IEEE 7th World Forum on Internet of Things
14 June–31 July 2021 // New Orleans, Louisiana, USA

TOP4: Data and Data Engineering

Date: Tuesday, 7 April 2020
Time: 8:30am-6:30pm
Room: Commerce Room

Data and Data Engineering Track Description

IoT is no longer the “next big thing” – it’s here, real, and transforming the world. McKinsey estimates that 127 devices are connected to the Internet every second; Statista expects 75B devices to be connected by 2025.  Yet the adoption of IoT technologies by the industry is moving slower than expected. While this can be attributed to multiple factors, the capability (or lack of) to cope with “data deluge” – how to store and manage data efficiently as well as to make sense of them – stands as a key one.

To support efficient storage and management of the vast data collected from IoT devices, the market has created solutions on two fronts: cloud infrastructure and embedded storage in edge devices. Even though, challenges still linger as for how to best distribute and process the data in the spectrum of varying computing power, storage cost, power consumption, and communication latency in the IoT networks. This has to be optimized to meet the application requirements and system constraints.

Data provides no value unless we transform it into actionable insights. To that end, IoT practitioners have been utilizing data analytics tools to unlock the value hidden inside the seemingly chaotic data vaults.  And let’s not lose sight of the ultimate goal of IoT – to help make better decisions, whether it’s improving user experiences, optimizing manufacturing, or personalizing marketing campaigns. The advance in data analytics, including AI, has produced early encouraging results.

In this track, we will focus on the challenges and opportunities surrounding, but not limited to, the following themes:

  • Data Analytics & Applications

    • Advances in predictive and prescriptive analytics, including AI and ML methodologies

    • IoT applications

  • Data Management

    • Data quality and data trustworthiness (Provenance)

    • Data semantics interoperability, information sharing

  • Data Storage, Computing, and Networks

    • Storage and computing for IoT data analytics

    • IoT network management and diagnostics


Tuesday April 7th, 2020 – Commerce Room

8:30am-10:30am Session 1: Data Analytics, Storage, and Computing

Tlk1: Mark Cusack (Teradata)

“Putting Machine Learning to Work to Deliver Business Outcomes”

Talk2: Yaniv Iarovici (Western Digital),

“Storage Can No Longer Be an Afterthought in the Industrial Internet of Things”

Talk3: Jun Deguchi (Kioxia)

“Co-optimization of Hardware and Algorithm for Energy-efficient Edge Computing with Convolutional Neural Networks”

Panel Discussion: Moderated by Chung-Min Chen

10:30am-11:00am Networking Break

11:00am-01:00pm Plenary Session

01:00pm-02:00pm Lunch

02:00pm-04:00pm Session 2: Data Management and Quality

Talk 4: Martin Bauer (NEC Labs),

“Sharing IoT Information and Making it Available across IoT Platforms”

Talk 5: Khalil Drira (CNRS),

“Semantic Interoperability for IoT Service Platforms”

Talk 6: Tim Shen (Splunk)

“Big Data System Acceptance Validation Framework(aka. BAVF)

Panel Discussion: Moderated by Chonggang Wang

04:00pm-04:30pm Networking Break

04:30pm-06:30pm Session 3: IoT Networks and Applications

Talk 7: Bin Xie (InfoBeyond),

“DiagSoftfailure: Soft-failure Diagnosis for High-throughput IoT Network by Learning”

Talk 8: Tao Zhang (NIST),

Title to be provided

Talk 9: Speaker TBD

Title to be provided

Panel Discussion and Wrap-up: Moderated by Munir Cochinwala

Track Co-Chairs

Chung-Min Chen, Lutron Electronics

Chung-Min Chen is currently with Lutron where he provides data science consulting in various business areas.  Prior to that he was VP of Data Science at Sidecar Interactive, a startup leveraging machine learning in digital marketing, and Sr. Director of Data Science and Analytics at iconectiv (an Ericsson company) where he helped build Ericsson’s analytics products that enable telecom carriers to better understand and manage mobile user experiences.  He had also held research and management positions with Bellcore, Telcordia, and subsequently Applied Communication Sciences.

His interests span across database systems, distributed and parallel computing, mobile computing, and machine learning. He has published over 50 papers in leading ACM and IEEE journals and conferences including JACM, TOSN, TKDE, TMC, TON, JSAC, SIGKDD, PODS, SIGMOD.  He received a PhD in CS from University of Maryland, College Park and BS in CSIE from National Taiwan University.


