How an Active Storage can Democratize Artificial Intelligence

Artificial Intelligence (AI) is no longer the exclusive domain of powerful supercomputers and humongous datacenters. From smart thermostats to wearable health trackers, AI is quietly powering more and more of our everyday devices. But as intelligent technology spreads, so does the challenge of making it efficient, accessible, and sustainable — especially in devices that aren’t exactly powerhouses.

Traditionally, most AI tasks --like training models or crunching big data-- happen in the cloud. But moving massive amounts of data back and forth between your device and a distant server isn’t always ideal. It eats up bandwidth, slows down performance, and raises concerns about privacy and energy use.

From the ICOS project we strive to develop a meta operating system for the cloud-edge-IoT continuum. In these infrastructures, it is crucial to efficiently use resources and be able to run artificial intelligence workloads across the continuum. Active storage systems are a key enabling technology that seamlessly introduces data processing capabilities within the continuum (through a distributed storage system).

Don't disturb, AI Domain Specialist at work

An excellent AI workload starts from the beginning, with good ideas and small scale testing and exploration.

In this environment, the expert will make use of devices, use GPU/TPU devices and experiment. However, they will eventually need to step up to a distributed environment.

Scaling up an AI application is a daunting task. Our proposal is part of the BSC software catalog (and component of ICOS): dataClay, a distributed active object store. With this software, applications objects and application logic can be distributed across the computing continuum while maintaining the programming model.

This needs (almost) no changes on the application, and it supports all the libraries. But with this step, we have obtained a distributed application! This results in an agile environment for efficient AIOps on the continuum.

The Future of AI is distributed

By using active storage and supporting AI across the continuum, we're making intelligent technology faster, more sustainable, and available to everyone --not just those with access to massive cloud infrastructure.

At ICOS, we believe that AI should work wherever you are, with whatever devices you have. And thanks to tools like dataClay, we’re getting closer to that vision every day.