variety of
non­uniform performance scenarios.
1 Introduction
Scalable I/O systems form the basis for much of the high­
performance computing market. In recent years, manufactur­
ers have found that growth in customer appetite for I/O capacity
is outstripping Moore's law [40]. Cluster systems are a key

  

Source: Arpaci-Dusseau, Remzi - Department of Computer Sciences, Department of Computer Sciences, University of Wisconsin at Madison
Hellerstein, Joseph M. - Department of Electrical Engineering and Computer Sciences, University of California at Berkeley
Zakhor, Avideh - Department of Electrical Engineering and Computer Sciences, University of California at Berkeley

 

Collections: Computer Technologies and Information Sciences; Engineering