CT Scan for Your Network: Topology Inference from End-to-end Measurements
# 175
Abstract
“Network topology information is needed for a number of management functions at network layer and above. However, the traditional ways of obtaining this information require either access to the control plane (e.g., SNMP, OpenFlow) or cooperation from internal nodes (e.g., traceroute). In this talk, I will give an overview of an inference-based technique for obtaining topology information based on end-to-end measurements in the data plane, known as network topology inference. As a branch of a broader family of inversion problems known as network tomography, topology inference tries to jointly infer the routing topology and the link metrics of a blackbox communication network from end-to-end performance measurements (e.g., delays, losses). The talk will cover the basic idea in topology inference, the state of the art and limitations, and an example of how the inferred information can facilitate upper-layer applications in the context of overlay networks.”
Ting He is an Associate Professor in the School of Electrical Engineering and Computer Science at the Pennsylvania State University. She received the Ph.D. degree in electrical and computer engineering from Cornell University. Her interests reside at the intersection of computer networking, performance evaluation, and machine learning. Dr. He has served as Associate Editor for IEEE Transactions on Communications and IEEE/ACM Transactions on Networking, General Co-Chair of IEEE RTCSA, TPC Co-Chair of IEEE ICCCN, and Area TPC Chair of IEEE INFOCOM. She received multiple top contributor awards from IBM and ITA, and multiple paper awards from IEEE Communications Society, ICDCS, SIGMETRICS, and ICASSP.