Section: Overall Objectives
The recent evolutions in computer technology, as well as their diversification, goes with a tremendous change in the use of these architectures: applications and systems can now be designed at a much larger scale than before. This scaling evolution concerns at the same time the amount of data, the number and heterogeneity of processors, the number of users, and the geographical diversity of these users.
This race towards large scale computing questions many assumptions underlying parallel and distributed algorithms and operating middleware. Today, most software tools developed for average size systems cannot be run on large scale systems without a significant degradation of their performances.
The goal of the MESCAL project-team is to design and validate efficient exploitation mechanisms (middleware and system services) for large distributed infrastructures.
MESCAL's target applications are intensive scientific computations such as cellular micro-physiology, protein conformations, particle detection, combinatorial optimization, Monte Carlo simulations, and others. Such applications are constituted of a large set of independent, equal-sized tasks and therefore may benefit from large-scale computing platforms. Initially executed on large dedicated clusters (CRAY, IBM, COMPAQ), they have been recently deployed on collections of many-core architectures. The experience showed that such sytemrs offer a huge computing power at a very reasonable price. MESCAL's target infrastructures are aggregations of commodity components and/or commodity clusters at metropolitan, national or international scale. Examples of target infrastructures are grids obtained through sharing of available resources inside autonomous computing services, lightweight grids (such as the local CIMENT Grid) which are limited to trusted autonomous systems, clusters of intranet resources (Condor) or aggregation of Internet resources (SETI@home, XtremWeb).
MESCAL's methodology in order to ensure efficiency and scalability of proposed mechanisms is based on mathematical modeling and performance evaluation of target architectures, software layers and applications.