I wonder if the post's focus on Linux/Android is just because that's where they started, or because they're leveraging techniques that are only possible on Linux/Android?
(Tailscale cofounder) Yes, that’s most of it. Most packet-level optimizations are necessarily platform specific and this post was mostly about the linux dataplane.
That said, the architectural multiqueue improvements and memory optimizations are useful starting points to build on for every platform.
I like to think of them like Docker. None of Docker's functionality was new, they just wrapped it in a smooth enough DX for it to reach escape velocity.
I used to LOVE tailscale. Then I put wireguard on my home network exposed to the internet with a dynamic DNS provider and it immediately became irrelevant. Not only is raw wireguard more stable (I don't have to fight the DNS issues on my mobile phones) it feels faster and is amazingly simple to set up.
Care to share your setup? I did some research into self hosting my own wireguard for my nuc and rpi, before ultimately settling on Tailscale because of how much simpler and plug-and-lay it was to add/remove devices compared to self hosting wireguard, not dealing with certificates, maintenance, etc.
There are no certificates to share with Wireguard. Nothing to rotate if you don't want to. Once it works, it works.
I've even got a backup wireguard server running on a Pi 1b. Works fine. We currently run wireguard on our router (and it seems more and more routers are supporting it).
There are keys to configure for each client, but once you have the configuration for one client, the rest come very quickly and easily.
I should add that I don't have any experience with Tailscale, but compared to OpenVPN and other VPN solutions, Wireguard is lightweight, simple, and easy to setup/configure.
We use it on all our mobile devices (phones, laptops, tablets) to tunnel our traffic through our home network with all the filtering it offers (along side access to private services we host).
I use an Island Router with Wireguard server built in, it handles DDNS and even the base Island router is beefy enough to give me up to 954mbps or so of Wireguard to right inside my home network. It took 2 minutes to set up.
That being said, I do understand the appeal of Tailscale and have used it.
Tailscale takes two minutes to setup and you can add more devices with zero configuration.
WireGuard takes 30 mins to an hour to set up, you'll need to configure port forwarding, DDNS, create keys for each device, and add them to each device manually. But you have 100% control.
Performance-wise, I haven't noticed a difference. My internet connection maxes out way before Tailscale hits any performance limits.
No you are being disingenuous if you think raw wireguard is amazingly simple to set up. Sure it is simpler than IPSec. But it’s absolutely not simpler than Tailscale. I migrated my raw Wireguard setup to Tailscale because after a few months without any tinkering I simply forget details of my setup. I’d rather outsource it to Tailscale.
It is slow. It cannot achieve speeds of greater than 1Gbps on clients systems (Windows & Mac), where you'd normally see it being used. On Linux, it struggles to achieve 10Gbps even when using a synthetic large packet benchmark [1]. With an IMIX benchmark, it would not be competitive whatsoever.
This problem is fixable. WireGuard achieves higher performance (Kernel vs Userspace implementation) and IPsec implementations can achieve 100Gbps/400Gbps (DPDK/XDP). Zero-copy networking.
From this blog post, I can say Tailscale still seems to not have the appetite for that, which is a shame.
And if any tailscale employees are reading this - https://github.com/tailscale/tailscale/issues/15724 please fix this too. Regular users not using some sort of enterprise saas DNS (whatever their thing is?) deserve DNS privacy too.
Tailscale's netstack is barely even WireGuard and they aren't compatible whatsoever. It's all marketing at this point.
So it's not that simple: it's impossible for Tailscale to use any existing kernel or accelerated WireGuard implementation. They could derive inspiration, but a kernel module for Linux won't fix Windows & Mac. With that said, I feel they have enough funding to maintain a few platforms (:
(Tailscale cofounder) That’s a good callout on DoH support, thanks.
That said, note that if you run your own DNS server on your tailnet, the regular UDP DNS is automatically private because it’s carried over Tailscale. That’s the most common setup for non-SaaS DNS servers. DoH doesn’t really add anything in that arrangement. (And it’s more fiddly because you need to get and refresh a TLS cert.)
You're correct, kernel isn't faster by default. With that said, the following is true:
1) the WireGuard kernel implementation, despite not even being zero-copy, exceeds the performance of the userspace implementation
2) implementations utilizing the userspace network stack have a maximum potential performance (context switch + memcpy is very slow, and that affects UDP disproportionately). It's the wrong approach for meaningful improvement.
(Tailscale cofounder) I see a few comments here that using kernel wireguard would make it faster; it’s not really that simple. In fact, for a while (and we wrote a blog post about it), our optimizations made wireguard-go faster than kernel wireguard because it was better optimized. They adopted some of those improvements and now we’re on to the next order of magnitude together.
