Background

Recent advances in parallel and distributed computing architectures made evident that there is a huge gap between the rough computing power available and the programming models and tools available to develop high performance applications on these architectures.

In the field of programming models, advanced models (algorithmic skeletons, design patterns, components) are becoming more and more interesting and look like to be able to overcome traditional models (plain message passing or shared memory models as well as data parallel ones, such as HPF) in a short time. In the field of compiling techniques, new techniques are being designed that take advantage of the knowledge relative to the underlying target architecture as well as of the knowledge derived from (structured) source code to produce highly optimized object code. Concerning run time supports, layered implementation techniques, frameworks and JIT techniques all provide effective mechanisms and ways of implementing very efficient, portable and extensible run time supports for high performance programming environments. Last but not least, computing GRIDs introduced new problems, such as those related to heterogeneity and dynamicity that must be faced and solved in order to be able to take complete control (and advantage) of these new distributed architectures.

On the application side, irregular problems (Barnes Hut, Fast Multipole Method, Adaptive Multigrid Methods for the solution of partial differential equations) pose several problems with respect to both memory constraint and completion time when uniprocessor architectures are considered. As a matter of fact, only parallel machines provide the computational power and the main memory capacity required to solve such large scale problems. Due to the features of irregular problems, the parallel solutions based on the most common current high level parallel programming tools (like HPF or Open MP) achieve low efficiency, and low level tools, like MPI or a shared memory, are required, at the moment, to achieve reasonable performance. Therefore a standard parallelization methodology and or higher level programming tools are needed to cope with such irregular problems.

The research group is closely following the research activity in these fields and actually, most of the members are involved in research activity concerning some of these topics. Group members have been and currently are responsible of different activities in several national research projects, either as national coordinators or as responsible of project work-packages involving different university, national research council and industrial partners. In particular, M. Vanneschi is the coordinator of the big national FIRB project GRID.it. Also, group members have been involved in several activities in the EU research framework on grids and high performance architectures. In particular, M. Danelutto participated to the NGG3 (Next Generation Grid Expert Group) meetings that lead to the document ``Future for European Grids: GRIDs and Service Oriented Knowledge Utilities, Vision and Research Directions 2010 and Beyond'' (available at
http://cordis.europa.eu/ist/grids/ngg.htm
) setting up the scene for the grid related research in Europe for the next decade.

The research group is currently also involved in several distinct EU funded projects:

• CoreGRID, an EU Network of Excellence (NoE) on Foundations, Software Infrastructures and Applications for large scale distributed, GRID and Peer-to-Peer Technologies. The Network is operated as a European Research Laboratory (known as the CoreGRID Research Laboratory) having six institutes mapped to the areas that have been identified in the joint programme of activity. One of these institutes, namely the one on Programming models is currently leaded by M. Danelutto. Within the CoreGRID NoE, the research group is actively cooperating with INRIA (Sophia Antipolis and Rennes groups) (F), with the University of Muenster (D), with the Haute Ecole Specialisée de Suisse Occidentale in Fribourg (CH) and it is currently having research contacts with several other partners of the NoE.
• GridCoord, an EU Special Support Action aiming at strengthening the co-operation amongst the funding authorities in order to better co-ordinate the planning of future activities in the field of Grid research, an ERA objective, at enhancing the already ongoing collaboration among the research actors and users and at developing, based on the above, national and EU Programme visions and roadmaps enabling Europe to play a leadership role in Grid technologies and applications. M. Vanneschi has been the coordinator of this SSA.
• GridComp, an EU STREP aiming at the design and implementation of a component based framework suitable to support the development of efficient grid applications. The framework basically derives from the one designed within the CoreGRID Programming model Institute and it will implement a kind of ``invisible grid'' concept as it will properly abstract all those specific grid related implementation details that usually require high programming efforts to be dealt with. M. Danelutto leads the Work Package 3 on Non functional component features.
• BEinGRID, an EU IP on Business Experiments in GRID. The main objective of the BEinGRID project is to foster the adoption of the so-called Next Generation Grid technologies by the realization of several business experiments and the creation of a toolset repository of Grid middleware upper layers. BEinGRID will undertake a series of targeted business experiment pilots designed to implement and deploy Grid solutions in a broad spectrum of European business sectors (entertainment, financial, chemistry, etc). BEinGRID Exploit European Grid middleware by creating a tool set of Grid services from across the Grid research domain and to use these services to deliver a set of successful business experiments that stimulate the early adoption of Grid technologies across the EU.
• SFIDA, a MIUR FAR-ITC on Soluzioni informatiche per FIliere, Distretti ed Associazioni di piccole-medie imprese, aiming at developing a Grid-based interoperable platform able to support next generation applications specifically addressing the needs of SMEs, such as Supply Chain Management applications matching various typical industrial cases, spanning from automotive, textile, food, white goods, and media retail. The SFIDA architecture should support componentization (e-services), intelligence (mining), collaboration (c-business), and customer business-processes orientation concepts on top of ASSIST, a Grid-aware high-level programming environment.