Just as companies have begun to gear up their digital transformations, CIOs and IT professionals have been tracking the arrival of the Internet of Things (IoT) with a combination of excitement and dread. Gartner forecasts there will be 8.4 billion connected “things” in use worldwide by the end this year, with that number growing to 20.4 billion by
The explosion of these smart devices will open many business opportunities, create new industries, and deliver huge advancements to the mundane – think connected lightbulbs, thermostats, and sprinkler systems. While the spotlight might be on smart cars, cameras, conveyors, and gauges, the true opportunity in IoT is about using the massive new streams of data that will be generated by the billions of edge devices. Not surprisingly, the promise of IoT brings with it significant IT management, security, and data challenges.
But some see the IoT-driven disruption extending well beyond those challenges. Peter Levine, general partner at the venture capital firm Andreessen Horowitz, argues that IoT will swing the pendulum away from the existing centralized cloud computing model to a new age of massively distributed computing occurring at the network edge.
In a presentation ominously titled “Return to the Edge and the End of Cloud Computing,” Levine makes a compelling case for IoT’s disruptive impact. Fortunately, his vision for the cloud isn’t quite as apocalyptic as the title suggests. Cloud computing will persist, but in a highly transformed way. Still, CIOs need to prepare now for the future that Levine describes, as aspects of it could well arrive faster than many of us expect.
At the core of Levine’s argument are the growing number of IoT devices and applications that must perform their computing locally. A clear example of this requirement is that of self-driving cars. By 2020, approximately 10 million self-driving cars will be navigating roadways globally, predicts BI Intelligence.
When a self-driving car’s “vision” system sees a stop sign or a pedestrian, Levine notes, it must take immediate action. There’s no time for the data to travel to the cloud for analysis, and for a suggested driving response to be transmitted back to the car. The input, calculation, and response must happen near instantaneously to avoid a potentially tragic outcome. Sensing, decision making, and action must happen immediately and predictably. In other words, this marks the fall of centralized computing and the rise of edge computing.
Today’s high-end luxury cars already contain about 100 CPUs, and self-driving cars may have twice that many, Levine says. In essence, cars are already becoming “data centers on wheels.” When you connect thousands of these cars together in peer-to-peer networks, “it becomes this massive distributed computing system at the edge of the network,” Levine says.
Cars are just one example of what Levine expects to eventually balloon to trillions of IoT devices. Sensors everywhere will be generating real-world data – location, acceleration, gravity, temperature, pressure, etc. – in staggering volumes. A single self-driving car, for instance, generates about 10GB of data for every mile that it travels, he says.
So what becomes of the cloud in this edgy scenario? Levine predicts centralized computing will become the hub of data analysis and machine learning. Information will be curated and filtered at the edge, with subsets sent to the cloud. There, machine learning algorithms can sort through the data and propagate any “learnings” back to the edge devices to enable tighter “sense, infer and act” loops.
Stepping back from the edge
Is it possible that cloud computing is destined for such second-tier status as compute power becomes more widely distributed? Let’s not be too hasty in jumping to that conclusion.
For starters, the emergence of peer-to-peer IoT networks at the network edge won’t replace the huge amount of processing being done today in the cloud, or the data increasingly stored in cloud infrastructure. Indeed, enterprise cloud computing
will continue to grow, as illustrated by IDG’s latest cloud computing survey. Organizations surveyed expect to have nearly 60% of their total IT environments in a public, private, or hybrid cloud by 2018, up from 45% in 2015.
At the same time, the rise of IoT will place new demands on the cloud. In line with Levine’s vision, edge devices that can sense, decide, and act better will become our collective measure of success. How will they be made better? By analyzing all of the data collected (sensed), all of the decisions made, and all of the actions taken. In fact, machine learning works best when vast amounts of data are analyzed centrally to identify new and better ways to sense, decide, and act. Rather than fading away, cloud will arguably become even more important in an IoT age. The key takeaway is that the cloud we are building and using today may not be what we need in the future.
In his talk, Levine tacitly acknowledges as much. The trillions of IoT devices will need to do more than just produce data streams for machine learning. These intelligent edge devices will need to be managed, coordinated, and secured, he notes. That and the critical activities of analytics and machine learning will be the role of the IoT platform. All of which is essentially software in virtual machines that could run anywhere – even in cloud.
And then there’s the idea that while edge computing is about real-time, time-sensitive tasks, centralized computing is capable of operating at a more relaxed pace. Enterprise cloud management, integrated enterprise cloud analytics, and enterprise cloud security will all need to evolve and advance to deliver the oversight services that edge computing networks will require. And while these management and security functions may not be under the same instant-reaction pressures as those faced by the IoT devices themselves, these cloud activities – unlike, say, machine learning – will also need to occur in near real-time.
This means that current efforts to leverage analytics that automate data center operations must accelerate. And quickly. Enterprises already require IT automation to manage everything from cyberthreat recognition and response to placing real-time offers in front of web shoppers. With IoT devices and IoT-generated workloads set to explode, data center automation will become an even more pressing need.
Finally, it’s worth thinking about how the IoT-driven future will affect application development and coding. Levine expects the “if/then/else” type of logical programming to give way to data-centric programming that relies more on mathematicians and data scientists than traditional coders. If true, CIOs may need to prioritize these newer skill sets to meet some of tomorrow’s programming needs.
Even so, continuous integration/continuous development (CI/CD), DevOps, and cloud will remain critical players in tomorrow’s IT and business world. If anything, development and deployment technologies, ranging from virtualization to containers and microservices, will become even more important as your company’s digital transformation embraces the IoT trend.
The shift to a more distributed, IoT-based computing environment means organizations, and their IT infrastructure and code, must become even more flexible and agile. At the core of these new environments will be what makes today’s best deployments – the ability to gather, manage, and share more data – and to gain more insight from that data in real-time. To deliver the best business outcomes in the future, CIOs and their teams must take full advantage of the tools and technologies already at their disposal.
Want to learn more?
To learn more about how Hitachi Data Systems can help you along your digital transformation to get the most out of your cloud investments – both for today’s needs and those of an IoT-enabled future – visit our Cloud Solutions page.