Transparent Gif

Department of Computer Science

University of California, Santa Barbara

Abstract

Multivariate Resource Performance Forecasting in the Network Weather Service

by: Martin Swany and Rich Wolski

Abstract:

This paper describes a new technique in the Network Weather Servicefor producing multi-variate forecasts. The new technique uses theNWS\'s univariate forecasters and emprically gathered CumulativeDistribution Functions (CDFs) to make predictions from correlatedmeasurement streams. Experimental results are shown in whichthroughput is predicted for long TCP/IP transfers from short NWSnetwork probes.

Keywords:

Supercomputing, Grid Computing, Distributed Computing, Performance Evaluation

Date:

May 2002

Document: 2002-12

XHTML Validation | CSS Validation
Updated 14-Nov-2005
Questions should be directed to: webmaster@cs.ucsb.edu