Microsoft Build 2024, the annual developer conference, has just concluded, and it was packed with exciting announcements and updates. In the next few months, we should expect to see a new wave of AI-powered and integrated software solutions! (although that's not really a shock)
As someone who thrives in infrastructure, my focus isn’t on crafting AI solutions, but rather on creating the robust frameworks they’ll operate within. This is where Microsoft Build proves invaluable, offering a glimpse into the future.
Nathan McGee - Lead Cloud Consultant
Microsoft didn’t disappoint, delivering a plethora of updates to keep us infrastructure enthusiasts engaged. Here’s a snapshot of the resources I’ll be diving into:
- AI Reference Architecture - Who doesn’t love a good reference architecture?
- Service guidance - Great to see the updates on the Azure Machine Learning service, it really helps when trying to answer those design questions: "What's the best practice?"
- Cost-effective RAG - With RAG taking centre stage, seeing more options and increased capacity with cost in mind will make it much easier to deliver large-scale search services.
- Prompt shields - I've had some fun demonstrating some basic prompt injections, but the reality is that as we adopt more LLM-based solutions, this threat becomes very real. It's nice to see solutions to address and protect against it!
- AKS Automatic - AKS Automatic feels like halfway between AKS and ACA. It's always great to have different levels of service in our design pick list, and this feels like it might be a sweet spot for offloading Kubernetes management but still having access to the API.
- Dev Box updates - I know I'll be highlighting the "Dev Box hibernation" feature. It's an easy win for reducing cloud spend, whilst still keeping the Devs happy.
Amidst these easy wins, several intriguing discussions emerged, sparking thoughtful consideration:
- Copilot+ PCs - Using local hardware to run models means improved performance and security, and with that opens up great new potential. Will this bring the need for new hardware forward? Will off-loading AI model inferencing to Azure for incompatible hardware come at a cost?
- Microsoft Copilot in Azure - I saw some great demos of how we can use Copilot in Azure to easily understand the cost of resources and help guide us through unfamiliar tasks. It'll be great to explore how this can be incorporated into everyday ops, using Copilot to generate cost estimates before actioning changes, or detailing planned changes to support a change control process.
- GitHub Copilot for Azure Migrate - I can recall cloud adoptions that have stalled when they hit the "refactoring" stage. This is normally because refactoring software or data is expected to be a complex and time-consuming effort. I'm eager to see if GitHub Copilot for Azure Migrate will help with these estimations by answering simple questions like "Can I migrate this app to Azure?" or "What changes do I need to make to this code?"
- Azure deployment environments - Deployment environments is aiming to be a Platform Engineers best friend. And the latest announcements see it moving towards that goal. The added support for Pulumi is very welcome, opening up a simple way to define infrastructure using languages that Devs are more familiar with. I'm sure I'll be running a few demos of this with our developers, especially single-click deployment templates!
Finally, I'm a big fan of Satya Nadella's philosophy of continuous learning: Don't be a Know-it-all; be a Learn-it-all. Hence, I'm naturally drawn to tools that foster this mindset. One standout is Visual Studio Code for Education: a complimentary, no-setup coding platform that embodies the spirit of “just start learning”. Great work Microsoft!