The Wechsler Consulting Cloud News – October 2020 – episode sheds light on new and very promising features in Azure Active Directory (Continuous Access Evaluation), Event Grid integration of Key Vault and App Services Private Endpoints, all of which are great assets in creating modern, scalable and efficient Cloud solutions.
Focus Topic this month is Tiny ML, a completely new approach using ML models on small embedded devices, bringing intelligence closer to places where things are happening. This can turn into nothing less than a game changer in the IOT industry!
In nearly every IOT project I had the opportunity to work in, time series data played a very important role. The problem for this type of data is that it normally comes in larger volumes and is therefore not always great to handle. This is especially true in projects, where you have to cope with small storage on devices and no central data store, which makes it very hard, if nearly impossible to get a global view on the behavior of these solutions in time. One could work with thresholds and alerts, but this approach never gives you the chance to detect trends and get “ahead of the wave” to react better, faster and more precise to certain events. Some of the industrial communication standards, such as OPC UA and SCADA, try to tackle this issue by providing historic data functionality in their communication layers, but this is just a single aspect of a comprehensive data solution.
Cloud architectures are able to help in this case, if you have the chance to collect time series data either centrally, or on the edge. A very valuable asset in Azure and in this context is Time Series Insights. It is a cloud service allowing you to handle query, transform visualize and correlate your different data streams into comprehensive views and insights. There are also connectors into reporting tools such as Power BI available. Using the M365 infrastructure Power Automation or Azure Logic Apps and Functions, serverless integration into corporate business process and control processes is also not a problem.
Bandwidth terror through an abundance of chatty sensors!
For example, in a manufacturing building, network traffic would go through the roof, if “everything” would be directly connected to the Internet (although, it is the Internet of Things) and, of course, it would be a security nightmare, too. Well, let us leave the latter topic aside for a moment and stick to the traffic requirements. The Cloud promise was, connect everything to the Internet and in the Cloud, magically, everything gets done! While this is not false in a lot of scenarios, it is not always true! Depending on solution use cases in focus, there are quite a few scenarios where distributed smart architectures have significant benefits over a centralistic approach. In these cases devices on the “Edge” come into play. They are gateway devices running pre-processing logic and providing storage capabilities to handle part of the overall system workload on-premise, on the edge to the Internet. By doing this, enable the transformation of raw events into higher quality events, such as e.g. the reporting of temperature sensors only, if set limits are exceeded. The higher quality events are passed into the Cloud solution and are handled there to trigger related business logic. Benefits of this design include
Significant lower traffic on central system
Better manageability / monitoring and security of data flow
Robustness against network outages (at least in some of the scenarios)
…
but, it comes with challenges, as well, such as:
Handling of business logic on the edge
Device management of IOT devices as well as edge devices
…
This needs to be taken into consideration!
However, if you are a vendor creating Cloud solutions experiencing a lot of data ingress, sooner or later you end up installing Edge devices to sort out raw events spamming your backend. You will buy devices and or talk to the customer to install these devices in the on-premise data-center. The drawback with this approach is that it adds a lot of upfront cost to your solution. Edge devices might be quite capable, full-fledged and therefore expensive servers that a customer may also want to include into his system management to keep them patched and secure. This triggers often time consuming approval processes to get things into place. All of this may kill your project or POC before it even has started!
So, is there a smarter way to approach this issue? As you may have guessed from the title of the post, there is. Microsoft is extending its Cloud-native rent-my-system approach to edge hardware and software. In this case the system of interest is called Azure Stack Edge (fka Azure Data Box Edge – Microsoft likes the renaming game, as we know). There is a very informative IOT Show episode on this solution, still using the old name.
Benefits of the Cloud-rental approach are that the Stack Edge devices are managed centrally via an Azure service. The service allows the installation and management of Azure IOT edge modules taken over the responsibility for the distributed logic in the system. Looking at IOT devices, it is especially interesting that these devices can be connected to a local (Edge) instance of IOTHub and also be managed from there, which gives you the best of both worlds: The devices are safe behind the firewall, but still accessible via the Edge gateway for administrative purposes! Microsoft operates the Edge devices as an appliance, which means it takes over responsibility for any OS (Stack Edge is running on Linux) or runtime patches. IOT Edge modules deployed can be available building blocks from Microsoft or 3rd party vendors, as well be self-developed Edge modules suiting the implemented solution.
This is really powerful, because it leaves solution developers the flexibility to draw existing commercial building blocks e.g. for AI or highspeed processing from Azure Marketplace and focus on the business needs of the solution.
If a customer needs a test at their location, the rental model is not to beat. Just ship your devices and a pre-configured Edge device and a POC can be up and running in minutes, not costing a fortune and hardware. If not suiting, it can be stopped any time with out wasting more costs and energy.
Looks like a quite innovative, efficient and modern approach to me! 🙂
