InfiniCortex and InfiniCloud : a distributed way to exascale

Yves Poppe
SpeakerYves Poppe
TrackLightning Talks Europaea Room
DescriptionResearch and Education Networking has a major new networking alternative to consider: long distance Remote Direct Memory Access (RDMA) enabled by long distance InfiniBand as a software defined network of choice for High Performance Computing, Big Cloud Computing and Storage, data centre replication and recovery.
Commencing in June 2014, the A*STAR Computational Resource Centre (A*CRC) team in Singapore, together with SingaREN and  dozens of partners world-wide, have been building the InfiniCortex based on the integration of four concepts: high bandwidth intercontinental connectivity, InfiniBand range extension technology, connection of InfiniBand subnets with different topologies and running workflows on such a distributed computational infrastructure.  
We have successfully demonstrated InfniCortex prototypes at the SC14 and SC15 conferences. The infrastructure comprised of computing resources residing at multiple locations in Singapore, Japan, Australia, USA, Canada, France and Poland. Various concurrent applications, including workflows, I/O heavy applications real-time interactive applications, and in-situ real-time visualisations were demonstrated.
The conclusive results have  led us to set up an InfiniBand network connecting the major universities and research centers as well as industrial partners to our new petaflop level NSCC (National SuperComputing Centre) as well as setting up an InfiniBand PoP on the US West Coast with internet2 and one in the UK with Géant.  
In our presentation we briefly report on basic ideas behind the InfiniCortex construct, our recent successes and some ideas about further growth and extension of this project.

All talks