Back to home page

Research activities

 

Collaborations

Researches described below have been done in close collaboration with co-authors. I would especially like to acknowledge the PhD students I supervise(d) :

Research actions

Emerging media and high performance communication infrastructures development (Internet, Intranets, metropolitan and local networks) require database research to face three important issues : the increase of the volume of data managed in very large databases ; the increase of the transactional load supported by the data servers ; the complexity of new queries and new operators. In this context, we focus our activities on collaborative multimedia information services. After studying parallel multimedia servers, our work now tackle the problems of the modeling and processing of query by image content systems and server collaboration.

Once upon a time... (about distributed shared memory systems)

In parallel with our work on multimedia systems, we have conducted, until 1999, researches on parallel programming environments, more specifically on distributed shared memory (DSM) systems. The system we have developed, called DOSMOS, offers several specificities :

Some papers about DOSMOS.

This work is at the core of Laurent Lefèvre's and Olivier Reymann's PhD theses. Laurent Lefèvre is now assistant professor at the Claude Bernard-Lyon I University where it continues developing DOSMOS. For more details...

Finally, more recently, we developed a high performance Java interface for Myrinet-connected PC clusters.

Parallel databases and video servers

Thanks to their scalability and remarkable performance/price ratio, parallel architectures appear as a very interesting hardware solution for the server overload described above. In particular, emerging interconnection technologies like Myrinet or Giga Ethernet, and the co-development of high performance (low latency, large bandwidth) PC clusters and Networks of Workstations (NOWs),open the way to a new type of coarse/middle grain architectures of low cost... if one can develop new data management heuristics adapted to these architectures. We have focused our work on two target applications : very large decision support databases and documentary and multimedia servers. In this framework our work concerned the query optimization and processing and the implementation of collaborative cache management protocols

We thus proposed a novel formalism for representing parallel relational queries, the DPL graphs. DPL graphs offer powerful description mechanisms. In particular, they allow including scheduling and mapping information within parallel execution scenarii. Based on DPL graphs, we exhibited a new class of execution scenario, the serialized bushy trees, and proposed new optimization heuristics that experimentally showed very interesting performances, especially when processing ressources are limited.

We also formalized the concept of parallel extension of data servers. The idea consists in coupling an application server (e.g. a DBMS) with a network of workstations/PCs. This requires : a partial duplication of the data, synchronization mechanisms, a parallel processing engine, communication protocols. In this context, we developed a parallel query optimizer called MPO (Modular Parallel Optimizer) and a parallel extension module for relational DBMS, Enkidu. Experiments were made using Oracle avec base DBMS.

These research have been supported by an Esprit Technological Transfert Node Program and a Région Rhône-Alpes program (Santé & HPC).

Some papers about DPL graphs and serialized bushy trees.

Finally, within the framework of a project supported by France Telecom R & D, the Sésame project, we have recently studied the design and implementation of parallel video servers for News on Demand applications. Our idea is to integrate all available indexing information (also called "annotations") at the heart of the management mechanisms and heuristics : query scheduling, cache replacement policy, video sequence prefetching, video stream management. The proposed heuristics have shown a performance improvement of the order of 15-25% with respect to classical heuristics. Based on these concepts, a video server prototype, called MINA, was designed and implemented.

Some papers about parallel video servers.

 

Modeling Content-Based Queries in Multimedia Database Systems

Modelling Content-Based Query in Multimedia Database Systems The question of integrating multimedia data such as image, video, sound, etc. in to Database Management Systems (DBMS) is these days one of the challenging subjects of research. Since 1999 we are engaged with the work of modelling multimedia data (particularly image data) for a content-based retrieval in an image database system. As a result, we have proposed an image data repository model that supports a combination of similarity-based and relational operations. We then have formalized a similarity-based algebra that can enable us perform similarity-based operations on image representations identical to the operations in relational algebra. The novel operators we defined to develop this similarity-based algebraic system includes: similarity-based selection, similarity-based join, symmetric similarity-based join, mine (or inversed selection), etc. We further have studied the properties of these novel operators and have identified useful properties for similarity-based query optimisation. The work in this domain is still ongoing and we expect interesting results for a Content-Based Multimedia Database System.

Some papers.

