Grid news
After almost 2 years of background research and the specification of user requirements, ACGT has completed the architecture design of its Grid infrastructure. Put simply, a grid infrastructure is a collection of servers and communication protocols that allow highly complex and compute-intensive tasks to be shared by the computers in the grid in a safe and efficient manner.
Based on standard Grid technologies like Globus, Gridge technologies, the ACGT Grid infrastructure is managed by the Poznan Supercomputer Center in Poland and at the time of publication of this newsletter includes computing nodes in Malaga, Brussels, SanKt Augustin, Crete and Poznan with plans to add more in the near future. In total 5 servers running mostly Linux provide a stable, distributed startup test bed with sufficient computing horsepower to handle a variety of compute-intensive applications such as the National Technical University of Athens? Oncolsimulator tool.
The sensitivity of the personal data being accessed and the life-science applications that are at the core of ACGT together with the computational complexity of some of these, have put very strong security and load distribution efficiency requirements on the infrastructure.
To meet these, ACGT has adopted standards commonly used in the Grid technology area; for example the security element is based on GSI technologies. Security tools are concerned with establishing the identity of users or services (authentication), protecting communications, and determining who is allowed to perform what actions (authorization). All ACGT users and services are obliged to identify themselves using certificates.
GSI technology supports authentication based on X.509 certificates while authorization decision is provided by the Grid Authorization Service ? a higher level tool fully compliant with GSI.
For load distribution, ACGT exploits the grid meta-scheduling system called GRMS (Grid Resource Management System). GRMS which is part of the PSNC Gridge Toolkit, is able to react dynamically to changing states of the Grid environment and balance the workload of the available Grid nodes according to the current load of the processors.