High-Performance Data Grid Toolkit
A SciDAC Program Project
Constructing the distributed computing machinery to support SciDAC
goals requires efficient, high-performance, reliable, secure, and
policy-aware management of large-scale data movement. An effective
solution requires an integrated approach to the development of "Data
Grid" infrastructure that recognizes and exploits commonalities
in two independent dimensions: among different levels of the "service
stack" so that network, middleware, and application-specific
services are developed in a coordinated fashion; and across different
applications, in order to develop infrastructure that is truly general
purpose.
The goal is to develop a set of essential network and
middleware services required to support large-scale, data-intensive
Collaboratory applications.
The objectives of this project are
- Develop new protocols taking advantage of unique properties
of data grids
- Develop innovative techniques for co-reservation of computer,
network, and storage resources with guarantees for each stage
- Develop market brokering services to make efficient decisions
in the face of constraints, when multiple resources are available
to fulfill a request
- Investigate variants of two-phase I/O strategies used in parallel
I/O systems for data transfer optimization
- Develop intelligent, adaptive recovery and performance strategies
based on knowledge of end-to-end routes and guarantees
This project is a collaboration among Argonne National Laboratory,
University of Southern California, and the University of Wisconsin.
For more info, see:
|