Avetec

Advanced Virtual Engine Test Cell

Avetec Case Studies

NASA Center for Computational Sciences & University of Maryland: Distributed & Parallel File Systems

Overview

This was a joint project between the NASA Center for Computational Sciences (NCCS) at the NASA Goddard Space Flight Center (GSFC) and the University of Maryland Institute for Advanced Computer Studies (UMIACS) at College Park, Maryland.  The NCCS at NASA GSFC is dedicated to providing scientists and engineers with the supercomputing resources and simulation tools required to carry out critical NASA missions and to make new scientific discoveries. The mission of UMIACS is to foster and enhance interdisciplinary research and education in computing across the University of Maryland, College Park campus. To learn more about these organizations, visit www.gsfc.nasa.gov and www.umiacs.umd.edu.

Challenge

Develop technologies that allow easy sharing of high performance Storage Area Network (SAN) data in wide area environments, while maximizing network resource allocations and addressing issues of data security.

Objective

Develop a management framework that dynamically allocates and dismantles IP-based wide area SAN on an as-needed basis.

Results

A management framework was developed using Apache Tomcat and Web Applications built on WSS4J, which dynamically allocates and dismantles IP-based wide area SAN on an as-needed basis.  The framework proved to be reliable in simulated wide-area environments with latencies up to 160 msec and packet failures up to four percent. The overhead associated with managing inter-site SAN extensions through web services was found to be small in comparison to the transfer times of large data sets and the runtime of most data intensive applications. Future implementations will expand the test to a remote site to evaluate performance and usability over actual wide area environments.

Ultimate Value

The management framework developed allows easy and flexible data sharing in wide area environments while minimizing network resource requirements and security exposures. Large increases in network latency were found to have a relatively small impact on the overall performance. The system is reliable enough to be suitable for most production wide area networks.

 

Case Studies Home