For really high bandwidth cases, things like DPDK are the long term best choice and are primarily userspace, for good reasons. Kernel mode is not the pure benefit it once was (if it ever was).
Separately, wireguard itself has a problem that the crypto suite it uses is not supported by hardware accelerators. So if we want to get into the hundreds of gigabits range, we will possibly need to switch packet formats entirely. (But, wireguard also needs to update to support post-quantum so maybe they’ll fix both problems at the same time and we can join in.)
Wireguard with PQ won't be Wireguard, anymore. It'll just be a rehash of IKE+IPsec. What made Wireguard better was the very simple handshake and minimal state, but no PQ algorithms can support that simplicity because the keys are too large and/or not as simple to use as ECC.
Might as well switch to IPsec. Everything is already in place, including hardware acceleration. But most people won't, and we'll live in a world with duck-tape hacks built around a compromised Wireguard-ish layer.
(Tailscale cofounder) I’m a little more optimistic; wireguard was always going to need a v2 eventually, that’s just the nature of cryptography. It’ll always be simpler than IPsec as long as it avoids live negotiation (you need to specify your suite up front for each node you talk to; v1 and v2 are the only suites) and continues to go over UDP instead of IPsec’s “it’s a different transport protocol!” madness.
Things like Google’s PSP already achieve that while still being hardware acceleratable: https://github.com/google/psp
Post-quantum negotiation will kinda suck but you don’t have to sacrifice everything.
Considering how core Wireguard is to the whole Tailscale value proposition, are they funding the future of that V2 roadmap on the open source side? If the answer is "no", and because the security world isn't going to stay put (particularly with AI math advancements happening lately are accelerating medium term risk), then one has to assume Tailscale is working on it's own in-house "V2" equivalent protocol for post-quantum support to enable that vision? I doubt you're doing nothing.
IPSec suffers from the "it can do everything" syndrome.
Storytime: 15 years ago I co-founded a startup to build easy-to-use infrastructure management for AWS. At that time, it did not have cross-region VPC peering or routing, so you couldn't easily and safely have apps that communicate between regions.
So I started working on creating an overlay network. My idea was to use IPsec, it even has an RFC that documents its kernel interface. So that when a host wants to send a packet to the secure network, the kernel goes to my userspace daemon, that in turn goes to the central server that provides it the key for the given host pair.
And it turned out that the interface lacked a crucial part - on-demand key negotiation for incoming packets. It had this for _outgoing_ packets, but not incoming. The only sane way to make it work was to create a proactively updated database of all the hosts.
Well, I did that. It also did not work (tm). I found so many issues with broken MTU handling, broken NAT traversal, etc.
I eventually gave up on that idea and started working on a simple TUN/TAP-based overlay. I almost made everything work, but our startup got acquired by AWS, and this line of work was abandoned.
IPsec is a terrible fit. Tailscale uses a control plane for peer & key distribution and renders almost all of the complexity in the IPsec protocol useless.
(Tailscale cofounder) Many years ago I made a VPN for experimental purposes that kept IPsec’s data plane but abandoned IKE. It’s actually a pretty good arrangement and might be applicable here.
My implementation relied on the extremely finicky Linux kernel IPsec and would never have worked in userspace macOS or windows. But yes, a Tailscale control plane with a per-node switchable data plane, one of which is IPsec with PQC, is a real option that would work.
By locking down the control plane it becomes incompatible with classic IPsec, but also avoids most of the security gotchas that plagued IPsec.
I don't think this really follows. Negotiation would be problematic, but you can just version the protocols and do WireGuard v1 and WireGuard v2, with v2 in a PQ configuration. As long as the configuration about which to use is static, you're not running up against the IPSEC problem.
I don't think there's anything necessarily complicated about MLKEM that would make this too difficult.
Switching to IPSEC loses you other WireGuard benefits; the wins don't end at "just one carefully curated set of cryptography primitives", but extend into things like DoS prevention and a design that admits to processing incoming frames without dynamic allocation.
It's only half the story. WireGuard uses chacha20-poly1305 and certain implementations can still achieve 100Gbps. Hardware acceleration is the path to 400Gbps and power efficiency though.
But, wireguard also needs to update to support post-quantum
It's quantum-resistant if you define a PSK. You still have to distribute the key out of band, but you already have to do that anyways for the public keys of the peers.
1) Besides marketing, I see no reason that Tailscale needs to concern itself with the WireGuard protocol spec at all. Tailscale is already incompatible.
2) Tailscale's control plane model for peer distribution avoids every concern affecting IPsec in regard to protocol negotiation security and compatibility.