 

Collaborative multimedia information services

Collaborative Distributed Multimedia Information Service The emergence of high-speed network now a days initiates the design of collaboration mechanisms between remote site systems that are giving the same service. A condition for this collaboration is to define the protocols for sharing and exchange of relevant data. Our approach and proposal inspires in one hand the services of proxies and on the other hand our previous work on the management of parallel video cash. We thus have recently proposed a logical architecture that guarantee effective collaboration of data server proxies. These proxies index the content of text document while updating the content of "carte des thématiques chaudes" in real time. The analysis of this card permits on one side the optimisation of the management of cash and pre-chargement, and on the other side the optimisation of inter-proxy document exchange. This finally provides the user with a powerful indexing and searching engine.

First experiments, performed in the context of video archives and Web proxies are very encouraging.

Some papers.

 

References

On-line papers

... or that will soon be available on the Web :-) :

Parallel programming environments - Distributed shared memory systems -High performance Java

L. Brunie, L. Lefèvre "New propositions to improve the scalability and the efficiency of distributed shared memory systems", IEEE ICA3PP'96 (International Conference on Algorithms And Applications of Parallel Computing), Singapour, juin 1996, p. 356-364.

L. Brunie, L. Lefèvre, O. Reymann "Execution analysis of DSM applications : a distributed and scalable approach", ACM SPDT'96 (Sigmetrics Symposium on Parallel and Distributed Tools), Philadelphie, mai 96, p. 51-60

L. Brunie, L. Lefèvre, O. Reymann "High Performance Distributed Objects for Cluster Computing", 1st IEEE International Workshop on Cluster Computing (IWCC '99)}, Melbourne, Australie, déc. 99 ; IEEE Computer Society Press, p. 229-236.

L. Brunie, M. Exbrayat, N. Gimmillaro "Combining Myrinet and Java in the context of a query evaluation application : first experiments and results", Myrinet User Group Conference, Lyon, France, septembre 2000.

Parallel query optimization - Parallel extension of relational DBMS

L. Brunie, H. Kosch "Control strategies for complex relational query processing in shared nothing systems", ACM Sigmod Records, vol. 25, n° 3, septembre 1996

L. Brunie, H. Kosch, W. Wohner "From the modeling of parallel relational query processing to the query optimization and simulation", Parallel Processing Letters, vol. 8, n° 1, mai 1998.

M. Exbrayat, L. Brunie, A PC-NOW Based Parallel Extension for a Sequential DBMS, PC-NOW 2000, Cancun, Mexique, mai 2000.

Parallel video servers

Mostefaoui, L. Brunie, H. Kosch, L. Boeszoermenyi, Multi-clip Query Optimization in Video Servers, IEEE International Conference on Multimedia and Expo. ICME'2000, New-York, juillet 2000

L. Brunie, H. Kosch, A. Mostefaoui, Semantic Based Prefetching in News-On-Demand Video Servers, Multimedia Tools and Applications Journal, éditions Kluwer ; accepté, à paraître.

A. Mostefaoui, C. Perez, L. Brunie, Serveur de séquences audiovisuelles parallèle sur réseau haut débit : concepts et expérimentations, Rencontres Francophones du Parallélisme, RENPAR'11, Rennes, juin 99, 127-132.

Collaborative information systems

Lionel Brunie, David Coquil, Serge Simon, Software architectures for collaborative proxies in wide area information systems (position paper), Fourth International Workshop on Parallel and Distributed Databases : innovative applications and new architectures (PaDD'2001), Münich, septembre 2001, IEEE Computer Society Press.

Lionel Brunie, David Coquil, Ahmed Mostefaoui, Semantic optimized heuristics for multimedia databases caching and prefetching, SPIE International Symposium on Voice, Video, and Data Communications Multimedia Networks and Management, Denver, Colorado, Etats-Unis, août 2001.

Modeling multimedia queries - Query by content applications - Image databases

Solomon ATNAFU, Lionel BRUNIE, and Harald KOSCH, Similarity-Based Algebra for Multimedia Databases, In the Proc. the 12th Australian Database Confrence (ADC'01), Gold Cost, Australia, Jan. 29 - Feb. 2, 2001, IEEE Computer Society Press, pp. 115-122.

Solomon ATNAFU, Lionel BRUNIE, and Harald KOSCH, Similarity-Based Operators and Query Optimization for Multimedia Database Systems, International Database Engineering & Applications Symposium (IDEAS'01), July 16-18, 2001, Grenoble, France; IEEE Computer Society Press. (zip format)