ML grid programming with ConCert

  • Authors:
  • Tom Murphy, VII

  • Affiliations:
  • Carnegie Mellon University, Pittsburgh, PA

  • Venue:
  • Proceedings of the 2006 workshop on ML
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Grid computing has become increasingly popular with the growth of the Internet, especially in large-scale scientific computation. Computational Grids are characterized by their scale, their heterogeneity, and their unreliability, making the creation of Grid software quite a challenge. Security concerns make the deployment of Grid infrastructure similarly daunting.We argue that functional programming techniques, both well-known and new, make an excellent practical foundation for Grid computing. We present a prototype Grid framework called ConCert built entirely in Standard ML which allows for the trustless dissemination of Grid programs through the use of certified code. The framework is fault-tolerant and relatively easy to implement, owing to a simplified network abstraction. This network abstraction is tedious to program for directly, so we present a high level ML-like language Grid/ML and a compiler Hemlock for the language.