IoT on Google Cloud Platform

Google wants people to use its Cloud Platform for connecting and managing IoT devices through IoT Core and use other GCP components like BigQuery to analyze data produced by those devices. While these products are fantastic, they also have some real world challenges.

IoT Core provides a managed service for connecting IoT devices. It talks with both HTTP and MQTT protocols and features one-click integration with Cloud PubSub easing most of the infrastructure tasks. However there are some limitations:

1. You cannot use any random MQTT topic to send/receive messages as you would expect on a custom MQTT bridge. There are special topic formats to send messages and also to receive commands.
2. IoT Core uses Public-Private Key cryptography to secure devices. All IoT devices must first authenticate using the Public Key in a JWT token and then start sending and receiving messages.

While these may seem like reasonable restrictions, one has to keep in mind that hardware vendors are still stuck in the 90's. Many of them use outdated languages and are reluctant to change any aspect of their devices script to make them talk to IoT Core. Vendors also want to move a subscription model where they charge a monthly fee to access the MQTT bridge hosted by them where their devices would send messages. As such their reluctance to adapt to IoT Core is much higher. The only real option left is to create a component which bridges the gap between hardware vendor's MQTT bridge and IoT Core which is a bad idea for so many reasons. Google must release a pre-approved list of vendors whose devices are compatible with IoT Core so that they can increase the adoption of GCP.

The other problem that one faces while using GCP is in building the UI. Most of IoT customers want real-time graphs for their devices and while Google does have Data Studio, it is not meant for real-time streaming data. To create a custom solution, one has to hook directly into PubSub and plot the streaming data received from there. For historical graphs, Data Studio along with BigQuery might be a option but one has to be very wary of costs which can escalate quickly while using BigQuery. Also the only way to embed a DataStudio graph in a web page is to use IFrame which leaves a very bad user experience. Google must provide a managed service for building charts (real-time or historical) which should be much better than what they currently provide with DataStudio.

GCP is a fantastic platform for building IoT solutions and if Google listens to its customers and provides a fully managed end-to-end solution for most scenarios, it will be a game changer and real challenger to Azure and AWS IoT solutions.