Wondering about one thing. Currently running a fusionhub solo connecting with client with valid prime care. This works fine.
I know that max throughput is 1gbit, but that doesn’t interest me, I’m wondering about the number of clients, basically the max size of state table.
Each client is producing very little traffic, more so in terms of connections/sessions.
Just trying to figure out how to scale this. Of course I’m gonna setup more hosts, but hosting also costs, so just trying to optimise in terms of clients.
Any ideas about this?
I’ve tested same clients on pfsense and that lead to exactly that issue, state table got so large that everything halted. After some testing we figured out ish client number per hardware and instance spec.
No it’s not. Throughput is only limited by the presented physical connectivity and allocated hardware resources. I have FusionHubs deployed in regions which regularly see more than 2Gbps of throughput over SpeedFusion.
As you note - it depends - mainly on concurrent session counts and the management overhead of those sessions.
We do a lot of enterprise SDWAN service extensions internationally - extending Versa SDWAN to new places over bonded 5G and Starlink. This is a lower session count than thousands of IoT devices, or wifi users who might then create hundreds of active sessions.
The only way to test is with actual network loads from real devices and extrapolate.
What I would say is we have had hundreds of peers connected to FusionHubs in the past but prefer today stay to below 250 peers per hub. Purely because the management interfaces get cumbersome when you have to scroll through that number of peers. Also its question of customer maintenance / management. Spreading a larger number of FusionHubs lowers my risk.
That said. Peplink wouldn’t sell a 4000 peer Fusionhub license unless it was possible to use it - and demanded by customers…
Cool, thanks for input.
We have it running now, and so far it performs great.
But yes, like you say, I can see the issue when you start having substantial amount of peers on one hub, but that is purely because of overview. Thanks to VRF this is easy simply works. Have some small issues with configuration of sub tunnels, but I’m playing around with it now to see what works and what doesn’t.