Lack of a single unified programming language has been a key obstacle to the production of a robust middleware infrastructure for wide-area distributed computing. Operational computational grids have been built from a collection of software systems but the brittleness of these systems remains a barrier to their more widespread and routine uptake by applications developers and end-users. Commercially viable cloud computing requires a stable and well-integrated software framework. The Erlang programming language is a strong candidate for building production quality middleware. The Erlang process-based concurrency model has been used to reimplement metacomputing infrastructure models such as that of the Distributed Information Systems Control World. This article presents an architectural analysis and some performance results from an Erlang middleware making use of distributed tuple-spaces for resource discovery and management. We also explore Erlang provisions for in-situ upgrading of a midleware software daemon while it is running live.
Keywords: metacomputing; computational grid; process control model; Erlang language; concurrency.
Full Document Text: Not yet available.