Getting started with IoT

Editor's Notes

1. Presentations This webinar is based on a couple of findings. One, most of the IoT initiatives never land in a production environment and, two, there are still several organisations wondering if IoT fits them During this webinar, we will draw the current IoT landscape and what an IoT adoption means. Ralf will also describe the impact of 5G on IoT. Top use cases for IoT-enabled by 5G - Microsoft Industry Blogs We will then review what an IoT infrastructure looks like and explain how to address some typical challenges. And then, we thought it would be interesting to share with you some customer stories which illustrate the various points discussed before.
2. [Max]: So, Ralf, my first question is about the sectors that would benefit from IoT. I would assume that there are sectors that are more sensitive to IoT. Is that a correct assumption ? [Ralf] [Max]: What would be the main motivations for a transition to IoT ? [Ralf] [Max]: Looking at the near future, which strategies and technologies should organisations keep an eye on ? [Ralf]
3. [Max]: As I mentioned in the intro, there are still today a decent percentage of IoT initiatives that don't reach production. How do you explain that ? [Ralf] [Max]: Can you illustrate some of these steps ? [Ralf]
4. [Max]: As a Digital Transformation Lead for a TelCo, this subject must be hot for you. Can you tell us more about it ? [Ralf]
5. [Ralf]: Turn to technical. Ask Max to describe what is a typical IoT ecosystem [Max] Ecosystem is, actually, the appropriate word as we have a physical world and a set of organisms or components. [CLICK] First and foremost, we need devices acting as an interface with the physical world. These devices collect and/or create data. Depending on the device capabilities, the protocols used and the platform, these devices can dialog directly with the backend system or not. Unfortunately, some devices do not support internet protocols and therefore a protocol translation is required. This is done via a gateway which will also facilitate the dialog with the cloud infrastructure. I will discuss this further later. Once these messages or events land, they reach the IoT platform where data is collected, stored in the right place. [CLICK] Following that, automation will take place. This is where business rules will be executed, some anomaly detection could take place, and some alerts be raised. All these events and derived data populate your data lake, based on which you can elaborated analytics and machine learning models and extract insights about what's going or what could happen. [CLICK] The last step can be called the feedback layer, whether it is through reports or actionable decisions that will justify to adapt and adjust what’s happening in the real world [Ralf]: IoT ecosystem is a hybrid environment with a set of things and gateways and THE IoT platform which sits in the cloud.
6. [Ralf]: IoT ecosystem can be vast. Ask Max to illustrate which component sits where [Max] [CLICK] If the trend today is to move as much as you can towards the cloud, this is partially applicable to IoT. Some of the processing has to happen on the edge so on the client side and potentially on the device side. Client here being the infrastructure closer to the devices and not in the cloud. [CLICK] A distributed IoT solution is based on four nodes: smart devices, edge gateway, micro data center and cloud data center. These 4 are spread across 3 tiers: Cloud, Fog and Edge. The intention is to extend the cloud paradigm to the edge with the assumption that consumption and storage can take place anywhere and not only in a dedicated cloud datacenter whether it is regional or global. Those global data centers run global applications, allow deployment of multi-location updates or execute large-scale analytics. Today, cloud providers invest in more regional cloud centers to bring customer’s solutions closer to their own premises. Despite this regional strategy, it is not enough and this is where we come to fog layer. Micro data centers are here stored in a network layer or on premise and aim to address specific needs such as latency and privacy. [CLICK] As a TelCo, Proximus Luxembourg is particularly involved in this, through the rise of 5G connectivity. It brings new opportunities for our customers thanks to fatser broadband, reduced latency and support high density of devices. If we move another step further, we reached the edge layer. This one contains small edge devices which can execute some logic with limited resources. [CLICK] Lastly, the closer you are to your devices, the more bandwidth you have with a low latency which allow you to react quickly on some insights. On the cloud, latency becomes bigger as you don’t have the bandwidth but you have the computing capacity. [Ralf]: Why this distributed approach is critical: reduce management, decomplexify solution, …
