Multi-hop on IVPN halved my throughput — what are the rest of you actually on?

Been on IVPN about a year, mainly for the multi-hop and the port forwarding, neither of which the obvious alternative does.

Problem: multi-hop costs me roughly half my throughput, consistently, across three different exit pairs. Single hop is fine. I understand why in principle — two tunnels, two handshakes, whatever the second hop's peering looks like — but I'd like to know whether that's everyone or whether I'm picking bad pairs.

Related question: what does everyone actually run for this, and does anybody use multi-hop in earnest, or is it a feature people pay for and then switch off?

Multi-hop protects you against one specific thing: a provider that logs, correlated against an exit observer. If you believe your provider logs, the correct move is a different provider, not a second tunnel through the same one.

Two hops through one company is one company. People run it because it feels like more, and it is not more, it is slower.

There are threat models where it matters — where the two hops are two companies in two jurisdictions and you arranged that yourself. That is not what a checkbox in a client gives you.

Fair hit and I don't have an answer. I turned it on because it was there.

Mullvad, three year now. Flat five euro, no email, account is a number they generate and you write down. Pay Monero and is ten percent off. Audited, and the audit is a real one not a marketing one.

But is honest to say what you lose. No port forwarding since they remove it — this is real problem if you need incoming connection, and from your post sound like you do. No multi-hop either. So for your case maybe is not upgrade, is downgrade.

Anyway for the boring case of i want a tunnel and nobody know who i am, is nothing better and nothing simpler.

Port forwarding removal was a policy change, not a technical one. Worth saying, because people assume it's coming back.

On the throughput question specifically, since nobody has answered it: half is about right, and it isn't your pairs.

The second hop doesn't just add latency, it adds a second round of encapsulation on a link that was already carrying encapsulated traffic, and the MTU arithmetic gets worse each time. If you are seeing exactly-about-half rather than a variable penalty, that is the shape of an MTU problem rather than a peering problem, and it is worth a look before you conclude the feature is bad.

I have not tested this on anything above a gigabit link and I would not assume it scales the same way.

MTU was it, or most of it. Dropped the tunnel MTU and I'm at about seventy per cent instead of fifty.

Still turning it off though, for nullroute's reason rather than the speed one.