Abstract

The open science communities supported by DOE/SC (Department of Energy–Office of Science) operate and use a large number of unique scientific instruments and a wide range of compute and storage resources. The DOE/SC ASCR (Advanced Scientific Computing Research) facilities division manages a collection of high-performance computing resources and a national high-performance network to link these compute resources and other scientific instruments to the national and international science communities located at DOE laboratories, universities, and other research centers attached to other Research and Education Networks. The ASCR program office research division funds research in several critical areas including high-performance networks and middleware services needed to ensure effective use of this complex infrastructure.
The Energy Science network (ESnet) is engineered to provide extremely high data transfer capacity. There are two major concerns however which are often voiced by DOE user communities.. The first concern is that end-to-end performance is not predictable due to a variety of factors. Even when some performance forecasts or predictions can be made, they often cannot explain the reasons why some predictions fail. While deploying measurement tools to capture and display time series throughput and/or delay data is a good first step, it is not sufficient to adequately determine if the observed behavior was correct, or how to predict what future behavior should be.

Date

September 11, 2012

Authors

Constantine Dovrolis, David M Nicol, George Riley, Jay Aikat, Prasad Calyam, David Clark, Ian Foster, Richard Fujimoto, Dan Gunter, John Heidemann, Raj Kettimuthu, Ezra Kissel, Kevin Mills, Jelena Mirkovic, Inder Monga, Kalyan Perumalla, George Polyzos, Taghrid Samak, Martin Swany, Venkatram Vishwanath, Wenji Wu, Zhi-Li Zhang, Richard Carlson, K Claffy, Srinivasan Keshav, Don Towsley, Peter Barnes, John Bodnar, Bryan Lyles, Raju Namburu, Ken Renard, Jun Yi, Dantong Yu

Journal

Washington DC.: US Department of Energy