Very large scale high performance computing and instrument management for high availability systems through the use of virtualization at the Square Kilometre Array (SKA) telescope

Abstract

The Square Kilometer Array (SKA) Telescope, is an ongoing project set to start its building phase in 2018 and be ready for first light in 2020. The first part of the project, the SKA1 will be comprised of 130.000 low frequency antennas (50 MHz to 350 MHz) and 200 mid frequency antennas (350 MHz to 15.5 GHz). The SKA1 will produce a raw data rate of ~10 Tb/s, require a computing power of 100 Pflop/s and an archiving capacity of hundreds of PB/year. The next phase of the project, the SKA2, is going to increase the number of both low and mid antennas by a factor of 10 and increase the computing requirements accordingly. The key requirements for the project are a very demanding availability of 99.9%, computing scalability and result reproducibility. We propose an approach to enforce these requirements - with an optimal use of resources - by using highly distributed computing and virtualization technologies.