7. [Ralf]: Talking about these edge components, which key features should they have ? [Max] Device lifecycle management Key here is the Provisioning: make sure devices can authenticate and register themselves in a secured way. After successful provisioning, device will start exchange data. The device should be able to reprovision itself in case of security breaches or other issues that could happen Because IoT edge devices run more logic than IoT device. The difference being that an IoT device sends telemetry data only, it is very important to consider the configuration. You should be able to update and adjust the configuration of them, remotely Because we have lot of logic running, monitoring is crucial. This will help to understand the health of the device and the services it provides. Device & data protection These devices must be secured by design and make sure that the communication channel is protected so the device can only dialog with your infrastructure. The idea is to prevent triggering a restricted or forbidden action in physical world. Convey data and commands The next one is fairly obvious. Commands is a key here. It can be an update, a reset, reboot, whatever Logic and data analysis And lastly, you would expect these devices to perform some tasks on the edge and not use cloud resources
8. [Ralf]: So far, presuming that we start from scratch, but from experience it is not the case. Throughout our projects, we have realised that organizations who decided to invest in IoT, realize fairly quickly that their current stack of devices is not really designed for modern communication. Just to be clear, we're talking about devices that only implement very old and legacy communication protocols. But, still, these devices generate a lot of data which is very valuable because it contains lot of insight A typical mistake is to ignore the current device stack and to focus on a new stack only. This is, in our experience, a mistake because the organisation will be invest money, time and people's resources effectively in building something that none of the existing devices are going to be able to benefit from. And, by doing so, the time to value is going to be measured in years. So you can easily image what the reaction of business managers, owners, or CFO/CTO will be when you will explain that is going to take years to derive value from that investment. This is a good example of retro-fitting problem that loads of organizations are facing. [Ralf]: Max, how technically can I address this ? [Max] The most effective way to handle that is by using field gateways. These gateways are located close to your on-premise devices and convey the data and command from and to the IoT platform. They act as a communication translator or facilitator and are able to process the data sent by the devices. Now the question is why should you consider fields gateways
9. [Max] The following content is extracted from a study done by IDC on industrial automation but, in our experience, it is equally applicable to other industries [CLICK] 85% of devices are not internet-ready so that means they don't understand modern communication protocols, they don't understand modern reliable encryption technologies, etc. They are pretty much designed to provide a simple functionality and, maybe, implement some communication interfaces but not something that you would trust. [CLICK] Fields gateway come with native functionalities that would require a huge amount of development effort if they had to be implemented. [CLICK] Also it's important for you to be in a position where you can perform edge processing What I mean by that is that data will be processed by a field gateway, on the edge and not on the cloud By doing so, you can achieve up to 60% reduction on your cloud costs This is really important because we see customers transmitting all data coming from the device, which represent a big volume of data sent to the cloud every second or, sometimes, every 100 milliseconds. This is also due to the fact that the device doesn't understand yet how to filter data before it's being sent back to the cloud and if you send everything to the cloud you're going to have to pay a lot of money in terms of transmission costs, storage cost and processing costs. So having the capability of pre-processing on the edge will allow you to perform dramatic savings in your operating costs [CLICK] There's also the topic of network optimization If you have multiple devices in the same physical location for example, which would be the case for a manufacturing facility for example, using a field gateway will give the possibility to use features like message batching and compression. Message composition is another interesting feature. This is where you take a message from one device and you merge it with messages from other devices to create one single composed message. You can also do things like message queuing so that you make sure if the network connection on the field is interrupted or down you still receive the messages on the field gateway. The gateway will queue those messages, persists them in memory or maybe on disk and when the network connection is re-established then all the messages waiting in the queues are transmitted so they don't get lost. You can also look at fluttering capabilities which is when maybe you have limited bandwidth and you may have a lot of devices sending data to the cloud but you want to optimize the network utilization [CLICK] It's crucial that the gateway gives you a consistent implementation of infrastructure capability so the credentials management and the security of the devices is handled in the same way, regardless of the number of development teams involved.
