Measuring performance from different regions and different clouds is core to what we do and providing developers, managers and product owners with the tools needed to understand what the impact of different locations and clouds is critical. To help people with this information we’ve set up a network of HTTPbin servers for developers to use to test against.
In each of the seven current Google Cloud regions, we have set up a small f1-micro VM running the HTTPbin API service. They are set up within a shared VPC network, which is fronted by the Google Cloud Load Balancer. They also each have an external IP address allocated, so they may be accessed directly.
The external IPs for the Load Balancer and the VMs were assigned to DNS records for both apimetrics.xyz and api.expert – the diagram below only shows the api.expert domains, the naming pattern is the same for both. We are using Google Cloud DNS for the nameservers, with the TTL of each record set to 24 hours to allow caching for a reasonable period of time.
We are running tests from every APImetrics agent each of these endpoints. This will allow us to determine which locations are closest to Google’s infrastructure, and if any locations get mis-routed. It will also give us a base-line for expected latency between any two points on the globe. It will also allow us to see how DNS lookup times differ across different cloud providers.
Our Google Map shows the last 24 hours of data graphically, but please feel free to use this for calls and if you’re not a user, why not sign up for an account today?