Invited Speakers
Aruna Balasubramanian
Stony Brook University, USAVisit Homepage
Efficient and Sustainable NLPMuch of the recent transformative advances in Natural Language Processing (NLP), including ChatGPT, are driven by advances in language models and deep neural networks. However, these advances have come with staggering computational and energy costs. For example, a state-of-the-art GPT-3 model used in ChatGPT3 has 175 billion parameters and requires significantly more energy to train than the average lifetime fuel consumption of a car. In the first part of the talk, I will describe systems optimizations we have developed that significantly reduce the compute and memory requirement of NLP models. The optimizations we developed can be applied broadly and results in over 10x reduction in latency when deployed on mobile devices. In the second part of the talk, I will describe our recent work on predicting energy consumption of NLP models. Existing energy prediction approaches are not accurate, making it difficult for developers and practitioners to reason about their models in terms of power. We use a multi-level regression approach that produces highly accurate and interpretable energy predictions. Finally, I will describe some future problems in this space and the role of systems and networking in addressing these problems.
Aruna Balasubramanian is an Associate Professor at Stony Brook University (and currently a visiting faculty at SUNY Korea). She received her Ph.D from the University of Massachusetts Amherst, where her dissertation won the UMass outstanding dissertation award and was the Sigcomm dissertation award runner up. She works in the area of networked systems. Her current work consists of three threads: (1) making AI efficient and sustainable, (2) improving usability, accessibility, and privacy of mobile applications, and (3) designing measurement-driven approaches to improve performance of next generation networks. She is the recipient of the SIGMobile Rockstar award, a Ubicomp best paper award, a Computing Innovation Fellowship, a VMWare Early Career award, several Google research awards, and the Applied Networking Research Prize. She is passionate about improving the diversity in Computer Science and leads the diversity committee in the department, is the faculty advisor for the WiCS and WPhD groups at Stony Brook, and is an active member of the N2Women group.
Radha krishna Ganti
IIT Madras, IndiaVisit Homepage
Massive MIMO with increasing number of antennas at the base station is an integral part of 5G and Beyond 5G due to the tremendous spatial diversity and multiplexing gain offered by such a large number of antennas. However, the huge number of data streams generated by it can be cumbersome to process and transfer, especially in the fronthaul link between the Remote Radio Head (RRH) and the Baseband Unit (BBU) of the base station. Matrix representation is an elegant and convenient way to visualize this data and analyze it. Emergent from this analysis, we propose an iterative matrix decomposition technique for fronthaul load reduction in the uplink for massive MIMO systems utilizing the convolution structure of the received signals. In this talk, we analyze the performance of the method under different practical scenarios and constraints. We also explore an almost blind demodulation method for MIMO-OFDM signals that uses this iterative technique to provide estimates of the user data and channel using a single pilot, irrespective of the size of the OFDM signal.
Radha Krishna Ganti is an Associate Professor at the Indian Institute of Technology Madras, Chennai, India. . He received his B. Tech. and M. Tech. in EE from the Indian Institute of Technology, Madras, and a Masters in Applied Mathematics and a Ph.D. in EE from the University of Notre Dame in 2009. His doctoral work focused on the spatial analysis of interference networks using tools from stochastic geometry. He received the 2014 IEEE Stephen O. Rice Prize, and the 2014 IEEE Leonard G. Abraham Prize and the 2015 IEEE Communications society young author best paper award. He was also awarded the 2016-2017 Institute Research and Development Award (IRDA) by IIT Madras. In 2019, he was awarded the TSDSI fellow for technical excellence in standardisation activities and contribution to LMLC use case in ITU.
Aniruddha Gokhale (Dinesh)
VanderbiltVisit Homepage
5G and beyond cellular networks are promising technologies for Industrial Internet of Things (IIoT) because of their support for high bandwidth, low latency, reliability and differentiated services that are provided through a virtualization technique called network slicing. However, the dynamic nature of IIoT traffic and their differentiated QoS requirements makes deploying multiple network slices on the same shared physical infrastructure a challenging problem. To address this problem, this paper presents an approach based on a digital twin of the 5G network that is used in predicting network traffic to improve network resource utilization, reliability, and efficiency. Specifically, the digital twin combines Holt smoothing and time series transformer models to predict end-to-end 5G network traffic over multiple time steps. Empirical evaluations comparing the performance of different machine learning algorithms used by the digital twin on two datasets: LTE 4G network traffic (497, 545 instances with 4 features) and Unicauca-Version2-87Atts (3, 577, 296 instances with 87 features) reveal that the combination of Holt smoothing and time series transformer models yields the best results. The prediction results from our approach are then used in deploying elastic network slices.
Dr. Aniruddha S. Gokhale is a Full Professor of Computer Science and Engineering in the Dept of Computer Science (primary) and Dept of Electrical and Computer Engineering (secondary), and a Senior Research Scientist at the Institute for Software Integrated Systems (ISIS) all at Vanderbilt University, Nashville, TN, USA. His current research focuses on developing novel resource management algorithms using data- and model-driven techniques to address emerging challenges in the areas of edge-to-cloud computing, 5G and beyond systems, real-time stream processing, and publish/subscribe systems as applied to cyber physical systems, such as smart transportation and smart cities. He is also working on using cloud computing technologies for STEM education. Dr. Gokhale obtained his B.E (Computer Engineering) from University of Pune, India, 1989; MS (Computer Science) from Arizona State University, 1992; and D.Sc (Computer Science) from Washington University in St. Louis, 1998. Prior to joining Vanderbilt, Dr. Gokhale was a member of technical staff at Lucent Bell Laboratories, NJ. Dr. Gokhale is a Senior member of both IEEE and ACM, and a member of USENIX. His research has been funded over the years by DARPA, DoD, industry and NSF including the NSF CAREER award in 2009.
Sergey Gorinsky
IMDEA Networks Institute, SpainVisit Homepage
Internet users spend an increasing portion of their day on video streaming, which has become a runaway traffic leader among Internet applications. Quality of Experience (QoE) represents the overall satisfaction of the user with the streaming service. Because QoE is a complex subjective notion, measurement and modeling of QoE poses difficult challenges. Prior work produces a large number of diverse QoE models for video streaming. Moreover, QoE models play a prominent role in design and evaluation of streaming systems. In this paper, we focus on the QoE problem and, in particular, construction and usage of QoE models. Based on personal research experience and real-world datasets, we analyze the state of the art in the area from accuracy, complexity, interpretability, and other angles. The analysis identifies common pitfalls and provides practical guidance on how to build QoE models and utilize them in design and evaluation.
Sergey Gorinsky is a tenured Research Associate Professor and leads the NetEcon (Network Economics) Group at IMDEA Networks Institute in Madrid, Spain. He joined the institute in 2009 and was a Ramón y Cajal Fellow funded by the Government of Spain between 2010 and 2014. From 2003 to 2009, Dr. Gorinsky served on the tenure-track faculty at Washington University in St. Louis, USA. Sergey Gorinsky received his Ph.D. and M.S. degrees from the University of Texas at Austin, USA in 2003 and 1999, respectively, and Engineer degree from Moscow Institute of Electronic Technology, Zelenograd, Russia in 1994. He graduated four Ph.D. students and has primary research interests in computer networking, distributed systems, and network economics. Sergey Gorinsky made research contributions to real-time scheduling, deployment of content delivery networks, economics of Internet interconnections, buffer sizing, learning-based caching, service differentiation, multicast, congestion control, networking education, data-plane algorithms, routing, video streaming, and bulk-data transfer. His work appeared at top conferences and journals such as SIGCOMM, NSDI, CoNEXT, INFOCOM, Transactions on Networking, and Journal on Selected Areas in Communications. He served as a TPC chair of ICNP 2017 and other conferences, as well as a TPC member for a much broader conference population including NSDI (2024), SIGCOMM (2012, 2016, 2022), CoNEXT (2015-2017, 2019, 2021, 2023), INFOCOM (2006-2019, 2021-2024; area chair: 2019), and ICNP (2008, 2010-2017, 2019-2023; area chair: 2013, 2016, 2019-2023). Sergey Gorinsky won the INFOCOM Distinguished TPC Member Award for the record seven times (2015-2018, 2021-2023). Dr. Gorinsky contributed to conference organization in many roles, such as a general chair of SIGCOMM 2018 and ICNP 2020. He also served as an evaluator of research proposals and projects for the European Research Council, European Commission, COST Association, Swiss National Science Foundation, Israel Science Foundation, United States National Science Foundation, ITRA-Mobile, and numerous other funding agencies.
Praveen Jayachandran
IBM ResearchVisit Homepage
Multi-cloud computing is rapidly becoming the norm, with enterprises deploying applications across an increasingly large number of public, private and edge clouds. This talk will highlight the trends in this space and the challenges it poses on networking and observability across these geo-distributed cloud environments. Specifically, the talk will explore challenges and approaches in supporting “application-aware” network connectivity across heterogeneous cloud environments, leveraging Observability and AI-based operational analytics. We will also discuss how insights derived from analytics can assist intelligent automation for managing applications communicating across heterogeneous multi-cloud environments.
Praveen Jayachandran is a senior technical staff member and senior manager of the Hybrid Cloud operations department at IBM Research, India. His work spans network management, observability, and managing systems and data at scale, specifically for multi-cloud and Edge environments. He is an IBM Master Inventor, a member of the IBM Academy of Technology, and a senior member of IEEE. He holds a PhD from the University of Illinois at Urbana-Champaign, USA.
Markku Juntti
University of Oulu, FinlandVisit Homepage
We discuss and summarize the architectural options for radar type sensing in cellular networks to enable integrated sensing and communications (ISAC) operations. The key characteristics of monostatic and multistatic radar operation are briefly reviewed and the design challenges for cellular ISAC are discussed. We focus then more closely to the design problem of downlink transmit beamforming at a base station serving multiple users and performing simultaneously monostatic target sensing. The waveform design principles are introduced for single and multiple targets. The implications of wideband multicarrier operation are described and the performance optimization in this regime is discussed. Both optimization-based algorithms and deep unfolding based data driven JCAS design are introduced. Finally, we discuss the timely research challenges in the area.
Markku Juntti (Fellow, IEEE) received his his Dr.Sc. (EE) degree from University of Oulu, Oulu, Finland in 1997. Dr. Juntti was with University of Oulu in 1992–98. In academic year 1994–95, he was a Visiting Scholar at Rice University, Houston, Texas. In 1999–2000, he was a Senior Specialist with Nokia Networks. Dr. Juntti has been a professor of communications engineering since 2000 at the University of Oulu, Centre for Wireless Communications (CWC). He serves as the Leader of CWC – Radio Technologies (RT) Research Unit. His research interests include signal processing for wireless networks as well as communication and information theory. His research interests include signal processing for wireless networks as well as communication and information theory. Dr. Juntti is also an Adjunct Professor at Rice University, Houston, Texas.
Igor Kotenko
SPC RAS, RussiaVisit Homepage
Modern Internet of Things networks combine many devices and sensors that transmit and process large amounts of data. Security tools identify security events that contain information about detected system or network states. In turn, high-performance data anomaly detection methods are required to ensure stability and reliability of work processes. Information about the correlation of identified security events can be used to detect and explain deviations from normal states. This study proposes an anomaly detection approach based on the casual correlation of security events using machine learning. The proposed approach does not require prior knowledge of event scenarios. Using cluster analysis and a recurrent neural network, we construct a security state correlation graph corresponding to the normal behavior of the system. Cluster analysis determines the similarity of events to each other. A recurrent neural network, represented by an LSTM, analyzes the temporal relationship of events. Using the identified event correlation thresholds, we look for anomalies in real time. Experimental results on an Internet of Things sensor dataset show that the proposed method is efficient in anomaly detection tasks.
Igor Kotenko is a Chief Scientist and Head of Research Laboratory of Computer Security Problems of the St. Petersburg Federal Research Center of the Russian Academy of Sciences. He is also Professor of ITMO University, St. Petersburg, Russia, and Bonch-Bruevich Saint-Petersburg State University of Telecommunications. He is the Honored Scientist of the Russian Federation, IEEE Senior member, member of many Editorial Boards of Russian and International Journals, and the author of more than 800 refereed publications, including 25 books and monographs. Main research results are in artificial intelligence, telecommunication, cyber security, including network intrusion detection, modeling and simulation of network attacks, vulnerability assessment, security information and event management, verification and validation of security policy. Igor Kotenko was a project leader in the research projects from the European Office of Aerospace Research and Development, EU FP7 and FP6 Projects, HP, Intel, F-Secure, Huawei, etc. The research results of Igor Kotenko were tested and implemented in multitude of Russian research and development projects, including grants of Russian Science Foundation, Russian Foundation of Basic Research and multitude of State contracts. He has been a keynote and invited speaker on multitude of international conferences and workshops, as well as chaired many international conferences.
Vinayak Naik
BITS Goa, IndiaVisit Homepage
Demand for energy in buildings is growing exponentially, and cooling systems contribute to more than 50% of buildings’ energy consumption. With global warming, we need to save energy on the cooling systems. This paper targets spaces where multiple ductless-split cooling systems are deployed, commonly known as split ACs. Unlike ducted centralized cooling systems, they do not have central sensing and control. To optimize energy consumption by the ductless-split cooling system, we propose a Model-assisted Optimal Control (MaOC) algorithm that observes the thermal environment of the room and generates an optimal schedule of execution for the cooling system. We observe that the mathematical model developed for cooling systems follows the properties of the convex function. We define a MINMAX problem to minimize energy consumption and maximize efficiency. We use the statistical distribution of cooling systems’ efficiency to generate a long-term control trajectory. We deploy and evaluate MaOC for ductless-split cooling systems in a real-world environment. We compare it with solutions based on Greedy Algorithms and Reinforcement Learning.
