I'm seeing some unexpected latency results when pinging between my VMs in different Azure regions. Specifically, when I ping between VMs located in the West US2 and East US2 regions, the latency is around 73ms, which aligns with the figures published by Microsoft. However, when I ping from my private datacenter in Ohio to West US2 via our site-to-site VPN, the latency is only 55ms. This doesn't quite add up for me. Shouldn't the Azure network backbone have lower latency compared to my VPN connection over the public internet? Can anyone clarify this situation for me?
2 Answers
It's a good question! The physics of latency can't be ignored. General latency between major cities is around 60-70ms, so what you're seeing isn't far off. Even when you think about the Azure backbone, the speed of light and other factors play a significant role. No provider gets to bypass those laws, including Microsoft. So while Azure’s infrastructure is robust, it's still bound by physical limits just like any other connection.
Microsoft is trying to upgrade the speed of light, but I think they're stuck in marketing because all the good names are taken!
Keep in mind that the internet often finds the most efficient path to its target, which could sometimes involve fewer hops. The Azure backbone may not always have the least congested routes, and while Microsoft has powerful infrastructure, it doesn't guarantee lower latency compared to a localized VPN that’s effectively optimized for direct routes.
Totally get that, but you'd think Microsoft would have less congestion and more dedicated paths compared to a regular internet user.
That's a little harsh, but yeah, I think there should be fewer hops on a private backbone like Azure’s compared to the internet.