Updated: May 24, 2019
Say "Internet of Things" and the second phrase that typically follows is "cloud."
Cloud computing is great for IT applications, e.g. your HR software. Your HR doesn't care what's on the other side of her browser so long as she can create the new hire records. The HR software can be running on a local server, desktop or in the cloud on another continent. Once your HR person finishes creating new hires in the system there isn't much data that goes into the application.
However, the scenario is very different with IoT. Unlike IT applications that predominantly interface with human users, IoT applications interface with machines that generate data in much higher orders of magnitude compared to their IT counterparts. Many companies are realizing that relying too heavily on the cloud when it comes to operating and maintaining an IoT project in the cloud at production scale can be expensive and inefficient.
Storing All Your Data in the Cloud is Expensive and Inefficient
The average person creates less than one GB of data per day between work and personal activities - and most of us average 8 hours of rest out of a 24 hour cycle. In comparison, one connected car generates on average 25 GB of data per hour. Needless to say, unlike people, industrial machines - many of which are "on" 24/7 -- generate a lot of data. Further, unlike people, machines cannot differentiate between what data is important or actionable without programming or human intervention.
A recent Forrester report found that up to three-fourths of all data within an enterprise goes unused. It doesn't take much to figure out that sending every temperature or on/off reading from a sensor or machine to the cloud and paying for every byte transferred and stored isn't a common sense strategy. Further, while cloud-scale storage systems give you the notion that storage space is virtually unlimited and cheap, transferring all the data from all your machines and sensors into the cloud for analysis and/or storage can get real expensive and inefficient very fast.
A More Efficient Cloud Starts With Managing Machine Data Intelligently
If not all data needs to be in the cloud, then it makes sense that managing machine data should occur closest to the machines that generate the data. Bringing data management intelligence into the network edge allows IoT solution architects and device design engineers to steer the right data to the right applications while enforcing data retention policies. The bottom line is faster time to deployment, reduced operational costs and improved ROI.
Here are three issues to consider when building your data management solution
Decide what data needs to go where. What data can stay local and what data needs to be fed into a cloud-based application?
Implement various compression methods for cloud-bound data. Depending on the type of data and IoT application, there are various methods for more efficiently compressing or consolidating data that is cloud-bound.
Establish automated storage mechanisms for managing different datasets in the cloud. Not all data needs to be stored the same way or for the same amount of time once it gets to the cloud. Depending on the type of data and your company's data retention strategy, data in the cloud may be stored in a time series, NoSQL or SQL database.
Here at Machinechat, we're working on the next generation of hardware-agnostic intelligent data management solutions that will enable IoT solution developers and architects to reduce man-years of development and eliminate complexity in deploying IoT projects successfully into the enterprise and Cross the IoT Chasm.
To keep up to date on how we plan to accelerate connecting the next billion IoT devices, sign up for our mailing list here. Like what you read? Controlling the clout and cost of the cloud is just one of three IoT development trends that will impact 2019.