Building an IoT Supercomputer

Computes.io is now a distributed supercomputer mesh network (more details). Each node on the mesh network acts as both a core and kernel (more details) — meaning that each node can perform Javascript computations for other machines as well as request computations from itself and other machines.

Computes.io uses IPFS’ pubsub capabilities to create a signaling and messaging engine for distributing computations across the mesh network to any machine with available resources (more details). The new messaging interface supports broadcasts as well as private messages to other nodes. IPFS pubsub is similar to the MQTT pubsub protocols used in most Internet of Things (IoT) applications and devices but with one major difference. IPFS does not require a dedicated message broker to store and forward IoT messages. IPFS nodes form a mesh network and can communicate with or without the Internet — no message brokers required!

Here’s how the computes.io messaging engine works:

Once connected to computes.io (job.on(“ready”) event), your node is immediately available for processing remote computational requests. job.message is used to send/broadcast messages to all nodes on the mesh network via the domain key or to a specific computes.io node via their IPFS-generated node id. Your node will fire a job.on(“message”) event when it receives either a broadcast or private message.

Now that we have the messaging basics covered, let’s use Johnny-Five to toggle an Arduino’s (running Firmata) LED on and off via computes.io messaging! We live-coded this historic demo (first IoT message on computes.io) at Monday night’s IoTPHX meet-up. Here’s a photo of the team involved.

Here’s a short video demo of IoT messaging in action on computes.io.

Here’s the source code used in the video above.

We believe that distributed, realtime computes along with IoT messaging all delivered in a mesh network is game changing. In addition to collecting/streaming sensor data and actuating devices, these machines will be able to apply machine learning and automations in the mesh network utilizing any other machine with the capacity to assist — all acting as one massively scalable, worldwide IoT supercomputer!