Chonggang Wang, Principal Engineer, InterDigital Communications, USA


Chonggang Wang received his Ph.D. degree from Beijing University of Posts and Telecommunications (BUPT) in 2002. He is currently a principal engineer at InterDigital Communications, Inc. where his research interests span quantum internet, edge computing, and Internet of Things (IoT). He was the founding Editor-in-Chief of IEEE Internet of Things Journal (2014-2016). He is a Fellow of the IEEE for his contributions to IoT enabling technologies. He is also a member of Convida Wireless, which is a joint venture partnership between InterDigital and Sony focused on IoT research and development.



Munir Cochinwala, CEO and founder of Data-Khilari

Munir Cochinwala is CEO and founder of Data-Khilari, a small company focused om advanced solutions applying data analytics and AI techniques. He was previously CTO for NJIT, College of Computing as well as Executive Director and Chief Scientist for data analytics at New Jersey Innovation Institute.  He has also at Applied Communication Sciences (ACS), previously known as Telcordia and Bellcore, where he was an AVP and Chief Scientist/Executive Director for the Information and Computer Sciences Research Lab.

He has over twenty-five years of experience and has interest and expertise on all aspects of computer sciences with a focus on databases, data analytics, distributed and mission-critical systems.


Track Speakers

Mark Cusack, Teradata

Mark Cusack leads the Teradata Product Management team responsible for the industry leading analytics platform – Vantage. Previously, Vice President Analytical Ecosystem at Teradata with product management responsibility for data integration products supporting the analytics platform. Prior to that, Chief Architect for IoT analytics at Teradata and Chief Architect and founding developer at RainStor. Once a PhD physicist and researcher in parallel simulation. Mark has a PhD Computational Physics; MSc Computing and BSc Physics.


Talk Title: Putting Machine Learning to Work to Deliver Business Outcomes

Abstract: Advanced analytics are being adopted rapidly across industries. In the past, business leaders
relied on historical analysis of data to understand trends to help drive business decisions – so
called descriptive analytics. Today, leaders are looking towards new predictive and prescriptive
analytics techniques, founded on Machine Learning, to inform and recommend actions in areas
as diverse as supply chain management, industrial IoT and customer experience.
However, the challenge faced by businesses is how to turn what are often “science
experiments” in ML into production-ready applications that can drive near-real-time decision
making. Enterprises face difficulties on the road to operationalizing models: in acquiring the
large quantity of data needed to drive model development and training; in preparing the data
for training and scoring; in model management; and in making the output of the models
In this session, we’ll discuss the challenges faced in operationalizing advanced analytics, and
consider how the largest enterprises are addressing these on the path to putting Machine
Learning to work to deliver outcomes for their business.


Tim Shen, Splunk

Libin Shen, graduated from Harbin Institute of Technology with a major in computer software, has more than 14 years of experience in quality assurance of large software and big data products.  He has in-depth research in big data system quality assurance, distributed system testing, automated testing and etc.




Talk Title: Big Data System Acceptance Validation Framework(aka. BAVF)

Abstract: When we have a big data system being delivered, we are often confronting a lot of questions before really putting it in production mode. For example, how can we make sure it has a good quality, how can I trust it and it will not lose my data from PB level input sources, and etc. Here we define a framework called Big Data System Acceptance Validation Framework(aka. BAVF) to measure a big data system’s acceptance degree from volume, velocity, variety and veracity perspectives. It’s a big data system architecture independent validation framework. We have 4 categories with dozens of standards to measure the whole system’s quality. A detailed introduction will be given for this framework in this session.


Tao Zhang, NIST

Dr. Tao Zhang, an IEEE Fellow, has over 30 years of experience directing research, product development, and corporate strategies to create disruptive innovations and transform them into practical solutions. He is currently with the National Institute of Standards and Technology (NIST). He was the CTO for the Smart Connected Vehicles Business Unit at Cisco Systems, and the Chief Scientist and the Director of R&D on vehicular and wireless networking at Telcordia Technologies (formerly Bellcore, originally part of the Bell Labs). He cofounded the OpenFog Consortium, the Connected Vehicle Trade Association, and served as a founding Board Director for them. Tao holds over 50 US patents and coauthored two books “Vehicle Safety Communications: Protocols, Security, and Privacy” and “IP-Based Next Generation Wireless Networks”, and 70+ peer-reviewed papers. He served as the CIO and a Board Governor of the IEEE Communications Society, a Board Advisor for multiple organizations, an editor for numerous technical journals, and a Distinguished Lecturer of the IEEE Vehicular Technology Society. He cofounded and served on leadership roles for multiple international conferences and forums.



