CISE Research Resources: A Shared Data Cluster For Real Time Interaction With Massive Datasets

Northwestern University, Evanston IL

Investigators

Abstract

EIA-0130869 Benjamin Watson Northwestern University CISE Research Resources: A Shared Data Cluster for Real Time Interaction With Massive Datasets The research at Northwestern is increasingly concerned with the issues involved in providing interactive and affordable access to massive (terabyte scale) datasets. In particular, they focus on visualization of these datasets using perceptually based computer graphics techniques and display the visualizations using a combination PC CAVE and Active Mural. They also study the issues involved with ubiquitous access to high-resolution spatial datasets from mobile devices, and leverage a powerful wireless network. They focus on prediction-based adaptation and scheduling for distributed interactive applications and are building a compute cluster to support this work. What is missing in the existing shared research infrastructure is a facility to store and serve massive datasets interactively, with low latency and high bandwidth. The researchers will build such a shared data cluster. The data cluster will be very high performance and capacity PC RAIDs interconnected with a multi-GB/sec SAN. The SAN will connect to their compute cluster, CAVE/Mural, mobile devices and wired client machines. The combined data and compute clusters will be sufficient to pump 200 MB/s from disk to the CAVE/Mural screens. The combination of the data cluster, compute cluster, CAVE/Mural, local network infrastructure, and desktop and mobile clients will let them study interactivity in the large.