Cloud-Only IoT: Stepping into a Trap Open page
August 21, 2016
Amazon, Microsoft, IBM and other vendors are all dictating that enterprises should build IoT solutions leveraging cloud-based macro processing across gargantuan quantities of data. While those vendors are telling feel good outcomes about curing cancer with magical algorithms, they are leading companies down an immature, unproven and high-risk path.
Cloud-Only IoT solutions quickly run into three trouble areas:
If you are investing in an IoT solution your expectations will be that it meets certain performance metrics and responsiveness. The reality is that data traveling all the way to the cloud and back can be an extremely timely exercise. During the mobile wave we had an ideal benchmark of 250ms for every transaction. That target is now closer to 100ms for IoT as users expect things to respond immediately. The idea that a sensor relaying data across a wireless protocol, then to a gateway, then to a cloud where it bounces from an API manager to a cloud function, to a big database and all the way back in under 100ms is extremely optimistic.
It is more likely the request will take over 10–15 seconds and during that time:
- A tired contractor walked through a hazardous asbestos zone
- A lucky thief drove away with a full haul
- A gasping patient urgently called for help
- A panicked guest can’t find their child the theme park
- A hurried carpenter loses his finger in a band saw
Ultimately if a solution doesn’t solve these challenges, it fails to deliver the full promise and potential of IoT.
IoT solutions are fundamentally built on an active connection to the Internet, but the Internet is not always consistent and available. We must have other core capabilities to guarantee reliable connectivity and uptime. Whether it’s an ISP not delivering the quality of service expected, a cloud vendor having scheduled/unscheduled outages, or an internal networking issue, internet delays and outages are a fact of life for everyone. The consequences of using cloud-only IoT solutions can mean:
- A factory stops production
- A warehouse doesn’t ship any products
- A home owner’s doors won’t unlock
- A hospital postpones a surgery
IoT solutions must have the ability to run disconnected and independent of the Cloud vendors. Failure to build offline architectures results in financial losses, and real lives at risk.
Every IoT analyst is predicting there will be billions upon billions of connected things in the coming years. The IoT cloud platforms today have a transactional, easy to adopt but daunting at scale model for pricing. Each message costs you a little bit of money and in the beginning your solution costs only pennies. Sadly, if you achieve success, the micro costs quickly become exorbitant, far exceeding the cost of dedicated infrastructure or traditional licensing. This means that:
- Maintaining the temperature in your buildings will outweigh energy savings
- Tracking your warehouse inventory will cost more than the real-time benefit
- Hiring an employee to control access to a pharmacy will be cheaper than automated monitoring
The biggest risk to the success of IoT solutions comes in the form of failed business models. Nothing exposes a flawed business strategy more than an unsustainable cost model.
The cloud should be part of most IoT solutions that get created. Amazon, GE, IBM, Microsoft, Google are all adding value that should be leveraged. Unfortunately the tools and models they offer today box companies into their specific cloud solutions and lead them down a path of failure. Without additional vendor support and the addition of other 3rd party vendor solutions for edge computing, these early cloud-only solutions are doomed to fail.
IoT vendors that include gateway/edge based solutions, for real-time actionable events will be required for IoT Solutions to succeed from adoption to scale. Vendors like ClearBlade are addressing this challenge and offering customers real solutions today.
For you to realize the value of IoT, you will need more than just another Cloud vendor.