How microservices change the role of the network in next-generation application architectures.
Microservices, for the uninitiated, are essentially the decomposition of applications into multiple services. This decomposition is often based on functional lines, with related functions being grouped together into a service. While this may sound a like SOA, it really isn't, especially given that SOA was an object-centered methodology that focused on creating services around "nouns" like customer and product. Microservices, while certainly capable of being noun-based, are just as likely to be verb-based, that is to say, based on a functional grouping like "login" or "checkout." SOA was essentially an extension of object-oriented design while microservices are more about decomposition for purposes of rapid (agile) development with the bonus of having significant scalability advantages over object-oriented architectures.
Now, I told you that to tell you something else about the emerging relationship between proxies and applications. Or perhaps more accurately the merging relationship between the two. Microservices alone don't make an application, of course. There must be something else directing the user to the right services at the right time. That's generally what we call the "application." But increasingly that "application" is on the client, composed in the browser or using a native mobile app toolkit. What that means is a whole lot of direct calls from clients to services.
Except it doesn't, because that would be a networking and security nightmare. And neither do the local service proxies providing scale and security for those services. Because that'd still be a networking nightmare. Imagine all those services exposed to the public Internet. Consider briefly the impact on DNS, on firewalls, and on all the intermediate devices providing application services in between.
Exactly. No one does that. No one sane, at least.
There is still the need for a a single "app" which represents the server-side logic in this equation. public interface through which requests for services are delivered, where access control is enforced, and where end-user performance is monitored and measured.
These requirements are easily fulfilled by an application proxy. It offers a single, strategic point of control that can provide the visibility, security and simplicity needed to effectively scale, secure and manage the breakup of monolithic architectures into microservices by mediating between the apps (users) and its composite services.
The proxy is the app, for all intents and purposes, because it is the only public interface seen by the app whether that's a native mobile or browser-based app. The proxy is able to provide visibility and security as well as performance-related services on behalf of the "app" because it effectively becomes the app itself. It provides the app routing functionality required to ensure a request to "login" goes to the "login service" while a request to "checkout" goes to the "checkout service" while mediating with the client to ensure performance and security requirements are met.
The (local) service proxy, on the other hand, ensures scale at the service level while providing a service-specific view of performance and capacity. When combined with that information from other services, along with the overall "app" capacity and status at the app proxy level, fine grained policies can be coupled with real-time monitoring to adjust capacity across the services as well as the app.
Now, you could posit that a "real" app could do these things and you'd be right; it could. The question is what else that app might be doing that can't be duplicated by the app proxy (particularly if that app proxy is imbued with the right programmability)? Because there's going to be an app proxy in front of the app either way; scale and security are still necessary, regardless. So you can have both the app proxy and the app (probably many instances of it) or you very well might consider ditching the "app" and just using the proxy to do it instead*.
It doesn't make sense to complexify the architecture (and introduce additional points of failure and costs to scale) if the app proxy is smart enough to do what you need it to do.
The proxy can certainly be the app. And you never know, it might actually be better at being the app than an app might be itself.
* Assuming the programmatic capabilities exist in the app proxy in question to eliminate the app itself.