There remains a significant gap in grid and distributed computing between what can be achieved by advanced research groups and what is realized by typical scientists on the desktop. We argue that Java-based tuplespaces (JTS) can shrink this gap by lessening the impediments of complexity and institutional buy-in which traditionally accompany the development and use of distributed systems. Drawing from our benchmarking experience, we analyze the strengths of JTS for node configuration, scalability, and failure recovery. We then illustrate how JTS were leveraged to rapidly prototype a powerful end-to-end Virtual Observatory analysis thread within the Chandra Data Analysis System.
Astronomical Data Analysis Software and Systems XII
- Pub Date: