PANELS at the ICCCN 2019

Panel 1: IoT, 5G, and Edge AI: Challenges and Opportunities

On July 29, 2019

Description: With a projected worldwide spending reaching 1.2T in 2022, Internet of Things (IoT) has generated remarkable excitement across academia and industry, promising to improve efficiency in transportation, power grids, buildings, retail, manufacturing and agriculture, and create new opportunities in homes, environment, healthcare, and smart cities. The upcoming 5G cellular technologies promise to deliver higher data rates with lower latency, enabling a wide range of data-intensive applications such as remote gaming, telepresence, and VR. 5G is also anticipated to help realize the full potential of IoT, bringing IoT and edge devices even closer to the cloud. At the same time, many applications would benefit from, and sometime require, local computation and intelligence, prompting a new wave of research and excitement on artificial intelligence at the edge (Edge AI). In this panel, we bring together experts from IoT, CPS, and Edge AI to understand the challenges across these three topics, as well as explore opportunities together with the audience.

Tarek Abdelzaher, University of Illinois at Urbana-Champaign,
Suman Banerjee, University of Wisconsin-Madison,
Nirmit V Desai, IBM Research,
Koen Langendoen, Delft University of Technology,
Moderator: Xiaofan (Fred) Jiang (lead),
Columbia University

Panel 2: Machine Learning for Cloud Networks

On July 30, 2019

Description:Performance unpredictability is a major roadblock towards cloud adoption and has performance, cost, and revenue ramifications. Predictable performance is even more critical as cloud services transition from monolithic designs to microservices. As applications get more complex, and levels of hardware and software abstraction increase, the community is increasingly adopting the "collect everything" approach to monitoring. Leveraging Machine Learning techniques on this data is key to isolating root cause of problems. Can Cloud management be unsupervised? Can the metrics and logs point us to potential problems, instead of the current method of an expert sifting through vast amounts of data using domain expertise, intuition and experience. Can we identify if the root cause of performance hotspots is an application, systems or a network problem before they propagate and amplify across dependent services? This panel brings together experts on Cloud monitoring and Machine Learning for distributed systems for an interesting discussion on these issues.

Moderator: Ramya Raghavendra (lead) - IBM Research,
Prof. Theo Benson - Brown U,
Mike Cugini (Site Reliability Engineer, Dropbox),
Prof. George Kesidis (Penn State U),
Dr. Homin Lee (Data Scientist, DataDog),