Vinayak Naik is an OPERA-awardee Professor of Computer Science at BITS Pilani, Goa. He is also a faculty at APPCAIR, a center of excellence in Artificial Intelligence at BITS Pilani. He received his Ph.D. in CSE from Ohio State University in 2006 and BE in CS from VJTI, Mumbai, in 1999. He is a member of the Research Advisory Council of D Y Patil International University, Akrudi, Pune. Before joining BITS Pilani, Vinayak was with IIIT Delhi, IISc, and UCLA. He was a Research Advisor to the CTO of TCS and a member of the Governing Council of NIDHI Technology Business Incubator (TBI) at the Indian Institute of Public Health Gandhinagar. As a visitor, he spent time as an Associate Professor at IIT Bombay, a Researcher at Microsoft Research, and a Researcher at VMWare R&D. He was felicitated with the Gandhian Young Technological Innovation Award'16 for using mobile and sensors for healthcare.
Gomathi Ramachandran
AWS, USAVisit Homepage
Traditionally data-plane measurements have been used to understand application performance and to detect specific impairments with high confidence. Control plane effects on data-plane performance were incidental findings, especially in traditional IP networks where highly multiplexed streams were serviced by higher speed, highly protected, optical circuits. As we move to larger and more dynamic loads and to SDN networks where the routing logic is moved further away from the data-plane, it becomes more important to understand the connection between control-plane effects on data-plane performance in addition to forwarding plane impairments. We illustrate this connection with examples of how some common control plane changes affect data-plane metrics.
GOMATHI RAMACHANDRAN is a Senior Research Scientist at Amazon Web Services. Her major area of interest is network performance assessment, including the design of measurements and systems to manage, track, and analyze performance. Prior to her career at AWS, she was a principal technical staff member at AT&T Laboratories. She did her postdoctoral work in modeling DNA dynamics at the Courant Institute in New York, and before that received a Ph.D. in physical chemistry from Cornell University, Ithaca, New York, for work on quasi-classical ion-molecule dynamics. She holds a Master’s degree in chemistry from the Indian Institute of Technology (IIT), Bombay, and a Bachelor of Science degree from Bangalore University, India.
Sushmita Ruj
UNSW, Sydney, AustraliaVisit Homepage
Cloud service providers can act maliciously and tamper with the data. Proofs of storage were introduced to enable clients to verify data integrity. In the first part of the talk, we will present the concept of proofs of storage and present some constructions. We will then discuss limitations of existing proofs of storage and show how blockchains can address these limitations. In the last part, we will present applications of proofs of storage for data sharing, trading and improving blockchain scalability.
Sushmita Ruj is Faculty of Engineering Lead of UNSW Institute for Cybersecurity, IFCYBER and Senior Lecturer in the School of Computer Science and Engineering at UNSW, Sydney. Her research interests are in applied cryptography, post quantum cryptography, blockchains and privacy enhancing technologies. She designs practical, efficient, and provably secure protocols that can be deployed in real-world applications. She has won several competitive grants like Samsung GRO Award, NetApp Faculty Fellowship, Cisco Academic Grant. She is an Associate Editor of the Transactions on Information Forensics and Security. Before joining UNSW, she was a Senior Research Scientist at CSIRO's Data61, an Associate Professor at Indian Statistical Institute and an Assistant Professor at Indian Institute of Technology, IIT, Indore. Sushmita is a senior member of both ACM and IEEE.
Praveen Tammana
IIT Hyderabad, IndiaVisit Homepage
High-speed programmable data planes (e.g., P4 switch, smartNICs) provide exciting opportunities to realize fast, accurate, data-driven network management systems. The core of these systems has in-network packet-processing algorithms that continuously monitor traffic, maintain statistics of dynamic network events (e.g., congestion, failures), analyze these statistics, and respond automatically. Despite their network performance benefits, automatic response to network events leads to an increase in potential sources of adversarial inputs and, hence, an increase in attack surface. In this talk, I will present possible DoS-like attacks and their impact on data-driven network management systems built on top of programmable data planes. This is followed by our ongoing work to detect and defend against such attacks.
Praveen is an Assistant Professor in the Computer Science Department, at IIT-Hyderabad. Before IITH, he was a postdoctoral researcher at Princeton University, USA. He received his PhD from the University of Edinburgh in 2018. He broadly works at the intersection of Systems, Networks, and Security. His current focus is on designing and building networked systems that make networks easy to manage, secure, and robust, by using exciting emerging technologies such as Software-Defined Networking and Programmable Data Planes. Praveen has received the best paper award at ACM SIGCOMM SOSR, IBM academic award, TiHAN faculty fellowship award, and IITH teaching excellence award.
