Hi
I have two machines fairly far apart, at around 80ms. Tailscale works great, no problems whatsoever, but… I’d like to reduce the latency.
The general answer here is “you can’t improve the routing over the internet”, but that’s not entirely true. My ISP in both places has priority routing to google, and google’s network connectivity across regions is very good.
My thought was going from:
tailscale machine in country A <-> tailscale machine country B
To:
tailscale machine in country A <-> GCP machine <-> tailscale machine in country B.
And if that doesn’t give low enough latency, to setup two GCP machines in different regions to do this. Aka get into google’s prioritized routing network as quickly as possible, then go across regions, then out to the local node again.
Has anyone tried to do anything like this? Or any ideas on how I can set up the routing? In general with a setup like this tailscale would prioritize trying to go direct I’d assume. One thing I imagine can do - which is a bit hacky, is using subnet routing on gcp and setting up individual connections from the local machines to gcp, then subnet routing to connect them, but it’s … hacky.