IoT Mid-mile: The biggest challenge facing mass scale IoT deployment.

IoT Challenges: IoT companies building systems that incorporate the Things and Applications to handle the front-end and back-end of IoT are realizing that the challenges they face typically lie in the mid-mile. Fundamental questions that arise include: 1. Does all data from the thing need to be sent to the cloud before anything can be done with that data? 2. Does all data from the thing need to be sent to the cloud all the time? 3. What if there’s a network outage and I lose the network for a few minutes? 4. How secure is the data from the thing to the cloud? 5. What if I have multiple clouds the data needs to go to? 6. Can I run some of the analytics in the cloud at the edge itself? 7. Can I enable local service chaining at the thing layer itself? 8. How can I set policies to ensure only the appropriate people have access to the appropriate data? 9. What about QOS issues? 10. Do I need an agent for every single device in the Things layer? 11. What about all the different communication protocols that things use – do I need build a translator for every single protocol? 12. What about the fact that some Things use ZigBee, while other may use zeewave and other 802.11 and yet others 3g/4g or satellite? They are finding that ensuring security, data privacy, data efficiency, QoS, scalability, availability, interoperability, integration, and low TCO (Total cost of ownership) requires significant additional expertise and investment for each custom IoT application. Limitation of Current IoT platforms The biggest limitation of current IoT platforms is that they force a centralized model - all Things are assumed to be dumb data producers that just push data to the cloud where all the intelligence lives. This approach suffers from several flaws. • Legacy Devices: The vast majority of Things are low cost, low power, low duty cycle devices that are not designed to directly connect to the internet. For example, most Things may not have enough capability to support strong security mechanisms. Also, 85% of Things are estimated to be legacy devices that cannot be updated (for example to run an agent) to connect to the internet. • Data Volume: Although each Thing may send just a small amount of data, given that the number of Things that will connect to the cloud is expected to reach several billions, the huge volume of small data will likely overwhelm the network infrastructure. • Real-Time Actions: Time-critical applications that require real-time analysis of collected data, real-time decisions & actions, will be extremely hard to support if supported at all. • TCO: Since all the data has to be collected, processed, stored, analyzed, and acted-upon from the cloud, the costs associated with network bandwidth, cloud compute, memory, and storage will push the TCO even higher. • Network Outages: The Things may be deployed in locations where internet connectivity is poor or intermittent. With a cloud-centric approach, this would mean critical data from the Things may be lost or even worse; the Things may get crippled without internet connectivity. • Complex Customization: Another major issue with IoT platforms in the market today is that many platforms require significant and complex customizations to support new sets of Things and Applications. This is because these platforms were never purpose-built as extensible middleware platforms. Rather, they were originally built as a custom IoT vertical solution tightly bound to specific Things and Applications; as an afterthought the middleware was packaged as a platform.