Efficient Distributed Data Access and Cloud Technology

Existing grid technology can be separated into two major fields, the computational grids and the data intensive grids. One of the challenging problems for data grid is the optimization of the remote data access to large datasets. Our works in this field concentrate on the optimization of data access carried out at two levels: a global optimization and a local optimization. In the case of global optimization we study data replication mechanisms. In the second one the optimization of usage of local resources in storage nodes is studied, focusing on the tertiary storage systems. In order to use our optimization methods, a special system architecture should be applied in data grid environments. For that reason research about middleware layer architecture of data grid systems is being done by us also.

The global optimization of data access is achieved by file replication in our researches. In order to have an optimal distribution, with respect to files usage, we need the following optimization services integrated in the middleware layer: efficient replica manager and replica catalog tools. The local optimization of usage of storage node resources is useful, due to different kinds of data and different types of secondary/tertiary storage systems installed locally. The goal of the local optimization study is to provide data stored inside storage nodes in unified and optimal way and to estimate data access-factors needed for global optimization. In this case a new component-expert architecture was used and appropriate data time estimators were implemented.

The research described above is sponsored by many Polish and European grants (SGIgrid, CLUSTERIX, EU CrossGrid Project, Polish-Austrian Grants). It is completed in cooperation with many co-partners:

• CERN
• University of Klagenfurt, Austria
• PCSS Poznañ and Czêstochowa University of Technology
• Crossgrid Consortium

Researchers: D. Nikolow, R. Słota, M. Kuta, J. Kitowski