How Pandio’s Centralized and Decentralized Messaging Make IoT, AI, and ML Visions Reality
When people think about internet of things (IoT) devices, they often think about the things that keep us connected—at home, at work, and in the car. For example, IoT devices like your thermostat, which allows you to adjust your temperature from an app or by voice wherever you are, or your doorbell that feeds you video so you can see who’s at your home, even when you’re not there.
IoT devices and related automated services are making lives and work much easier. That’s why we’ve seen a consistent (and now rapid) increase in device usage over the past decade.
According to Statista, back in 2015, there were about 3.6 billion active IoT connections worldwide, representing the number of actively connected nodes, devices, and gateways with end sensors (not devices like cell phones or computers).
In 2021, it’s estimated those active connections will reach 13.8 billion, and by 2025, it will more than double to almost 31 billion active connections.
Other reports suggest that by 2030 there will be more than 125 billion active IoT devices, with the expectation that for every person who owns an IT device, they may own upwards of 15 connected devices by that year.
The Move Away from All-Centralized Data Crunching
As technology has evolved, so have the mechanisms that make these connections and data-sharing possible.
In the past, network connections primarily facilitated these connections that only enabled the IoT device to send and receive data to remote centralized data centers. But today’s technological advances, such as microprocessors, make it possible for that data to live right in the device itself—for example your washing machine, your car, or even your refrigerator. Often, without ever having to leave that device.
But the more data the device collects and analyzes, the more complex those data processes become.
Those complexities have, for a long time, prevented some companies, especially small and mid-sized businesses (SMBs), from pursuing the possibility of adding IoT capabilities to their existing product lines or innovating IoT products for the future.
There is hope, however, for companies of all sizes, thanks to simplified data exchange capabilities produced by Pandio, powered by Apache Pulsar.
But before we get into some of the many ways Pandio can help simplify distributed messaging and enable AI and machine learning for your devices, let’s take a closer look at IoT and data streams in those devices.
First, What is IoT?
Internet of things (IoT) is a term coined to describe devices (the things) with embedded sensors and software inside of them.
The sensors and technologies collect data and then transfer that data in a variety of ways over a network (the internet). IoT devices can talk to one another and transfer data without reliance on human interactions.
The more connected devices you have collecting and sharing data, the more automated (and potentially smarter) decisions you can make.
For example, instead of manually setting your thermostat to a cooler temperature every night and then increasing the temperature when you wake up, an IoT thermostat can track your inputs, learn from your actions (day, time, temperature settings, etc.) and then create a schedule based on your preferences.
If your device has the capability to analyze power usage, it can even make recommendations to help you decrease consumption.
Traditionally, when people think of these types of data exchanges and machine learning opportunities, they think of software in a device doing specific tasks—for example, your favorite social media platform.
In this example, you use your phone, laptop, or computer to activate software (for example a website or a web app). That software collects information about your activities and then uses a network connection to send that information to centralized data centers. The data centers are where your data lives, where it’s analyzed, and where recommendations or other functions are fed back to you.
IoT devices are different. The logic that has traditionally taken place in those data centers can live (and fully function) right in the IoT device.
What are Some Examples of IoT Devices?
There are a vast array of IoT device types. Here are a few examples of IoT devices in homes:
- Personal assistants, like Alexa and Siri
- Door locks
- Coffee pots
- Washing machines and dryers
- Wearables like smartwatches, smart rings, and smart bracelets
- Garage door openers
- Light and plug switches
While IoT devices are increasingly common in home environments, they’re also growing in usage for industries and business. In these settings, these devices are often referred to as the industrial internet of things (IIoT) or Industry 4.0, referring to the Fourth Technological Revolution we are experiencing in modern times.
Here are some examples of IIoT in these settings:
- Sensors used to monitor critical infrastructure like transportation, water plants, power plants, and more
- Robotic technologies, for example, on assembly lines
- Farm machinery
- Inventory and production management
- Medical devices
- Predictive maintenance for machinery
- Building automation system monitoring
IoT Data Processing
Earlier, we shared an example of how centralized data is generally processed in SaaS, cloud-based, or web-app platforms, like social media. But with IoT, that same data processing, can be decentralized, meaning it happens where the device lives, not transferred into a complex data center for analysis and tracking. This means a reduction in complexity and cost.
Here’s an example: Let’s say you have a smart washer in your home. That washing machine has a sensor inside that can give you insight into the machine’s performance, for example water temperature or cycle stage.
In this example, you want your washer to alert you when a load of laundry is complete and send that alert to your smartwatch while you’re working out on your smart treadmill.