TL;DR, diverging from WireGuard & supporting AES (via protocol negotiation with hardware acceleration), would be relatively painless.
I wonder if the post's focus on Linux/Android is just because that's where they started, or because they're leveraging techniques that are only possible on Linux/Android?
it sounds like they’re leveraging Linux features that may not exist 1:1 on Darwin and NT
(Tailscale cofounder) Yes, that’s most of it. Most packet-level optimizations are necessarily platform specific and this post was mostly about the linux dataplane.
That said, the architectural multiqueue improvements and memory optimizations are useful starting points to build on for every platform.
Leaves them where they landed. Slop check : yes +++ATH
Also makes it sound like they were the first to do stuff like NAT traversal, which goes WAY back.
I like to think of them like Docker. None of Docker's functionality was new, they just wrapped it in a smooth enough DX for it to reach escape velocity.
In our use case latency went parabolic at 250 sessions pumping just 60 mb/s through a tunnel.
I would like less battery usage with exit node.
You’re using a battery powered device as an exit node?
I used to LOVE tailscale. Then I put wireguard on my home network exposed to the internet with a dynamic DNS provider and it immediately became irrelevant. Not only is raw wireguard more stable (I don't have to fight the DNS issues on my mobile phones) it feels faster and is amazingly simple to set up.
Care to share your setup? I did some research into self hosting my own wireguard for my nuc and rpi, before ultimately settling on Tailscale because of how much simpler and plug-and-lay it was to add/remove devices compared to self hosting wireguard, not dealing with certificates, maintenance, etc.
There are no certificates to share with Wireguard. Nothing to rotate if you don't want to. Once it works, it works.
I've even got a backup wireguard server running on a Pi 1b. Works fine. We currently run wireguard on our router (and it seems more and more routers are supporting it).
There are keys to configure for each client, but once you have the configuration for one client, the rest come very quickly and easily.
I should add that I don't have any experience with Tailscale, but compared to OpenVPN and other VPN solutions, Wireguard is lightweight, simple, and easy to setup/configure.
We use it on all our mobile devices (phones, laptops, tablets) to tunnel our traffic through our home network with all the filtering it offers (along side access to private services we host).
Get a Unifi system. It's built in. Works great!
I use an Island Router with Wireguard server built in, it handles DDNS and even the base Island router is beefy enough to give me up to 954mbps or so of Wireguard to right inside my home network. It took 2 minutes to set up.
That being said, I do understand the appeal of Tailscale and have used it.
My favorite thing about Tailscale is how it lets you SSH/RDP INTO servers without having to open any ports.
Tailscale takes two minutes to setup and you can add more devices with zero configuration.
WireGuard takes 30 mins to an hour to set up, you'll need to configure port forwarding, DDNS, create keys for each device, and add them to each device manually. But you have 100% control.
Performance-wise, I haven't noticed a difference. My internet connection maxes out way before Tailscale hits any performance limits.
Tailscale wins for convenience.
No you are being disingenuous if you think raw wireguard is amazingly simple to set up. Sure it is simpler than IPSec. But it’s absolutely not simpler than Tailscale. I migrated my raw Wireguard setup to Tailscale because after a few months without any tinkering I simply forget details of my setup. I’d rather outsource it to Tailscale.
In my opinion, this is Tailscale's largest issue.
It is slow. It cannot achieve speeds of greater than 1Gbps on clients systems (Windows & Mac), where you'd normally see it being used. On Linux, it struggles to achieve 10Gbps even when using a synthetic large packet benchmark [1]. With an IMIX benchmark, it would not be competitive whatsoever.
This problem is fixable. WireGuard achieves higher performance (Kernel vs Userspace implementation) and IPsec implementations can achieve 100Gbps/400Gbps (DPDK/XDP). Zero-copy networking.
From this blog post, I can say Tailscale still seems to not have the appetite for that, which is a shame.
[1] https://tailscale.com/blog/more-throughput
Yes. Just fucking stop doing userspace wireguard on linux https://github.com/tailscale/tailscale/issues/426 - issue has been open for 6 years (and is locked now, lol), btw.
And if any tailscale employees are reading this - https://github.com/tailscale/tailscale/issues/15724 please fix this too. Regular users not using some sort of enterprise saas DNS (whatever their thing is?) deserve DNS privacy too.
Tailscale's netstack is barely even WireGuard and they aren't compatible whatsoever. It's all marketing at this point.
