John Tadrous is an assistant professor of electrical and computer engineering at Gonzaga University. He received his Ph.D. degree in electrical engineering from the ECE Department at The Ohio State University in 2014, MSc degree in wireless communications from the Center of Information Technology at Nile University in 2010, and BSc degree from the EE Department at Cairo University in 2008. Between 2014 and 2016 he was a post-doctoral research associate with the ECE Department at Rice University.
His research interests include Modeling and analysis of human behavior's impact on data networks in various timescales from seconds to hours. This research spans interdisciplinary fields of human behavioral science, economics, machine learning, and big data analytics. He also works on the design of optimal resource allocation strategies that harness developed human behavioral models for maximal network welfare. This research utilizes information theory, network optimization and control, and smart-data pricing. Dr. Tadrous' served a technical program committee member for several conferences such as Mohihoc and WiOpt. He also served as a peer reviewer for a multitude of scientific conferences and journals.
Modeling and Analysis of Interactive Data Traffic
The dominant portion of smartphone app traffic involves human interactions with the app. This research project aims at studying and characterizing specific features of wireless data traffic generated by interactive apps and developing short-timescale traffic models that facilitate more efficient service of smartphone traffic. Our preliminary investigations of different interactive app categories (e.g., web-browsing, online gaming, and travel) have revealed several interesting characteristics in the timescale of seconds, these promise of a considerable quality-of-experience (QoE) enhancement for end-users, e.g., reducing operational delays by 50% for every user-server interaction.
We have collected a dataset featuring packet level detail of 1500 sessions of interactive smartphone apps and made it available for online for scientific research. This link provides a guide to accessing and processing such dataset.
Leveraging App Interactivity for Scheduling Efficiency
Building on the outcomes of interactive data traffic models, we work on developing intelligent resource allocation strategies that serve multiple app sessions with the highest QoE possible. These strategies utilize two key properties of app interactivity in the timescale of seconds. Namely, (1) the generation of new user-server interactions is dependent upon the completion of the service of the current interaction, thus there is significant coupling between traffic generation and service, and (2) that end users spend a relatively long time in the order of seconds to process each server's response before they start a new interaction, thus creating much room of service opportunities for other sessions.
Our approach targets both the theoretical and practical design aspects of the scheduling problem. Theoretically, we investigate the optimal design of service strategies that achieve maximum network utility, while practically we give attention to complexity and scalability aspects of design.
Content Management in Next Generation Networks
Offering reliable service of data content is becoming a challenging problem with the emergence of throughput hungry applications such as 4K video streaming, online gaming, and cloud computing that demand significant bandwidth resources. Our research aims at exploiting large timescale behavioral characteristics of end users to best provision data content in a way that maximally utilizes network resources while guaranteeing highest levels of QoE. These characteristics include the high discrepancy between peak and off-peak demand levels, the predictability of content popularity over time, and the economic responsive of end-users.