Author: Derek Mak Date: January 23, 2020
Internet of Things (IoT) is one of the most significant catalysts for digital innovations. Tens of billions IoT devices are connected and are generating data that needs to be collected and mined for actionable insights. A high-definition camera at a street corner is creating 200GB of data per day, there are over 62 million security cameras in North America (2016 figure). An autonomous vehicle generates 11TB of data per day. A modern commercial aircraft generates 5-8TB per flight. Analyzing and combing data from different devices can help businesses make faster and more quality decisions, which can lead to productivity gain and better business outcomes.
IoT technology is pervasive. All major technology categories including networking, security, cloud, edge computing, blockchain, machine learning/artificial intelligence (ML/AI), autonomous vehicles, augmented/virtual reality (AR/VR) play a role in the IoT market evolution. Fortune Business Insights in a report, titled “GLOBAL INTERNET OF THINGS (IOT) MARKET: GLOBAL MARKET ANALYSIS, INSIGHTS AND FORECAST, 2019-2026” states that IoT technology holds significant potential in the ICT sector. As per the report, the global market was valued at US$ 190.0 Bn in the year 2018 and is anticipated to reach US$ 1111.3 Bn by 2026.[1]
The promise of IoT for mankind is a smarter tomorrow. That is based on deep learning and analytics of the data from the connected Things, or generally refer to Assets in the industrial sector. The data is generally coming from sensors embedded-in (new digital assets) or bolted onto Things (existing assets). Virtually all IoT projects involve one or more of the following objectives:
monitor the state of the asset to detect abnormality (e.g. improve worker safety).
collect and analyze historical data to predict future state of the asset (e.g. avoid downtime).
perform what if scenarios using deep learning techniques to optimize the performance of the operation (e.g. optimize operations).
The key is understanding the data. Without it, there is little value to IoT. Many IoT initiatives struggle to go beyond proof of concept due to common reasons such as misalignment to business objectives, underestimated technical complexity, lack of expertise, data quality issues and security concerns etc. I submit the number one challenge to IoT initiatives is the effort required 1) to establish connections to assets to extract data and 2) to normalize the data from multiple assets with different data formats and time sequence so meaningful analytics can take place. In addition, asset manufacturers are not quite cooperative in sharing their data. All of them recognize the value of data and are trying to bring others into their data ecosystem. Data ownership is an enormous huddle to resolve to break down the data silos present in all businesses today. I will take up this topic in my future blogs. Stay tuned.
The next huddle is processing of the data. Common approach today is to forward the data to the cloud for analytics. For latency sensitive situations such as manufacturing production line, vehicle-to-infrastructure and vice versa communications (V2I/I2V), fleet management etc., sending data to the cloud simply isn’t going to work for many reasons, such as cost and data privacy. Similarly, remote operations such as offshore drilling where communications are at a premium and reliability is questionable, the extract and forward approach will not work.
A better approach is to connect the data at the edge, meaning to extract the data from the assets where the assets are, processing the data at the edge so there is no time delay, and much less limitation to what data to keep. A computing device, a compute gateway or appliance, for example, can collect and process data at the edge of the network where assets are connected, and then send only the relevant data to the cloud for analysis, reducing bandwidth needs or it can send it back to the edge for real time application needs. Collecting, processing (including encryption), and storing the raw data at the edge of the network can increase security and privacy, providing the business with full control of the data.
Apart from the greater security and privacy benefits that edge computing enables, other key benefits are low latency and less dependent on communications infrastructure. These features enable more efficient real-time applications that are critical to real world scenarios. Applications such as self-driving cars, augmented and virtual reality, smart cities, automation systems and other that require real-time processing and response. By connecting the data, all the data, at the edge where data is being generated, raw data is available for deep learning and mining in order to formulate understandings, to make decisions based on facts, and to build a smarter tomorrow!
Comentarios