Fabcon Vegas 2025 - The main keynote announcements
After last nights keynote, we've so many announcements and new features to digest that it's going to take a few blog posts to get through everything. Given that I'm going to split the announcements to cover: Keynote, Power BI, Fabric, Purview, and preview features.
Let's get started. For this blog, if you have a specific feature you want to know more about do have a look at Arun's blog. Unless it's a really major preview announcement, I'm going to focus on features that have gone GA - the reason, without doing a risk/reward evaluation, I personaly take the view that preview features shouldn't be used in production. That means for the majority, the preview features need to be considered in future roadmap plans but are unlikely to be used today.
Up first, copilot.
Co-pilot
The big announcement in this space is that Copilot is now available at all SKU levels! This is a big one for the SME sector that couldn't justify the cost of an F64 SKU. Now you too can access the time savings that copilot brings today - and some of the frankly increadible stuff that's currently under wraps.
The down side, I haven't seen any announcement about an increase in the number of data centres that Azure AI is available within. For those in the EU, that means your grounding data still has to go to Paris - and for the UK, that means you still have to enable sending data outside of your region and risk it going to the US (ah Brexit, the gift that keeps on giving). Whilst the increase in availability is very much welcome, Microsoft need to be treating the UK as part of Europe from a data prespective given our equivalency agreement is still in place - we really shouldn't be exposed to the risk of data going to the US.
In other copilot related news, we've got an overhaul of the integration within Notebooks. With this update, we're starting to see this version of copilot being more useful as a tool to improve productivity potentially - especially as we no longer need to remember to initiate copilot each time and a history of interactions is maintained.
One Lake security
Another big update for the platform. Whilst it isn't quite in public preview yet, and so will be a while before it is recommended for Production use, this one is big enough to be aware of today and consider undertaking a review as to if the risk of using a public preview feature is enough of a reward for your situation.
Today, if we have specific security requirements, they have to be replicated in each Fabric item (e.g. Lakehouse, semantic model, etc). It's a pretty time consuming process, even if we use Entra groups to help make this easier to manage.
With OneLake security, this becomes a lot easier. We can define the security model down at the file level and it's respected by all Fabric objects. At the moment, we can't use the context of the logged in user to define DRLS and a few other restrictions are in place - but this is definately going to be the way to go to define security models in the future.
Comments
Post a Comment