Hey there. I’ve looking at a project for a Zerotier implementation, using Mikrotik routers on a bunch of IoT remote sites and in my lab tests using an RB5009 (ARM64 variable speed) as the remote router I’m getting about 300 Mbps which is more than sufficient for the IoT devices in question.
But on the backhaul side of things with around 200+ remote sites, I’m wondering whether I would get better performance by using one of their high-specced routers (ARM64 1700 Mhz) or go the CHR (Mikrotik RouterOS in an Intel VM) with Zerotier in a container. Or just a plain Linux VM with Zerotier setup for forwarding.
Are there any reference points for comparing Zertotier on ARM64 vs Intel at equivalent CPU speeds? And with with multiple endpoints, is the whole process single-threaded or can this be spread across CPUs for the individual connections?
Notes : Testing lab is internal with 10GbE backbone, so the physical transport is not a bottleneck.
Sorry we missed you. Mikrotik did a bunch of custom work in their integration, so it’s hard for us to speak to the performance. And we’re not super familiar with their hardware.
The current version of stock zerotier has single threaded IO and that is the main bottleneck.
The only ARM64 testing we’ve done is informally on macbooks (pretty fast!)
Have you already tested “plain linux vm on CHR” at this point?
Not yet - that’s my next step. I’m going to try and reproduce the remote site configuration that I’ve got on the Mikrotik on a bare bones Linux machine as well as a CHR. I imagine that bare bones will be marginally faster, but that from an administrative standpoint, having a homogenous park for the routing would be easier for management (update management, etc), but now I need to check into things like hardware compatibility etc with the machines that I have kicking around in the lab.