Next, we're going to run through the options for the different endpoint types. The underlying implementation and network connectivity will depend on the option you pick, so it's important to know what you need before deploying.

The first option follows the example given in the previous section: edge-optimized endpoints. An edge-optimized endpoint is a deployment that leverages the CloudFront global distribution network and is useful if you have clients connecting from multiple locations around the world. API requests are sent to the nearest CloudFront point of presence, or edge, which can improve latencies for connections. At the time of writing, there were 189 edge locations across 37 countries.

The next is a regional endpoint deployment type, which is similar to the first option but doesn't come with a CloudFront distribution by default. You can still create and connect your own distribution to an endpoint; doing this would give you the flexibility of being able to manage your own.

Finally, a long-awaited feature is the private endpoint type. This enables you to create an API endpoint inside your own VPC. Previously an API endpoint was public by default, so the introduction of this feature allows you to deploy API Gateway so it can be used by private resources in your subnets. This endpoint type uses VPC endpoints, so you will need to create and apply a resource policy that allows API invocations from your own source VPC.

As shown in the following, the endpoint type can be selected when you're creating a new API:

Another relatively new feature is the ability to create WebSocket APIs—more details on that next.