How Much is Google Cloud Platform‘s Inter-Region Latency?

Wonder how fast Google‘s data centers can talk to each other? As an infrastructure geek, I did too – so I ran some tests across Google Cloud Platform to find out.

Latency between cloud regions critically impacts where you place workloads. You want your services deployed closest to your users for fast response times. But how speedy is Google‘s own backbone between data centers?

I‘ll cover my testing methodology, then dive into the results and what contributes to Google Cloud‘s inter-region latency. By the end, you‘ll know:

  • Numeric latency benchmarks between 24 Google Cloud regions
  • How their network compares to AWS and Azure
  • Architectural factors that influence cloud latency
  • Tips for optimizing your infrastructure

So buckle up! I‘ll explain things plainly, as if chatting over coffee. Let‘s get to it…

Why Google Cloud Latency Matters

First, why should you care about latency between Google Cloud regions?

In a word: speed.

Low latency allows faster communication between your cloud services. And faster services keep your users happy.

As an example, I host Geekflare‘s web tools across multiple Google Cloud regions to be closer to users. This reduces public internet latency.

But my services still must talk between data centers. So Google‘s internal latency directly impacts the user experience.

I replicated tests in 24 regions to benchmark how speedy their backbone really is…

My Google Cloud Latency Test Methodology

To measure latency, I provisioned identically-configured VMs in each region and pinged between them.

Nothing fancy installed. Clean Ubuntu 20.04 LTS with 1 vCPU and 0.6 GB memory, the f1-micro type.

I pinged private IP addresses to test internal latency between regions, not Internet latency. Public pings would incorporate uncontrolled variables from ISPs.

Here‘s the command I ran 10 times each way, then averaged:

ping -c 10 <Private IP of VM in Other Region>

This tests how long 10 ICMP echo packets take to traverse Google‘s internal backbone from point A to B and back again.

I formatted all the data together to spot global trends…