Dataflow Computing in the Public Cloud
Public cloud computing has been taking the world by storm. A public cloud appeals to both senior management and to IT departments of many organizations from a wide range of domains. For senior management, a public cloud creates a second source for IT services, removing the lock-in with large internal IT teams and steadily growing infrastructure costs. For the new generation of IT professionals, the public cloud offers the power to manage a massive amount of resources with reduced effort and lower required expertise. Cloud providers are now becoming serious about high performance computing services: 2016 has seen the integration of Altera into Intel, on the back of a large deployment of Altera products at Microsoft; SuperVessel of OpenPower by IBM using Xilinx devices and more recently Amazon announcing their new AWS EC2 F1 instance as their top end performance offering. AWS F1 instances are fully compatible with Maxeler's MAX5 Dataflow Engines (DFEs) generation and come in single and eight-way configurations similar to the MPC-C appliances (one DFE) and the 8-way MPC-X DFE appliance in 1U form factor.
The above provides organizations with the ultimate elasticity to blur the boundaries between public and private clouds in case of hybrid, large-scale deployments. This is of special interest for large organizations in Finance and Data Business Analytics where confidentiality, service level guarantees, and critical business continuity services are of ultimate importance. The recently announced partnerships of Maxeler with Amazon Web Services and Hitachi Data Systems emphasize just two examples of the competitive edge our DFE technology brings on the cloud while providing solutions for extremely hard problems, e.g., computing minimal capital requirements using standard models (FRTB and Basel III) and high-frequency, transaction-related record keeping (Dodd-Frank and MiFID II). In this talk we will show how our Multiscale Dataflow Computing solutions leverage the Amazon AWS F1 instance and provide an elastic hybrid cloud solution to minimize both operational and financial, cost and risk.
Georgi Gaydadjiev is a computer engineer with close to 30 years in Reasearch and Development of various computing systems. He is the Director of Maxeler Laboratories and is a Visiting Professor at the Department of Computing at Imperial College. In the past he held different engineering positions at System Engineering Ltd in Pravetz, Bulgaria and Pijnenburg Microelectronics and Software and later CPS BV in Vught in the Netherlands. He was a faculty member at the Computer Engineering Department at TU Delft between 2002 and 2014. Between 2011 and 2015 he held a Chair Professorship in Computer Systems Engineering at the Computer Science and Engineering Department at Chalmers University in Sweden. In the fall of 2014, he joined Maxeler Technologies in London for a one year sabbatical that transformed in a more longer term involvement. His research expertise include, among others, reconfigurable computing, advanced computer architectures, embedded systems design and fault tolerant systems.