Dr. David Carrera seminar

Date and location: June 25th – 3pm – Room B

Title: Data Centric activities at the Barcelona Supercomputing Center

Abstract: In the recent years, the major change observed with respect to Supercomputing technologies over the last years comes from the appearance of new hierarchical storage systems based on the use of persistent memory technologies, and the adoption of new fast network fabrics. At the same time, the current race towards exa-scale systems poses new challenges in terms of energy consumption and programmability of the next generation supercomputers. To simplify the management of these every-time more complex infrastructures, a common solution is the use of Software Defined Environments, that is, the development and deployment of centralised controllers that continuously observe the infrastructure and provide virtual partitions that are managed according to user-defined goals. As a result of these changes in workloads and technologies, supercomputing infrastructures are in the middle of a clear transition: technology is moving fast from an era in which floating point capacity was the unique performance metric, to an era in which data-centric workloads are becoming the norm, and the use of virtualization, new programming models are growing to boost productivity and flexibility of supercomputing facilities, and energy consumption is a clear concern. Additionally, it can also be observed a growing demand for more distributed storage technologies, programming models and languages focused on productivity, more scalable and flexible network fabrics, and heterogeneous computing environments.
In this talk I will motivate the reasons behind the current convergence between Supercomputing and Big Data environments, describe the requirements of modern data-centric workloads and cover several activities being carried on by the Data-Centric Computing group at the Barcelona Supercomputing Center aiming to address the complexity of management of emerging supercomputing technologies.