A scalable and reliable fabric manager for high performance networks (224947)

Operators of data centres and supercomputers use a special kind of software, called a fabric manager, to ensure network reliability and stable operation. Think of it as air traffic control for computer networks. Without a fabric manager, every network device would have to be configured manually and a single fault or change in the configuration would disconnect the network. This would make the computing resources unreliable and unsuitable for running critical applications. Today?s fabric managers try to address these problems, but they are slow in addressing network faults and poor at balancing network traffic to changing workloads. They also impose strict rules on how the physical architecture of the network should be configured. This project will verify a new type of fabric manager, which will dramatically improve fabric management capabilities compared to existing solutions. It addresses several problems of fabric management related to fault tolerance, load balancing and routing. The technical goal is to build and launch a commercial quality fabric manager with novel features for fault tolerance, load balancing, and routing. The commercial goals are to recruit 2-5 pilot customers that can test the product in a production environment, then to find the first paying customers for the product.