Yaniv Iarovici, Director, Marketing for Industrial IoT and Edge, Western Digital

Yaniv Iarovici is the Director of Marketing for Industrial, IoT and Edge at Western Digital. A 20-year veteran with a diverse background in software, business development and product marketing in the high-tech industry, Yaniv is responsible for developing go-to-market strategies for storage solutions targeting the Industrial and IoT segments. He also oversees the ecosystem development for Mobile, Automotive, Edge Compute and Industrial/IoT markets.  Before joining Western Digital Corporation, Yaniv held engineering and management positions at M-Systems. Prior to that, he was a software engineer at IAI (Israeli Aerospace Industries) and Friendly Robotics. Yaniv holds a bachelor’s degree in Mathematics and Computer Science.

Talk Title: Storage Can No Longer Be an Afterthought in the Industrial Internet of Things

 Abstract: Industry 4.0, the fourth wave of the Industrial Revolution, is ushering in new opportunities for “connected everything,” allowing companies to utilize networked data from sensors, monitors, robots and other IoT devices to take advantage of artificial intelligence (AI) and machine learning (ML) for more efficient factories. Purpose-built solutions with storage, networking and compute – either in sensors, gateways, at the edge, in the cloud or on-premise – need to be architected in ways that are optimized for their part in the IIoT ecosystem. In the past, storage might have been acquired solely based on anticipated capacity needs or price. However, the autonomous applications in today’s Industry 4.0 elicit many more considerations. With the unique needs of IIoT, architects must proactively plan as they design for qualification, or they risk time-to-market advantage. To support IIoT’s promises to bring innovative use cases to light and new efficiencies to operations in this more connected world, you must take storage into considerations early when designing industrial applications. In this session, you will hear about the Industry 4.0 trends that are driving the need to think about storage earlier (hint: 5G, the edge, and M2M play starring roles!) and what unique considerations you must take into account when designing industrial applications.


Martin Bauer, Senior Researcher, NEC Laboratories, Europe GmbH, Germany

Martin Bauer is a Senior Researcher at the IoT research group at the NEC Laboratories Europe (NLE) in Heidelberg, Germany. He received his MSc in Computer and Information Science from the University of Oregon, USA, in 1998, and both his Dipl. Inf. and his doctorate degree from the University of Stuttgart, Germany, in 2000 and 2007 respectively. In 2005 he joined the NEC Laboratories Europe (NLE), where he has been working on context and IoT related research projects as well as standardization activities in the same areas. In particular, Martin Bauer is/has been working on a number of European research projects, most recently in the large scale pilots in IoT, Autopilot and Synchronicity, and in the EU-Japan project Fed4IoT. In standardization, he has actively contributed to OMA NGSI, oneM2M and currently ETSI ISG CIM, primarily on topics related to context management and semantics. Martin Bauer is a personal member of the FIWARE Foundation, which is managing, promoting and evolving the open source platform for smart IoT scenarios. He currently serves as a member of the FIWARE Technical Steering Committee (TSC). Furthermore, he is the AIOTI WG03 sub-group chair for the semantic interoperability topic. He has (co-) authored more than 50 technical papers and has been active as peer reviewer and program committee member for several journals, conferences and workshops. He is a member of IEEE, IEEE Computer Society, IEEE Communications Society and ACM.

Title of the Talk: Sharing IoT Information and Making it Available across IoT Platforms

Abstract: To achieve the full potential of the Internet of Things, the sharing and reuse of information in different applications and across verticals is of paramount importance. IoT information is often based on low-level data provided by sensors that it is not directly usable by applications. Thus, higher-level information or knowledge has to be derived and made available in a suitable representation. The important aspect for enabling the sharing is that there is agreement on the semantic concepts.

As a proposal on how this can be achieved, the NGSI-LD information model will be presented, which is being specified by the Context Information Management Industry Specification Group of the European Telecommunications Standards Institute (ETSI ISG CIM). NGSI-LD is based on JSON-LD and represents knowledge as a property graph. This graph enables finding relevant information on a suitable abstraction level, integrate meta-information like quality and provenance, and link to external information sources providing information not directly representable in NGSI-LD, e.g. video streams or 3D models.

