Position: Postdoctoral Researcher
Current Institution: Microsoft Research
Abstract: Towards Self-Managing and Performant Networks
Today, the vast majority of the applications we use on a day-to-day basis operate by relying on computer networks. For this reason, any performance degradation on these networks can result in significant impact on not just user experience but the revenues of the service providers (e.g., Google Azure AWS) and the services themselves. My research focuses on improving user quality of experience (QoE) in data center and wide area networks. To this end, I have looked at many different aspects of networking and systems research, including Traffic Engineering diagnosis resilience to failure and security. In my future research, I plan to work on creating networks that can manage themselves. The advent of new machine learning techniques as well as new programmable data-planes makes the time ripe for starting to move our networks to a place where they can manage themselves. In theory, a self-managed network would identify where and when to execute management tasks, would identify problems when and where they occur, and would automatically resolve them. This includes reconfiguring the network when necessary, predicting performance problems and mitigating them, and, finally, finding link failures security breaches and even congestion events. In my current work as a Post-Doc, I am working on problems that lay a foundation that would be useful to solve this over-arching problem: including work on diagnosis where my intern Robert MacDavid (Princeton) and I are looking at improving ticket resolution in Microsoft data centers and also in the space of security where we are looking at finding compromised VMs without visibility into customer VMs.