Analytics and The Cloud of Things

The Internet of Things has changed data analytics forever. The torrent of machine data spewing from sensors and devices presents challenges of variety, volume and timeliness. The IoT data river contains gold, if only it can be mined. Many organizations just aren’t equipped and let the data flow away. The cloud offers the solution: elastic storage to manage the volume of IoT data, and elastic compute to analyze it. This is why operations teams need ‘Cloud of Things analytics.’

Let’s take a cable TV provider with hundreds of thousands of customers. It streams content to smart set-top boxes which report back their status and other operational data. The provider wants to know the condition of each box to improve customer service, so each unit pings the hub, which collects the data. Some basic analytics will provide useful troubleshooting information—if a particular model has an issue with a firmware upgrade, for example. A more advanced platform might use predictive analytics to swap out set-top boxes that might soon fail. Bingo! Happy customers and reduced costs.

The reality, though, is that much of the data in this situation is wasted. When a device fails a status check, it floods the network with alerts. “Help me, help me!” it cries until the issue is resolved. If multiple devices fail, the result is a cacophony of alerts leading to further performance degradation. Without a cloud-based analytics system that can scale-up elastically to handle these intense workloads, there are two options. First, the analytics system is overwhelmed, leaving the company flying blind. More commonly, the data stream is ignored or downsampled, again giving only an occluded view of the operational reality. Both affect performance, time to resolution and customer satisfaction. That’s why cloud analytics is so important in the age of IoT. Yesterday’s on-premises analytics weren’t designed to handle today’s streaming data.

Storage is another driver for Cloud of Things analytics. Let’s go back to our hypothetical cable TV company. It wants to gather anonymized usage and performance data from across its network of set-top boxes. The data will show not just time of day, service and channel selections but also data streams about power consumption, bandwidth and traffic types. Cross-correlating these data, which are all time-stamped and coming from a known array of devices, should provide valuable historical and predictive analytics. And in fact that is the case; the data can inform content development decisions, pricing and support staffing levels. But there is so much data, the cable provider can only store the records for the past two weeks. After that it overwrites the old records with fresh logs. The cost of storing and maintaining the data on-premises is hard to justify. This leaves the company with a short-term memory problem.

A better solution would be to move the analytics and storage into the cloud. There economies of scale are far greater and cost per gigabyte of storage plummets, making data more viable to retain. Now the cable company can see further back, to last quarter and even last year for deeper trends. Once again, a private datacenter isn’t able to handle the data volume IoT sensors and devices generate.

Streaming data from IoT devices is valuable but problematic. Each data point in itself is all but worthless. But aggregating each pixel of data draws a sharper picture. The challenge is that you can’t watch a 4K movie on a CRT screen. Operations teams recognize this, but implementing Cloud of Things analytics isn’t as simple as reinstalling the platform on a cloud-based server. There are provisioning, application delivery, security, regulatory and access challenges to contend with. That’s what we’ll be looking at in the next post in our Cloud of Things analytics series.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>