Why disk latency differs in different DigitalOcean data centers?

[silving]

I've moved my droplet from one DigitalOcean data center to another and to my disappointment realized that the disk latency is noticeably higher in the new data center. It is the same droplet, just in other data center. What can be the reason?

https://user-images.githubusercontent.com/40499291/103486561-c0241c00-4e0f-11eb-8e3d-d75de18b0fda.png

[StratusBen]

Are you using general purpose Droplets? If so, DigitalOcean uses a heterogeneous mix of underlying hardware that has varying degrees of performance. Some of it is years old.

If you're looking for normalized performance, it's my strong suggestion you look at Droplets with dedicated resources. While I believe the disk virtualization layer is always technically "over-committed", these Droplets run on more modern hardware that typically has better and more normalized performance.

If you don't want to do that, an alternative strategy is just to destroy and recreate the general purpose Droplet to try and land on the better hardware by a stroke of luck...though that's probably not worth it.

I hope this helps you!

[silving]

Yes, general purpose Droplet. It's quite a big surprise and disappointment...

[robertcope]

Seems like a question you should be asking their support folks. We could guess, but...

[silving]

They did not give any explanation, just advised to do some tests using 'sar' to check disk latencies.

[wmf]

It could be a different hardware model or more VMs sharing the same disk.

[silving]

Or is it possible that in different data centers different hardware is used? I thought it's supposed to be the same.

The thing is that higher latency affects the whole system hosted on the droplet (Mongo database).

[silving]

So it's possible that other websites hosted on the same VM affect latency of my droplet's SSD?

[IcePic]

Ask yourself the opposite: Could there perhaps be a measurable difference if I am the only VM on the host, versus a ton of busy VMs next to mine?