All of the data collected and processed for that load of laundry only exists and lives within the washer. It doesn’t have to be sent over the internet to a data center for an action and then back to the machine. After determining the load is done, the washing machine uses your network connection to send an alert to you via your app or other option.
For the gamut of IoT applications and use cases, the washing machine example is fairly simplistic. Let’s look at a more complex example: a smart car.
In this example, the smart car is loaded with cameras and other sensors that help improve performance on many levels—for example, the ability to safely change lanes, alert you if vehicle is malfunctioning, if there are vehicles in your blind spot, how to safely park, or even valet your car to you without human interaction.
Unlike the few tasks handled by the IoT washing machine, all of the sensors and cameras in the smart car constantly collect data and work simultaneously. That data helps the vehicle make real-time data decisions, and that’s why data processing happens inside the vehicle.
Think of it like this: What if instead of processing data in your vehicle about whether or not you can safely change lanes, your vehicle has to collect that data, send it digitally to a data center, wait for it to undergo analysis, and then send information back to your vehicle so you can make a decision—do I want to change lanes or stay where I’m at?
Even at the fastest technological capabilities possible, the delay in data transfer could very well mean you don’t have the right information you need to make split-section decisions. While you’re waiting for the data to bounce out and back to you, you switch lanes, swiping the car next to you.
Handling that data within the vehicle speeds up response times and helps deliver near real-time, data-driven decision-making capabilities.
But, that two-way connection, in this example, is also not lost. Here, the vehicle can send certain data to and from the data center to improve performance, for example, when technologies need updates. This example marries centralized and decentralized data processing, which in the end, makes your decentralized data (data specific to your vehicle) even better.
The Big Data Conundrum
IoT and IoT devices fuel “Big Data,” and big data from these devices are critical for successful implementation of AI and machine learning models to make better data-driven decisions.
But when you have devices with a lot of sensors and a lot of data collection and analysis, each device must, in essence, serve as its own data center.
Traditionally, the technology and expense of building these devices so they can actually handle data at that level has been too taxing on many companies so they avoid IoT altogether or limit their thoughts about innovating with it in the future.
Pandio makes this attainable for your company, regardless of how big or small you are.
Remember the IoT washing machine device we talked about earlier? By enabling messaging in both centralized and decentralized manners, Pandio can handle everything from that simple alert when a load is finished, all on a device so lightweight it can exist like a tiny computer inside that washer, to more complex functions, like updating the operating system, and tracking maintenance and other performance issues.
And for bigger data projects, like that smart car, Pandio can run an entire system within your car and facilitate other external data transfers as needed. It can enable your smart car to create events, write those events to an input topic, read the topic, and then process an output, including written logs.
In both of these examples, you can take data collected from those devices (any IoT device) and feed it into AI and machine learning models to help you (and your device) make better decisions.
Think of Pandio as the software that does the data collection within your device, and then, when and if needed, using configured logic, it can send that data off to a centralized location for AI and ML.
Pandio can fuel increased adoption of IoT, AI, and ML by decreasing operational costs and simplifying implementation and data management, making things you never thought were viable for your company a real possibility.
Here’s an example in manufacturing: You have multiple machines handling a variety of tasks on your shop floor. Right now, you have an employee responsible for tracking maintenance schedules and charting when the next maintenance should take place. With Pandio, you can leverage sensors within those machines to continuously track performance and alert you when there is an issue and/or when scheduled maintenance should occur.
Pandio gives you the ability to compute at the edge of the actual sensor on-site because it reduces the amount of data you have to transfer to your centralized data center, while also eliminating human interaction, manual tasks, and the possibility of human error.
With IoT devices, you need a streaming system with durable log storage and Apache Pulsar has both. It supports cloud, edge, and device. And, Apache Pulsar has a unique ability to separate processing from storage meaning your processing can scale as you need it, without negatively impacting storage (think costs, servers, etc.) Because it’s multi-tenant, you can always scale up and down as your IoT device needs it.
When you add in the layers of ML and AI for those functions, you can run real-time analytics, learning from your devices over time, and predicting core decisions for you, with actual data, not guesses.
Overcoming the Fear of Shutdown
Another barrier to IoT adoption has been the fear of the processes associated with setup and deployments causing device shutdown or data loss.
One of the many great things about how Pandio works is you have multiple deployment options so you can always ensure one component remains completely functional and operational, while your new one, for example, a new sensor, gets set up to run alongside it.
You can even have this additional sensor learning your existing IoT protocols and fully control the “flip-the-switch” when the new one operates the way you want, producing the data and information you need.
Regardless of industry, Pandio helps enterprises implement or optimize their IoT strategy by reducing complexity around data collection, dramatically reducing the costs of data processing, and streamlining data orchestration and analysis within your organization.