Towards A Dynamically Optimal Self-Adjusting Network

The vision of developing self-adjusting networks is now a reality: programmable optical networking switches are already being used in big tech companies such as Google and Microsoft. However, theoretical foundations are lagging behind in this area: previous methods have been designed by having fixed and limited networking resources in mind. In this talk, I detail the mathematical background and explain two solutions for problems inspired by data center networking and packet classification needs. I will show how online and randomized algorithms are being used to develop more efficient techniques that are optimized toward and match the traffic workload they serve.