Network Architecture Lab

We work on fundamental questions regarding the design and construction of practical and dependable network systems. Our core interest is the functionality deployed inside the network infrastructure, e.g., inside Internet routers, data-center switches, or intermediate nodes in a sensor network. We aim to determine what functionality should be deployed inside vs. on top of these network elements—when should the network support sophisticated packet processing vs. being merely a pipe for packets? We also work on how this functionality should be implemented, e.g., should the network process packets on customized hardware or on general-purpose processors? In short, our research goal is to identify and build the right network layer for different communication architectures. 

Here are some examples of specific problems and questions we address in our work:

• Trustworthy network debugging: what is the minimum functionality that we need to deploy inside a network, such that we can ask the network how much loss and delay it introduced in a certain traffic flow and get accurate, trustworthy responses?
• The practical boundaries of network tomography: what is the most one can infer about a real network based on end-to-end measurements?
• Fast, programmable packet-processing platforms: how should we program and configure a software router so as to optimize its performance and prevent different packet-processing applications from stepping on each other's toes?
• Dependable sensor networks: what is the minimum amount of mechanism required to support dependable data collection in a sensor network, i.e., ensure that all messages reach the sink with very high probability?