Operations | Monitoring | ITSM | DevOps | Cloud

Previous articles walked through RDMA (Remote Direct Memory Access) as a programming model and InfiniBand as the fabric that was built around it. Both led to the same conclusion, even if it was never stated outright: moving data, not compute, becomes the bottleneck once systems scale. So what happens when you want RDMA, but you’re already running an Ethernet network you’re not keen to replace? That’s usually where RDMA over Converged Ethernet (RoCE) enters the conversation.

Tokens per watt (TpW) – the measure of useful AI work produced per watt of energy consumed – is the metric at top of mind for CEOs, heads of AI, and infrastructure teams alike. With the tremendous cost of GPU clusters, extracting as much value as possible from the expense is critical. But in the pursuit of tokens, it’s important to remember that hardware efficiency isn’t the only factor influencing data center operating costs, or the output of useful, revenue-generating AI work.

Welcome to this blog series which explores innovative uses of Ubuntu Core. Throughout this series, Canonical’s Engineers will show what you can build with our releases, highlighting the features and tools available to you. In this blog, Asa Mirzaieva, engineer from the Silicon Alliances team, will show you how to deploy optimised AI models on Renesas RZ/V series hardware using the Dynamically Reconfigurable Processor for AI (DRP-AI).

One of the important offerings of the RISC-V Instruction Set Architecture (ISA) is the ability to customize and extend the base instruction set. An initial reaction to hearing this is often to worry about software portability and compatibility, since if every RISC-V CPU offers a slightly different set of instructions, software won’t be portable.

When distributed workloads stall because nodes cannot exchange small messages quickly and consistently, the network is the limiting factor. How do you solve that problem? InfiniBand offers one solution. InfiniBand is an interconnect, meaning the end-to-end communication system that links compute, storage, and accelerator nodes. It is implemented as a purpose-built network fabric, the switching and transport layer engineered to deliver high bandwidth and low, predictable latency between those nodes.

Some support cases are straightforward. Others lead deep into legacy code, where a single logic bug can quietly turn a routine command into a major performance problem. This series looks at how Canonical Support and Sustaining Engineering work together to investigate, patch, and upstream difficult issues that standard troubleshooting alone cannot solve.

Today, Canonical announced the release of Workshop, a solution for launching development environments with a single command. These environments are configured once, and can be reproduced on different machines. This means consistent workflows across development machines and deployment pipelines, and less time managing dependencies.

Open source thrives on engineering-driven processes. Fast feedback loops, terminal tools, Git workflows: they’re the lifeblood of how we build software in the open. But for software to truly excel, we need to create user experiences that empower people to use them. I wanted to bring this conversation into the spotlight as part of Canonical’s Open Design initiatives. What better way than at FOSS Backstage 2026 Berlin?

Canonical, the publisher of Ubuntu, today announced the general availability (GA) of Managed Kubeflow on the Microsoft Azure Marketplace. This solution enables AI teams to get a fully managed, production-ready MLOps platform in their own tenant. Upstream Kubeflow is a powerful tool for machine learning, but it remains notoriously challenging to deploy and maintain.

I’ve yet to meet a developer that enjoys working with metered AI APIs. The need to pay for every API call in development works in direct opposition to the ethos of rapid iteration, and it’s easy for the costs to get out of hand. That’s why Canonical has created a different approach to building AI-powered applications; one where the model lives inside your app, not behind a pay-per-token HTTP call.