NGSI-LD is used as the internal common format of an IoT virtualization platform developed in the Fed4IoT project. The idea is to provide users with a virtualized view of the information, i.e. only the relevant information in the required representation using the desired IoT platform, e.g. oneM2M or FIWARE. To enable this, information from heterogeneous source platforms is integrated into the virtualization platform, transformed into the common format, and then translated to the respective target platform. Thus information can be shared, bringing it to a common basis and then can be made available across different IoT platforms.


Jun Deguchi, Kioxia Corp. (formerly Toshiba Memory Corp.)

Jun Deguchi received the B.S. and M.S. degrees in machine intelligence and systems engineering and the Ph.D. degree in bioengineering and robotics from Tohoku University, Sendai, Japan, in 2001, 2003, and 2006, respectively. In 2004, he was a Visiting Scholar at the University of California, Santa Cruz, CA, USA. In 2006, he joined Toshiba Corporation, and was involved in design of analog/RF circuits for wireless communications, CMOS image sensors, high-speed I/O, and accelerators for deep learning. From 2014 to 2015, he was a Visiting Scientist at the MIT Media Lab, Cambridge, MA, USA, and was involved in research on brain/neuro science. In 2017, he moved to Toshiba Memory Corporation (the company name has been changed to Kioxia Corporation since Oct. 1st, 2020), and has been a Research Lead of an advanced circuit design team working on high-speed I/O, deep learning/neuromorphic accelerators and quantum annealing.

Dr. Deguchi has served as a member of the technical program committee (TPC) of IEEE International Solid-State Circuits Conference (ISSCC) since 2016, and IEEE Asian SolidState Circuits Conference (A-SSCC) since 2017. He has also served as a TPC vice-chair of IEEE A-SSCC 2019, and a review committee member of IEEE International Conference on Artificial Intelligence Circuits and Systems (AICAS) 2020.

Talk Title: Co-optimization of Hardware and Algorithm for Energy-efficient Edge Computing with Convolutional Neural Networks

Abstract: A lot of accelerators for convolutional neural networks (CNNs) have been developed and utilized in the cloud for a variety of applications such as image classification, object detection and semantic segmentation. In order to deploy CNN accelerators on edge devices, it is essential to drastically reduce memory and computational costs due to their limited battery capacity and hardware resources. Although quantization of activations, weights and their vectors is one of promising approaches to reduce the costs, recognition accuracy of CNNs tends to be degraded when neural network models are deeply quantized with e.g. single-bit representation. Therefore, in order to practically reduce the costs while maintaining recognition accuracy, co-optimization of quantization algorithms and hardware architecture of accelerators has become important recently. In this talk, co-optimization of our proposed filter-wise quantization technique and a specific hardware architecture with variable-bit-width MAC (multiply-and-accumulate) units is introduced.


Khalil Drira, Research Director, CNRS, France

Khalil Drira is Research Director at CNRS, the French National Center for Scientific Research. He obtained the Computer Science Engineer degree from ENSEEIHT/INP Toulouse in 1988, and the PhD degree in 1992, and Research Habilitation in 2005 from Univ. Toulouse 3. He joined CNRS, the French National Center for Scientific Research, in 1993 as a Tenure Researcher. He was the head of the Network and Communication Department at LAAS-CNRS, from 2012 to 2015, and the head of the research team SARA at LAAS-CNRS ( His research interests include design, modelling, analysis, implementation, and provisioning of smart networked services for IoT platforms and applications. He is author of more than 200 papers in international journals and conferences. His is involved in standardisation activities in IoT/M2M services platforms. He is expert for the European Telecommunications Standards Institute (ETSI), and served as member of the Specialist Task Force 547 “Security/Privacy and Interoperability of standardized IoT Platforms”.  He is or has been involved in different European and French R&D projects (FP6, FP7, ITEA, H2020, PIA) in the field of distributed systems and IoT/M2M and software architectures. He serve as Associate Editor for several international journals including: IEEE IoT journal, Internet Technology Letters, and smart science journal. He has been guest editor of  Special Issues in international journals including recently JSS, CCPE, FGCS, IST.  He was co-editor of the following SPRINGER volumes: LNCS 11895, LNCS 2236, LNCS 7957. He was co-editor of over 20 international conferences and workshops proceedings. He chaired several international conferences including : ICSOC 2019,  ECSA 2013, IEEE-WETICE 2012, 2015, 2020 and SERA 2015.  He has co-organized international workshops and tracks including: ASOCA@ICSOC;  SISOS@ACM-SAC. He has served in the Steering Committee of the international conferences IEEE-WETICE and ECSA.  He has served in the Program Committee of over 100 international conferences.

