The ongoing trend to move more of our computations and IT services into the cloud implies that huge server parks to an increasing extent take on the role of crucial infrastructure. An essential part of such installations is the interconnection network serving as a communication backbone for the servers constituting a cloud. At Simula, we have conducted research on routing and management of interconnection networks of supercomputers for more than a decade, and our algorithms are running on some of the biggest supercomputers in the world. Currently we extend our line of research through three main initiatives: ERAC, NumaMultiConnect and Fabriscale. Through these three initiatives, the Cloud Department at Simula aims at staying on the forefront of research and innovation in the intersection between HPC interconnection networks, cloud computing, and big data.
Brief overview of activities
In the ERAC research project, we focus on selected problem areas where current technology is not well suited to meet the promises of cloud computing: rapid elasticity, massive scale, resilient computing, virtualization, ease-of-use, trust and energy efficiency. In particular, we conduct research related to self-adaptive network architectures and flexible and efficient virtualization techniques for high-performance clouds. These areas are of vital importance for realizing the potential of cloud computing and big data, where there are significant architectural and technological challenges that must be resolved.
In the research project NumaMultiConnect we work closely with the industry to develop the next-generation technology for connecting processors from multiple CPU manufactures and form large shared-memory systems enabling in-memory processing of big data. By making efficient use of inexpensive off-the-self components to create large, easy-to-use shared-memory systems, the project aims to make big data analytics available to a broader audience – in the industry as well as in the academic sector.
In the Fabriscale project, the initial goal was to create a commercial fabric manager for InfiniBand based on results from our previous and ongoing research on interconnection networks. After a very successful first phase, Fabriscale Technologies was established. Today this company offers a fabric manager with several unique features compared to its competitors, including extremely fast fault-tolerance, highly scalable and efficient topology agnostic routing, and a flexible SDN interface.