ASU Logo

Partha Dasgupta

   

Home Up Publications Teaching Service Vita Columns

 

The Milan Project

MILAN (Metacomputing In Large Asynchronous Networks) is a joint project of Arizona State University and New York University. The primary objective of the MILAN project is to provide middleware layers that enables the efficient, predictable execution of applications on an unreliable and dynamically changing set of machines. Such a middleware layer, in effect creates a metacomputer, that is a reliable stable platform for the execution of applications. These “metacomputing environments,” built from commodity machines present unique challenges for constructing metacomputers on them, because the component machines and networks may: (1) exhibit wide variations in performance and capacity, (2) become unavailable either partially or completely because of their use for other (non-metacomputing related) tasks. These challenges force parallel applications running on metacomputers to deal with an unreliable, dynamically changing set of machines and have thus, limited their use on all but the most decoupled of parallel computations.

The MILAN project has developed innovative mechanisms and algorithms the overcome these challenges. These techniques include eager scheduling, TIES, efficient, reliable distributed shared memory, programming interfaces, pre-emptive scheduling, granularity management, distributed cactus stacks, tunability, adaptivity and fault-tolerance.

Systems implemented under the MILAN project include Calypso, Chime, Malaxis, Charlotte and various other prototypes:

More Information: 
>>
Leaflets (short, glossy) on:  MILAN, Chime, Scheduling.
>> Papers on MILAN, Calypso, Chime, Scheduling.

Web Site at ASU
Web Site at NYU