Hosting

Static Content

The simplest sites are ones with no dynamic content, commonly referred to as static. The primary attribute of sites such as these is that you don't need any special backend logic in order to provide the functionality. Think of blogs, businesses homepages, reference information, that sort of thing, anything where all of the pages that could possibly be rendered could be generated ahead of time.

There are a lot of options out there, but an especially popular one is GitHub Pages. This allows you to host your static website at zero cost, moreover the bandwidth limitations are generous enough that they don't provide practical limitations unless your site is wildly successful.

Dynamic Front End Sites

A more recent trend has led to services allowing websites to be hosted with no dedicated backend server. Instead, all of the dynamic functionality is served by third party services, allowing deployment to static hosting providers while providing richer functionality.

This is a new enough concept that I wasn't able to find a term for it. For an example of this, I believe Firepoker where we did our estimations last class uses this approach. It's named after Firebase, which offers a number of services supporting this style of website.

There is certainly some appeal to the flexibility in hosting this option provides, but bear in mind it also provides a lot more information to malicious users. Instead of endpoints focused on just the functionality you provide, this forces you to share everything about how your site works. Although there are security controls you can put in place with these systems, it gives them the ability to cause a lot more damage if there's a hole in that security.

Platform as a Service

When you're using a fairly popular software stack, it's relatively simple to provide a more curated experience for it. This is the same functionality over flexibility tradeoff we discussed with Rails. Because it can make assumptions about your system, you're able to get a site up and running much more easily. On the other hand, if you need to do something novel it may not have the flexibility you need.

Since it follows the same tradeoff as MVC architectures, they rose to popularity around the same time. The first big name is probably Heroku, which made it dead simple to deploy a Rails app. Since then, other major players such as Vercel have also emerged, but the offerings among them are extremely similar.

Cloud Platforms

If your product has special requirements, you may need to turn to full scale cloud platforms such as AWS, GCP or Azure. These companies provide large suites of services allowing you to build using any combination of technologies necessary. In addition to being able to deploy virtual machines that are always running, they also provide the ability to set up different hosting and storage options that can scale much more efficiently.

These platforms are more difficult to set up, but they provide attractive tradeoffs when a site reaches a particular scale. On the other hand, although they have a fairly generous free tier these days, for a simple site that's not serving too much traffic you could easily find yourself in a situation costing you hundreds of dollars a month.

Another problem with these platforms is that what you have set up can get quite sprawling. It's easy to lose track of something you provisioned months ago, and that could come back to bite you later on when trying to diagnose an issue. If it's expensive, you'll typically notice it on your bill and quickly track it down, but it's not uncommon to get charged a couple dollars for something you accidentally set up or forgot about. Tools like Terraform are making that easier, but there are still a lot of rough edges when you're working this low level.

Physical Data Centers

Some businesses still build their own data centers, but making this decision is becoming less and less common. There is a certain economy of scale in operating a data center that large companies such as Amazon, Microsoft and Google can achieve, one that all but the largest companies have difficulty realizing. As a result, it is extremely rare these days for a business to decide to build a new data center rather than hosting on on a Cloud Platform.

The main reason people make the decison today is because of strict regulations their company is under. A decade ago, this was why large health systems eschewed Cloud Infrastructure, but these situations are becoming less and less common. Perhaps defense contractors still fall into this category, but I'm struggling to think of a second example that would truly require a physical data center.