Revenir au blog

Follow and Subscribe

Disponible uniquement en anglais

Cette page n'est actuellement disponible qu'en anglais. Nous nous excusons pour la gêne occasionnée, merci de revenir sur cette page ultérieurement.

Apps That Shouldn’t Be Built at the Edge | Fastly

Sean Leach

Chief Product Architect, Fastly

Until recently, many developers were curious (if not skeptical) about the power of edge cloud platforms, but hesitant about building applications and migrating complex logic there. But as innovators and early adopters from a variety of industries have harnessed the edge’s computing power, cost benefits, and security features — offering up incredible customer experiences as signs of success — one fact is becoming clear: for a more performant, secure web, nearly everything can and should be built at the edge.

To back up this bold claim, we originally set out to write a blog post about the types of apps that are best built at the edge. But as our list grew to more than 50 use cases, it made much more sense to flip the perspective and center our story around the (much shorter) list of apps NOT to build there.

Solving for big state

To be clear, the edge should not replace the central cloud completely — there are still important computing jobs to be done there. As a general rule, this includes any app that requires big state:

Big persistent data stores: Databases like Oracle or MySQL that persist lots of information for a long time are good candidates for the central cloud, where it’s easier to maintain their consistency. Other legacy data sources may still be confined to corporate data centers and aren’t even candidates for moving to the central cloud.

Legacy applications: Often built to expect direct access to centralized persistent data stores, legacy applications can be difficult to move to the edge wholesale. But they can benefit from gradually converting components to microservices, abstracting their data access behind APIs, and moving them to the edge piece-by-piece. This hybrid approach can have enormous benefits in terms of performance and security while avoiding the need to completely rewrite an app.

Machine learning training: Requiring enormous datasets, long-running training jobs, and specialized processors like GPUs, training machine learning models is well-suited to the central cloud. Once they are trained however, the edge is a great place to deploy models for high-performance inference because it combines low-latency access with the compute power necessary to run the models quickly.

Taking small, smart steps to the edge

So what if your app wasn’t built at the edge, but doesn’t fall into any of the cases outlined above? Here are some practical ways you can begin moving parts and pieces of your apps to the edge.

  • Microservices: As we saw above, separating a legacy monolith app into smaller services can allow you to move select components of your app to the edge over time. Once those components communicate via APIs you can take advantage of edge technologies like URL routing, content modification, and logging. We have a great technical brief on using Fastly with microservices.

  • JAMStack APIs: JAMStack sites work hard to move as many of their components to the edge as possible, but they often still go back to the central cloud to perform API calls. Those calls may run in multiple clouds and on origins located at different providers. Building custom JAMStack APIs that run at the edge and using them to stitch together content from central cloud APIs provides a scalable and highly performant way to reduce latency for clients. Even better, the content retrieved from those central cloud APIs is often cacheable at the edge, saving round trips, which helps from a performance perspective (fewer network calls) and from a cost perspective (third party APIs often charge on a per-query basis), so fronting JAMStack APIs at the edge is not only faster but cheaper as well.

  • Security auditing and enforcement: The edge provides a single, logical ingress path into all of a customer’s applications. This makes it a natural place to put security auditing and enforcement to ensure that all applications are covered and don’t have to be rebuilt individually as new applications are deployed. Customers can use the edge to audit all requests to any of their applications using real-time logging, as well as provide universal and consistent security enforcement for protection against the latest web application and API threats.

The most solid foundation for modern apps

The edge is now a powerful starting point for building innovative applications. And, as customer expectations of speed, security, accuracy, and personalized experiences continue to grow, so too will the importance and necessity of the edge.

If you’re not yet convinced, we’ll happily walk you through the capabilities of our platform and global network, case studies that apply directly to your business — even those 50+ use cases mentioned above. Join us at the edge. We’re here to help.