There was a time when people laughed at Amazon’s cloud offerings and scoffed at the thought of centralization. Fast-forward to now, and everyone’s stumbling over themselves to start public clouds. What’s interesting is the disruptors could officially be the disrupted if Peter Levine, a venture capitalist and partner at Andreessen Horowitz, knows what he’s talking about. Big Data and cloud computing have played a major part in advancing AI technology, but could it be at the cost of their own very existence? That particular point of view may be a bit farfetched, especially since a lot of the IoT is made up of small devices that are built for singular purposes. The whole advantage of having the cloud to do the heavy processing means IoT devices can be built light and cheap.
A good example of cheap and simple IoT hardware is the Echo Dot from Amazon. The Echo Dot is a hands-free, voice-controlled device with a built-in speaker that lets you control everything from your bedside lamp to your thermostat with just the sound of your voice. In fact, the use of Amazon Echo with IoT hardware has been so successful that some think the next IoT platform may be based on voice recognition.
PaaS and the IOT
The IoT already outnumbers humans in a big way if you include smartphones. To properly manage your IoT resources from the cloud, a PaaS solution is probably the best option, especially for beginners. Since managing IoT devices often involves rapidly deploying customized software, an SaaS solution can be bit stifling while an IaaS comes with a lot of unnecessary tasks like managing middleware and operating systems.
Another reason for the enterprise to choose PaaS is that it’s not easy to maintain and manage thousands of IoT devices that are almost always producing data of some sort. When your biggest asset is your data, it’s good to have control over it, and that’s another thing you forfeit when you choose an SaaS solution. Unlike Amazon’s cloud-centric approach where all the processing is done in the cloud, another approach is to analyze IoT data at the source and send only specific data to the cloud for processing.
Another thing is that the IoT is a mixture of all kinds of different hardware devices with IP addresses, and while a lot of it is cheap hardware, some if it can cost millions of dollars like drones and hospital equipment. In such situations, processing is done as close to the source as possible to minimize the risk involved. In this hybrid mix of cloud and edge computing that many are referring to as “fog computing,” PaaS plays an even more important role as this requires the combination of datacenter technologies with much more constrained devices.
Living on the edge
For those devices that need to behave and react in time-critical circumstances like drones, driverless cars, or hospital equipment, they need to have their engines on board rather than in the cloud. Man might have reached the moon but commercial optical transmission systems are only capable of single-channel data rates up to 100GB per second. That’s probably why cloud giant AWS had to take to physically transporting datacenters to the cloud, no matter how ridiculous that sounds.
Edge computing is all about keeping that processing power on the edge of the system, or as close to the source as possible rather than at the center. The advantage here is you can effectively make use of a large network of hardware devices for your computing needs rather than depend on servers that are centralized. If you count smartphones and all the IoT devices out there, you’re talking about a massive army of resources that if tapped correctly, could possibly negate the need for cloud processing. That’s not to say that the cloud has any chances of disappearing just yet because the ease of use it offers is too valuable right now. Pooling and centralizing digital resources have made hosting modern apps and websites in the cloud possible for absolutely no cost and the services make it possible for small teams to do big things.
A bit of this, and a bot of that
There has probably never been a point of time in the enterprise where there are just so many new technologies bombarding it from every side. With cloud computing, DevOps, containers, machine learning, and the ever growing IoT, it’s almost like they are all pieces to a bigger puzzle that we’re supposed to somehow solve. The future will probably look a lot less cloud dependent especially since smartphones and IoT devices are getting more and more powerful. The amount of processing power that cars, boats, airplanes, and almost all equipment are going to possess in the near future would effectively make them small datacenters in themselves.
Not to mention that since containers suspiciously enable us to run lots of cheap hardware as opposed to high-powered equipment, the IoT might just be the ideal place to run containers as well. Edge cloud requirements like cost efficiency, low power consumption, and robustness can be met by implementing container and cluster technology on small single-board devices. Again, PaaS comes into the picture with regards to application packaging and orchestration and a container-based edge cloud running PaaS architecture may not be a bad idea.
Outgrowing the clouds
The biggest factor here, however, is that the IoT is probably outgrowing the cloud. With millions of sensors all over the world consuming terabytes of data each, it will be physically impossible for the cloud to keep up. The very fact that AWS called their data-transfer unit Snowball Edge is probably some level of acknowledgement that all computing cannot happen in the cloud. In fact if we go by the popular predictions of how many IoT devices we will be looking at by 2020, that’s probably enough to crash the cloud many times over. Where cloud computing works best is when all the data is in the cloud; when data starts pouring in from the billions of sensors across the planet, we’re probably going to see a different cloud that’s built to deal with armies of IoT devices.
From heart-rate monitors, to biochip transponders, to satellites and CCTV cameras, the IoT is this monster army that just keeps recruiting and growing. Anything with an IP address that can send data across a network is effectively part of the IoT. With the recent IPv6 address space expansion, we can potentially address every atom on the planet Earth and still have enough IP addresses for another 100 planets. Cameras are notorious for capturing large amounts of data, and when that data is perishable, it doesn’t make any sense to send it to the cloud for processing. Perishable data is data that will be no use after a certain time. In the example of self-driving cars, a stop sign needs to be identified immediately. If the car is travelling at, say, over a hundred miles per hour, identifying a stop sign needs to be done “on-premises” and in a hurry.
Clear skies ahead
The increase in mobile computing, decrease in cost of hardware, and sheer number of networked devices make the IoT what it is today. If it’s putting a strain on resources now, then once the real hardware rolls out, it will be a nightmare. There’s all this talk about self-driving cars and how even regular luxury cars have about a 100 CPUs. A lesser known fact is that Microsoft and Airbus are working together on the first pilotless flying taxis that should be in working prototype mode by the end of this year. Can you imagine how many CPUs those things will have or how many terabytes of data each of the sensors on those “taxis” are going to generate? And can you imagine how much data a fleet of those flying taxis would generate? Judging by the amount of hardware that Microsoft and Airbus are going to put in them, that’s essentially a fleet of flying datacenters. Literally in the clouds.
The future will probably see intelligent devices sitting on the edge of the hybrid network dealing with time-sensitive material while the rest will probably be sent to the cloud for analysis and storage. If the cloud does one day go obsolete, hopefully we’ll all be flying around in private cars like the Jetsons without a care in the world.
Photo credit: Flickr / Francois Schnell