In advance of Microsoft’s announcement of the general availability of its Event Grid offering today, I took the opportunity to chew the proverbial fat with Corey Sanders, Microsoft’s head of all things container and serverless. The briefing was in part chance to give me a preview of the news, but also to elicit my broader thoughts on the direction of products such as this.

First up, let’s take a look at the GA announcement and what has changed from Event Grid in its preview days. In describing the thinking of Event Grid, Sanders explains the trends occurring in the world of enterprise it. As he explains, modern applications are moving away from monolithic architectures and instead using a set of distinct services, all working together. These services cover the gamut of what vendors offer, from foundational services offered by a cloud platform like Azure (Database, Storage, IoT, Compute, Serverless Functions, etc.) to application-specific services (inventory management, payment services, manufacturing processes, mobile experiences, etc.)

As we move to a far more modular and microservices view of the world, event-driven paradigms, the idea that services can communicate via a simple and consistent mechanism, become important. The term Event covers a myriad of situations and could include IoT device updates, cloud provisioning, storage blob creation, through to custom scenarios such as new users being added to HR systems.

Which is where Event Grid comes in. The product is, in my overly-simplistic terms, a virtual train station that reconciles arriving trains (the triggers) with departing passengers and their intended destination (the actions, and the external systems upon which those actions will occur.) Instead of continuously polling for changes, Event Grid is triggered on the arrival of the particular trigger in question, and customers only pay for the events in question.

Event grid also offers a fabric to reconcile horizontal heterogeneity and allows users to react to their particular events using Serverless offerings such as Azure Functions or Azure Logic Apps, using Azure Automation, or even writing custom webhooks for their code or third party services. What this means is that any service running anywhere can publish and subscribe to Azure Events.

Uptake while in preview

Sanders told me that, since its preview release, Event Grid has grown to the point where it currently handles 300 million events per week and has seen a range of use cases, from Outotec using Event Grid to rearchitect their hybrid integration platform to Paycore using Event Grid to unify the various different human capital management systems they use.

Broadening the franchise

As is often the case when a preview product makes its way into GA, Microsoft is offering new features for Event Grid. First up, as expected from a service that will now see usage in mission-critical settings, Event Grid will be available in a broader number of regions – specifically West US, East US, West US 2, East US 2, West Central US, Central US, West Europe, North Europe, Southeast Asia, and East Asia with reportedly more coming soon.

On top of that, Microsoft is offering more of its own third-party offerings on the Event Grid platform – General Purpose Storage, IoT Hub and Event Hub, for example. In what will be attractive for those looking to integrate different third-party offerings onto Event Grid, Microsoft is releasing new SDKs – Python, .Net, Node.js, GO and Ruby.

In an interesting change, the new Event Grid SLA includes a 24-hour retry policy for event triggers. This is of particular interest and is a reflection on the fact that, in the real world, connectivity isn’t always available. Especially in IoT use cases, the ability to have a remote device trigger an event, even when that device drops in and out of connectivity is useful.


I’ve been excited about Event Grid for some time. I see it as a broadly strategic move for Microsoft and one which allows it to build a strong business of the back of new approaches to building applications, but also one which provides it with a strong story around neutrality. In simplistic terms, I’ve always seen Event Grid as an infrastructural analog for Zapier or IFTTT, a service that allows triggers for anywhere and anyone, to create actions that occur anywhere or on any product or service. A single place for both ends of the event lifecycle.

Sanders indicated that, during the preview, there was more usage of external systems for Event Grid than he would have expected. This speaks to, in my view, the real value that a neutral event fabric offers. While the various vendors will do their utmost to encourage people to use their own proprietary systems, offering a fabric that welcomes all comers is a smart move, allowing Microsoft to play in both the open fabric space and the more propriety one as well. Personally, I’m fascinated to see how this use case expands, and am looking forward to talking with customers who are, for example, using Event Grid as the fabric to connect triggers from non-Microsoft applications, to action on non-Microsoft services.

Sanders and I have an ongoing joke (and, this is the technology industry, so apologies in advance that the joke isn’t overly funny) about Microsoft’s strategy of creating lock-in by love. That strategy is shown in Event Grid, there is nothing keeping people on the fabric other than the real value of connecting all of the different things they care about.

Event Grid is a smart approach, and I have some strong views on how it can be further extended to continue the theme of this new, more open, Microsoft. Interesting times…

Ben Kepes

Ben Kepes is a technology evangelist, an investor, a commentator and a business adviser. Ben covers the convergence of technology, mobile, ubiquity and agility, all enabled by the Cloud. His areas of interest extend to enterprise software, software integration, financial/accounting software, platforms and infrastructure as well as articulating technology simply for everyday users.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.