10. [Ralf]: As usual, should aim to balance efforts vs costs. Our recommendations is to focus on the value generators and leave to the platform the handling of commodities [Max] An IoT platform comes with loads of commodities that should be used. All this infrastructure plumbing is not really something that really add value to your business So things like Message handling, Transfer protocol implementation, How to manage credentials on the field so that your devices can authenticate in a secure manner How do you monitor that your devices are working okay by implementing health and reporting health and diagnostics capabilities How to persist messages These are all things that are not trivial to implement but still required to have a reliable IoT environment But, again, they don’t really bring value to your business and, yet, will impact the implementation phase and therefore your deployment plan. Instead of doing all that you should be spending your development time and resources doing things that generate values for your organization So that means investing time in implementing edge analytics. Maybe you don't want to send all the data from every device all the time. Maybe you want to implement some rules to generate alerts based on some thresholds Maybe you want to be able to push these rules and thresholds over to the field and make your devices acting based on them Now, and this is a bit contradictory, you might want to spend time implementing your own protocols, particularly if you have some proprietary protocols
11. [Max] Let's have a look at an ideal architecture which would address several processing modes: real-time, near real-time and batch First you have your device, things and, or data sources. This is where proper real-time mode is executed. [CLICK] The data are sent to the edge, to a field gateway for instance which will handle the data flow through a communication layer [CLICK] Then are then sent by the gateway to the cloud and land first on the device management and ingestion layer. [CLICK] Switching to near real-time, the streaming analytics path is activated. This path support the execution of complex event processing, time series analysis etc. It will also trigger actions and or interactions with other systems. [CLICK] The batch mode shouldn't be forgotten as it will dictate the [CLICK] type of long-term storage required and the [CLICK] cold analytics path [CLICK] The connectivity with existing "business" applications can then be set. So far we had a look at the edge-to-cloud flow. [CLICK] Looking at the cloud-to-edge, data and configuration will be transmitted back to the device or the edge [CLICK] And let's not forget the edge processing. [CLICK] Lastly, this capabilities rely on a combination of infrastructure, security and monitoring layers and a strong DevOps practice. Ok, I think it is enough with best practices and let's move on to our customer stories.
12. Bühler Group uses IoT in the food processing industry to minimize waste, improve quality, increase yields, reduce downtime and slash energy use. The result for their customers is more control of the food production process and huge cost savings. 
13. [Max] Gathering data from the machines and generate insights imply a platform with the appropriate capabilities. The first need here was to elaborate a scalable architecture considering that they had thousands of machines worldwide. Besides, some machines had no digitalization capabilities but had to be integrated into the digital landscape.  So, we recommended to combine Microsoft Azure and IoT technologies to form a comprehensive solution. Older equipment was retrofitted with Codit IoT gateway technology called Nebulus. This streamlined and speed up the delivery time. The pilot has been a complete success and has opened the door for other projects. Bühler Group estimated that they could save 15% of the food waste, labor costs would be reduced by 30% and energy costs by 5%
14. Hub One was looking for a solution to keep real-time track of passenger luggage and feed these information back to the airlines.
15. [Max] The biggest challenge was the drastic shift from an old and legacy system to a solution that combined IoT, Cloud, and on-premises systems. Considering the size of this project, Hub One asked Microsoft who recommended us. We, first, carried out a feasibility study in order to validated Hub One’s concept and assessed the value of the platform. We then setup the Azure-based integration platform where data was processed and sent to airlines’ data centres. Today, and thanks to this hybrid platform, airlines at Paris-Charles de Gaulle Airport receive real-time information about their passenger’s luggage. Besides, Hub One has been able to reduce their operational costs.
16. Televic provide an healthcare communication system to give patients the freedom and flexibility to move across healthcare facilities. 
17. [Max] Televic was looking for a tracking mechanism to provide continuous care to their patients. Thanks to Microsoft’s recommendation, Televic asked for Codit expertise in order to speed up their cloud-based IoT adoption. The devices is connected toAzure data platform via Codit’s Nebulus™ IoT Gateway. This signal then localizes the closest nurse or caregiver and sends an alert to their smartphone or monitoring system.  Televic currently serves 80% of the hospitals and 60% of nursing homes in Belgium and are going to branch out internationally, to Europe and China.
18. Do Secure senior business sponsorship Focus on business requirements Start small, think big Build multi-disciplinary teams Engage the right experience Keep security central in all you do Use agile development processes Design for flexibility and change Take an ecosystem approach to IoT Do not Confuse data with insights Build a solution for all your needs Over-pivot on the technology Focus on future requirements Develop your own IoT infrastructure See middleware as the IoT solution Think connectivity is not the problem Underestimate the importance of high quality sensors