With the Internet of Things now a reality in auto and other industries, companies are finding themselves remarkably changed–and challenged–by the availability and analysis of data from a multitude of sensors. Companies like Tesla are taking the lead with “over-the-air” remote maintenance, driverless technology and the connected infrastructures. These IoT advancements mark a growing need for a new database approach that offers both scalability for high-transaction workloads, and total data accuracy for high-value workloads. With the right approach, we may see the doors blown off the current pace of innovation–from redefining entire industries, to making commercial space travel feasible. The wrong approach could cause serious setbacks.
So far, the approach to scaling databases that many companies have taken involves modifying existing database structures like MySQL or NoSQL, but these are really only temporary solutions, at best. NoSQL is OK for analyzing market trends, but doesn’t offer the necessary accuracy for many IoT applications, while MySQL offers the accuracy but not the scalability, and modifying either of these creates problems down the road. Much-discussed surges in data volumes, live analysis and the need for high availability of data will soon demand profound changes of approach.
Current thinking is that a cloud-centric drop-in replacement for MySQL, like Aurora, will solve problems of scalability. What many don’t realize, however, is that Aurora and other comparable solutions are essentially built to scale in the cloud the same way that their predecessors scaled in the data center, and present a very real performance ceiling. They can’t “scale out” like other cloud applications, and you’ve got to use old-school, problematic methods to achieve performance increases.
In the IoT realm, having low standards for database infrastructure just doesn’t make sense. Nasty things can happen with traditional database architectures when they need to be pushed beyond what they were built for, because they, and the applications that run on them, have to be seriously modified. Ultimately, companies will require a form of SQL database that’s engineered from the ground-up for very high scalability. So why not move away from “traditional” thinking altogether and go with a big picture database to backup your big picture plans.
Planning for the Big Picture IoT
With the current mainstream eye fixated on household items like the Nest Thermostat, it’s easy to lose sight of how big the implications of IoT really are. Tesla’s remote maintenance provides just one example of a large scale IoT. We’ve only seen the start with driverless vehicles. Imagine, in the near future, a driverless car network being coordinated from a centralized application; this would require a combination of high throughput processing and analytics that we haven’t even fully envisioned yet. Done well, such a system would allow cars to move through the city like your blood moves through a healthy body, in perfect rhythm. But done wrong, with poorly engineered products and system performance issues, and you have the equivalent pain of millions of small heart attacks!
These “super applications” will undoubtedly involve much richer data sources for accuracy and detail. They’ll produce a massive stream of sensor data that must be ingested and organized for analysis, requiring large amounts of storage and computing capacity. Integration of large unstructured data sources like high definition video and images will also be a big part of the IoT story. Furthermore, requirements to make data available to multiple applications simultaneously will compound the need for the ability to rapidly increase or decrease database capacity.
This need for accuracy combined with scalability means that traditional solutions won’t be an option. Does this mean that the IoT-enabled future we envision is really just science fiction? Not at all. We’ve already “cracked the code” so to speak with ClustrixDB, the first relational database that scales out, allowing you to increase performance indefinitely by merely adding nodes and flex capacity up or down according to need; meanwhile ensuring data reliability at all levels of performance (it’s required to know how to complete IT asset disposition even). Scalable SQL is the only course that meets demands for both scale and accuracy, and it’s the direction that we must take to get into the IoT fast lane. The well-traveled MySQL, NoSQL road leads only to a traffic jam, figuratively….but perhaps also literally if it’s applied to a technology like driverless vehicles.