Paul van Gerven
28 May 2019

Researchers of Eindhoven University of Technology (TUE) have devised a new optical switch for data centers that can halve the costs and power consumption and reduce network latency.

Power consumption by data centers is increasing at an alarming rate. Although the data is already transported via energy-efficient optical fibers, most of the interconnecting switches are still electrical, which is less energy efficient and causes high latency. Optical switches are the obvious answer, but so far, lack of optical data buffering and the low throughput of direct connections between optical switches, have prevented their widespread adoption.

TUE PhD student Fulong Yan tackled this problem by buffering data packets on an electrical switch. Together with his colleagues, he demonstrated an 8-port prototype of this new switch. They also designed and simulated a 16-port version, which is expected to be fully operational by 2022.

TUE optical switch
The 8-port prototype optical switch developed by Fulong Yan. Credit: Eindhoven University of Technology

Furthermore, Yan came up with a fast flow control protocol that communicates between the optical and electrical switch. To solve the problem of low efficiency, he created a network architecture that avoids direct connections between optical switches. The new architecture, coined OPSquare, consists of two parallel networks interconnected by two kinds of optical switches, resulting in much lower latency. To jack up the overall network performance even further, Yan developed a load balancing algorithm.


Device lifecycle management for fleets of IoT devices

Microchip gives insight on device management, what exactly is it, how to implement it and how to roll over the device management during the roll out phase when the products are in the field. Read more. .

Additionally, Yan devised a network architecture that halves the number of optical transceivers, without losing any network performance. The architecture is scalable: 163,840 servers can be connected by adding a parallel layer in the architecture that involves the use of three parallel sub-networks.

Fulong Yan successfully defended his PhD thesis in April. His research was supervised by Nicola Calabretta and Ton Koonen.