Talk Title: Semantic Interoperability for IoT Service Platforms

Abstract: The focus of interoperability has been initially on technical interoperability (basic connectivity, network interoperability) and syntactic interoperability (Common Information Models with static information based on a pre-defined syntax). This was reflected in the work of standardisation with many great achievements.  However, as soon as the requirement on the information exchanged become more complex (e.g., systems from different sectors), static information is no longer sufficient, and the need arise for basing the exchange of information on its meaning (independently of underlying protocols). This is the role of Semantic Interoperability: making sure that the meaning of semantics can be understandable and processed by machines, and the most common way to achieve this is by using an ontology which is “an explicit specification of a shared conceptualization”. One of the main objectives of the standardization and research activities for the last decade was to enable the transition from the Internet of Things (IoT) to the so-called “Web of Things” (WoT). The ultimate objective for the next decade is to enable the transition from the Web of Things (WoT) to the so-called “Semantic Web of Things” (SWoT). This talk will investigate the challenges for semantic interoperability and autonomic management in IoT service platforms.   The landscape of standardisation and research initiatives will be explored. The hindering and enabling factors for architecting interoperable IoT applications will be discussed.


Bin Xie, Founder & CEO, InfoBeyond Technology, USA

Dr. Bin Xie is the founder & CEO of the InfoBeyond Technology (2008). He conducted substantial R&D works in the areas of Networks, Machine Learning, and Security. His R&D of Security Policy Tool technology is awarded as a Successful SBIR Story by NIST in 2018.  His R&D of Cyber Security Architecture Tool is also awarded by NIST. As a PI, he has received 8 million of R&D grants for DoD (e.g., Naval Research Laboratory, Air Force Research Laboratory, Missile Defense, and Marine Corps), DoEOffice of Science, NIST, and DoT. Some of his R&D technologies are transferred into the governments and commercial domains.

Dr. Xie has published 70+ articles. He is the co-editor/author of books titled Handbook/Encyclopedia of Ad Hoc and Ubiquitous Computing (World Scientific: ISBN-10: 981283348X, World Scientific Publisher) (Best Selling in 2012 &2013), Handbook of Applications and Services for Mobile Systems (Auerbach Publication, Taylor and Francis Group, ISBN: 9781439801529) and Heterogeneous Wireless Networks- Networking Protocol to Security, (VDM Publishing House: ISBN: 3836419270, 2007).

Dr. Xie severed as a member of NIH Special Emphasis Panel on System Science and Health in the Behavioral and Social Sciences, ZRG1 HDM-Q (50), 2012-2017.

Talk Title: DiagSoftfailure: Soft-failure Diagnosis for High-throughput IoT Network by Learning

Abstract: Many IoT applications (e.g., emergency cares) rely on low-latency or/and high throughput data delivery. However, performance diagnosis in an IOT network comes to a challenge due to a complex networking environment, especially at the bottleneck (i.e., edge) of the IoT network. A network soft-failure (e.g., an insufficient buffer in an edge device) is characterized by the degraded performance. This is different than a hard-failure that is an issue of network disconnection or function loss. Current soft-failure diagnosis highly depends on an expert to analyze the performance symptoms and from which to track down the cause and source of the failure. This results in a high time and labor cost for accurate diagnosis and troubleshooting. On the other hand, current approaches are not only limited to distinguish soft-failures accurately but also unable to identify the root cause of the soft-failure in the complex network. Powered by advanced Big Data machine learning technology, DiagSoftfailure is a novel approach in attempt to address the diagnosis challenges in a way to provide an intelligent and automated diagnosis tool that can be used for IoT edge devices. For such a purpose, DiagSoftfailure is trained by combining supervised and unsupervised machine learning to accurately identify both known and unknown soft-failures in a network. Substantial machine learning results will be demonstrated in this talk.