|Title||:||Data-Driven Techniques Towards Performance Optimization in Wireless Networks|
|Speaker||:||Ayon Chakraborty (State University of New York, Stony Brook, USA.)|
|Details||:||Fri, 24 Mar, 2017 3:00 PM @ BSB 361|
|Abstract:||:||In this talk, we will introduce two emerging issues in wireless
networks in the two extreme ends of the network protocol stack. The
first deals with managing quality-of-experience (QoE) in using mobile
applications and the second deals with large-scale monitoring of the
radio frequency spectrum. Essentially, for both the directions, we
show how data-driven techniques can be leveraged to model performance
of complicated real world systems and apply necessary optimization.
Guaranteeing good QoE for the user is a challenge in mobile apps due to their diverse resource requirements and the resource-constrained, variable nature of wireless networks and mobile devices. A key question we ask is: how to select the best network for a given application, or how to optimize the overall user experience for multiple users and applications in a given network? To address such issues, we develop a network-side system called ExBox (Experience Middlebox) that uses a new experiential capacity-based formulation to aid admission control and network selection. ExBox can handle multiple different types of applications in the network and dynamic network conditions.
The second issue relates to the problem of RF spectrum limitations in the face of exponentiating demands for wireless network capacity. This has promoted new spectrum sharing regimes where a diverse set of wireless technologies, including broadcast television, radar and various forms of wireless communications, co-exist in the same spectrum band while respecting specific spectrum rights. Our key question is: how effectively can we monitor available spectrum opportunities across time and space? One key technology in this space is spectrum databases that store spectrum availability information. We augment existing spectrum database technologies to improve their accuracies in a cost-effective fashion. The idea is to supplement existing model-based techniques with actual spectrum measurements. We also contribute towards making the spectrum measurements themselves scalable by developing techniques to perform spectrum sensing on mobile devices. These efforts culminate building a spectrum database system called SpecSense that can schedule and collect measurements from a distributed system of spectrum sensors in order to estimate spatio-temporal patterns in spectrum availability.
Bio: Ayon Chakraborty is a final year Ph.D. candidate at State University of New York - Stony Brook working with Prof. Samir Das. He is broadly interested in mobile systems and data driven performance analysis/optimization of such systems functioning in constrained environments. Many papers resulting from his research appeared in top venues including ACM SIGCOMM CoNEXT, IMC, MOBICOM and IEE INFOCOM. His work on characterizing quality-of-experience (QoE) among mobile virtual network operators (MVNOs) was nominated for the Best Paper Award at ACM IMC 2014. Before joining Stony Brook, Ayon completed his B.E. from Jadavpur University (Calcutta) in 2011, where he won the Department Gold Medal. Ayon is also a recipient of DAAD scholarship.
Host : Prof. C Siva Ram Murthy