Here is the updated scientific production of the HADAS research team (Heterogeneous and Adaptive Distributed dAta management Systems), part of the research area "Data and Knowledge Processing at Large Scale" within the LIG, Grenoble Informatics Laboratory.
Technological changes have reduced the cost of creating, capturing, managing and storing information to a sixth of what it was in 2005. This allowed a scale change in the size and distribution of data, the number of connected devices, and the number of users. Data can be numerical data coming from sensors, scientific data or personal data coming from heterogeneous and largely distributed data sources. They cannot be handled anymore by centralized data management systems with pre-established schemas. Modern data and knowledge management systems requires new data models, new services and algorithms that must be largely distributed and deployed over different types of large scale systems (grids, peer-to-peer networks, sensor networks, web infrastructures). The HADAS group has revisited and extended standard database systems to deal with dynamic and distributed data, to define data management systems as infrastructure comprise of distributed data series or composing data services for handling autonomy, dynamic behavior and heterogeneity of both users and data sources. The activities of the group focuses on:
Management of massive datasets:
-Adaptive and distributed storage and cache for storing large heterogeneous datasets
- Indexing data on the fly to facilitate efficient data manipulation.
- Economy and energy oriented integration of big datasets management: economic cost model.
-Quality-based continuous data/event stream processing and composition.
Adaptive querying systems:
-Declarative hybrid languages for expressing data (streams) processing.
-Learning-based distributed query optimization for efficient (continuous) query evaluation with scarce metadata.
-Query operators for on-the-fly data reorganization facilitating future data manipulations.
-Service Level Agreement guided optimization of continuous and mobile queries.
Of course data technologies have to be deployed on different types of architectures (grids, peer-to-peer networks, sensor networks, cloud, HPC, GPU, ARM/Raspberry).
We collaborate with other laboratories in several ANR projects and also with industry.
Results of our research have direct impact on applications dealing with huge amounts of data and resources largely distributed in pervasive environments, such as data spaces, smart grids and smart buildings, etc.
Keywords:
- Big Data management,
- Query optimization,
- Polygot persistence systems,
- Event-based systems,
- Service-oriented data management systems