Some Related Work

• LYDIA: Cost Function Metacomputing

• SPAWN: Market-Based Metacomputing

• Condor: Homogeneous Metacomputing

• Globus: Infrastructure for Global Metacomputing

• Linda and JavaSpaces: Tuples for Metacomputing

Previous slide

Next slide

Back to first slide

View graphic version

LYDIA (http://www.ics.forth.gr/pleiades/projects/LYDIA/) studies single-resource resource allocation on a system with many different kinds of jobs. Each “class” of jobs has different performance expectations. It uses a homogeneous cost to evaluate system performance.

SPAWN (http://www.parc.xerox.com/istl/groups/iea/www/spawn.html) and similar systems provide computational markets where tasks bid competitively for resources. These systems integrate economic principles with computer science.

Condor (http://www.cs.wisc.edu/condor/) is a mature resource allocation system for metacomputers. It takes advantage of unused machines to perform work, handling homogeneous resource allocation smoothly on large numbers of machines.

The Globus Project (http://www.globus.org/) provides a set of tools for large-scale metacomputing. Their specific goals are orthogonal to ours, but their general mission is much the same.

Linda (http://www.cs.yale.edu/Linda/linda.html) and JavaSpaces (http://www.sun.com/jini/specs/js101.html/) are metacomputing systems built around the idea of a “tuple space” that stores data and code objects. Programs on different machines put objects in and retrieve objects from this space. This is a powerful distributed programming paradigm. If the relevant programs monitor local resources to avoid resource exhaustion, the system becomes a metacomputer.