The implementation of the BSP parallel computing model on the InteGrade Grid middleware

摘要

It is a great pleasure to introduce the ACM/IEEE/IFIP MiddlewareWorkshop Program for Grenoble 2005. This is the fourth time thatMiddleware has included a Workshop Program, and the levels ofinterest and the standard of submissions continue to be extremelyhigh. This year we have five workshops. The Workshop on Adaptiveand Reflective Middleware (ARM05) is our most venerableinstitution, and is now in its fourth incarnation, having startedout at Middleware 2000 in New York and run at every Middlewaresince. Two of the other workshops are also well established, havingrun previously in Rio and Toronto. These are the Workshop onMiddleware for Grid Computing (MGC05), and the Workshop onMiddleware for Pervasive and Ad-hoc Computing (MPAC05). MPAC thisyear is joined by Didier Donsez and his team whose proposal for aworkshop on 'Middleware for Sensor-Based Services' has been mergedinto MPAC. As a consequence, this event is running over two daysrather than the usual one day.>Another workshop thatis continuing a lively tradition is theDoctoral Symposium on Middleware. This popular event has workedparticularly well in the past and represents a wonderfulopportunity for young reseachers in our field to present their workto a mock thesis committee of mentors and to receive valuable andsupportive feeback. This year we also have a new workshop: the 1stWorkshop on Aspect Oriented Middleware Development (AOMD05). Thishighlights what seems to be a promising area of synergisticresearch that has recently been receiving a lot of attention. Welook forward to a productive dialogue on the relationship betweenaspectisation techniques and middleware techniques.>In order to control a broad base of resources, Grid Computinghas a middleware layer controlling the allocation of resources andthe distributed execution of applications. By exploring thepossibility of a wider range of applications specially based ondata grids and content networks, a larger diversity of resourcesand devices can be thought participating ina grid system.>The first MGC workshop held on June 2003, in conjunction withthe Middleware Conference, in Rio de Janeiro, Brazil, MGC'2003, sawa variety of efforts in Middleware for Grid Computing. There were16 technical presentations and discussions on several stimulatingtopics including, Classic Grids, Object Oriented technologies,Service-based Grids, Open grid Service Architecture, Agent Grid,Interactive Grid, Grid Economy / Scheduling, and Portlets.Following the workshop, extended and thoroughly revised versions ofthe papers were invited to a Special Issue of Concurrency andComputation: Practice and Experience Journal published in 2004.>The 2nd workshop in the MGC series brought together researchesin the field of middleware for grid computing, addressing topicsthat emerged from 2003, and some novel ones such as, Strategies andProtocols for obtaining Quality of Services, Virtualization,Wireless Grids, Data Grid Middleware, Semantic Grid Middleware,Dependability and Fault Tolerance in Grid Middleware and ManagingInformation. This workshop received an unprecedented number ofhigh-quality submissions. Fifteen full papers and nine posters werechosen for the workshop proceedings out of the forty-eight papersoriginally submitted. The presenters highlighted issues andsolutions in one or more of the themes identified for the workshop.Following the second workshop, thoroughly revised selected paperswere invited to a Special Issue of Concurrency and Computation:Practice and Experience Journal to appear in 2005/2006.>MGC'2003 and MGC'2004, generated substantial interest in thecommunity and it is expected that this interest will continue. Farfrom exhausting the topics of interest, they have paved the way fora third edition of the workshop in the series. For MGC 2005workshop, researchers from the various GRID middleware communitieswere encouraged to submit and present original work to beconsidered for publication.>The goal of the third edition of the MGC workshop is to bringtogether researches inthe field of middleware, addressinglarge-scale and real world problems in grid environments, includingthe most interesting and stimulating topics emerged last year, andsome novel ones as Strategies and Protocols for obtaining Qualityof Services, Virtualization, Wireless Grids, Data Grid Middleware,Semantic Grid Middleware, Dependability and Fault Tolerance in GridMiddleware and Managing Information.>The third workshop received a good number of high-qualitysubmissions and sixteen papers were selected for these proceedingsout of the thirty-eight originally submitted. Additionally 4posters have been invited. The presenters highlight issues andsolutions in one or more of the themes identified for theworkshop.>As full papers there were contributions oniDependability and Fault Tolerance in GridMiddleware/i by Camargo, Cerqueira and Kon, presentingstrategies for storage of checkpointing data using non-dedicatedrepositories on grid systems. On iServiceDiscovery/i there wasa contribution by Banerjee et al.,exploring scalable grid service discovery based on UDDI. OniPortals/i there was a paper by Akram, Chohan andAllan. Addressing iPeer to Peer Protocols in GridComputing/i there were papers by Amoretti, Zanichelli andConte, Lima et al., and Santos et al. They present aservice-oriented framework for P2P-based grids, a peer-to-peerresource discovery in mobile grids, and an accurate autonomousaccounting in peer-to-peer grids. Exploring different topics andcommon issues related to iProgramming Models and Tools inGrid Middleware/i by Goldchleger, Godlman and Hayashida,Paventhan and Takeda, and Rodriguez, Milanes and Schulze. Theyfocus on a BSP parallel computing model implementation, on amutli-language CoG toolkit, and on managing jobs with aninterpreted language for dynamic adaptation. Addressing aspects oniAgent-based Approaches and Grid ApplicationFrameworks/i there were papers by Joita et al., d'Orazio etal., and Desertot, Escoffier and Donsez. They explore anapplication deployment using grid middleware, the building ofadaptable cache services, and autonomic management of J2EE edgeservers. On iResource Management and Scheduling/ithere were papers by Yarmolenko and Sakellariou, Vargas et al., andBoeres et al. They discuss aspects on the flexibility ofWS-agreement for job submission, on a hierarchical submission in agrid environment, and on hierarchical self-scheduling for MPIapplications executing in computational grids.>Additionally there were poster contributions by Lima et al., andYamin et al. on iProgramming Models and Tools in GridMiddleware/i, focusing on a framework for buildingcustomized grid environments and a middleware to support pervasivegrid applications. Pathak, and Bittencourt et al. present posterson commons issues on iResource Management andScheduling/i, focusing on dynamic resource management onthe grid and a path clustering heuristic for scheduling task graphsonto a grid.