So it's not that simple: it's impossible for Tailscale to use any existing kernel or accelerated WireGuard implementation. They could derive inspiration, but a kernel module for Linux won't fix Windows & Mac. With that said, I feel they have enough funding to maintain a few platforms (:
(Tailscale cofounder) That’s a good callout on DoH support, thanks.
That said, note that if you run your own DNS server on your tailnet, the regular UDP DNS is automatically private because it’s carried over Tailscale. That’s the most common setup for non-SaaS DNS servers. DoH doesn’t really add anything in that arrangement. (And it’s more fiddly because you need to get and refresh a TLS cert.)
Kernel networking is not automatically faster then userspace.
You're correct, kernel isn't faster by default. With that said, the following is true:
1) the WireGuard kernel implementation, despite not even being zero-copy, exceeds the performance of the userspace implementation
2) implementations utilizing the userspace network stack have a maximum potential performance (context switch + memcpy is very slow, and that affects UDP disproportionately). It's the wrong approach for meaningful improvement.
(Tailscale cofounder) I see a few comments here that using kernel wireguard would make it faster; it’s not really that simple. In fact, for a while (and we wrote a blog post about it), our optimizations made wireguard-go faster than kernel wireguard because it was better optimized. They adopted some of those improvements and now we’re on to the next order of magnitude together.
For really high bandwidth cases, things like DPDK are the long term best choice and are primarily userspace, for good reasons. Kernel mode is not the pure benefit it once was (if it ever was).
Separately, wireguard itself has a problem that the crypto suite it uses is not supported by hardware accelerators. So if we want to get into the hundreds of gigabits range, we will possibly need to switch packet formats entirely. (But, wireguard also needs to update to support post-quantum so maybe they’ll fix both problems at the same time and we can join in.)
Wireguard with PQ won't be Wireguard, anymore. It'll just be a rehash of IKE+IPsec. What made Wireguard better was the very simple handshake and minimal state, but no PQ algorithms can support that simplicity because the keys are too large and/or not as simple to use as ECC.
Might as well switch to IPsec. Everything is already in place, including hardware acceleration. But most people won't, and we'll live in a world with duck-tape hacks built around a compromised Wireguard-ish layer.
(Tailscale cofounder) I’m a little more optimistic; wireguard was always going to need a v2 eventually, that’s just the nature of cryptography. It’ll always be simpler than IPsec as long as it avoids live negotiation (you need to specify your suite up front for each node you talk to; v1 and v2 are the only suites) and continues to go over UDP instead of IPsec’s “it’s a different transport protocol!” madness.
Things like Google’s PSP already achieve that while still being hardware acceleratable: https://github.com/google/psp
Post-quantum negotiation will kinda suck but you don’t have to sacrifice everything.
Considering how core Wireguard is to the whole Tailscale value proposition, are they funding the future of that V2 roadmap on the open source side? If the answer is "no", and because the security world isn't going to stay put (particularly with AI math advancements happening lately are accelerating medium term risk), then one has to assume Tailscale is working on it's own in-house "V2" equivalent protocol for post-quantum support to enable that vision? I doubt you're doing nothing.
(Tailscale cofounder) Nothing to announce yet, but yes, we absolutely do sponsor wireguard, with money and with code.
IPSec suffers from the "it can do everything" syndrome.
Storytime: 15 years ago I co-founded a startup to build easy-to-use infrastructure management for AWS. At that time, it did not have cross-region VPC peering or routing, so you couldn't easily and safely have apps that communicate between regions.
So I started working on creating an overlay network. My idea was to use IPsec, it even has an RFC that documents its kernel interface. So that when a host wants to send a packet to the secure network, the kernel goes to my userspace daemon, that in turn goes to the central server that provides it the key for the given host pair.
And it turned out that the interface lacked a crucial part - on-demand key negotiation for incoming packets. It had this for _outgoing_ packets, but not incoming. The only sane way to make it work was to create a proactively updated database of all the hosts.
Well, I did that. It also did not work (tm). I found so many issues with broken MTU handling, broken NAT traversal, etc.
I eventually gave up on that idea and started working on a simple TUN/TAP-based overlay. I almost made everything work, but our startup got acquired by AWS, and this line of work was abandoned.
IPsec is a terrible fit. Tailscale uses a control plane for peer & key distribution and renders almost all of the complexity in the IPsec protocol useless.
(Tailscale cofounder) Many years ago I made a VPN for experimental purposes that kept IPsec’s data plane but abandoned IKE. It’s actually a pretty good arrangement and might be applicable here.
My implementation relied on the extremely finicky Linux kernel IPsec and would never have worked in userspace macOS or windows. But yes, a Tailscale control plane with a per-node switchable data plane, one of which is IPsec with PQC, is a real option that would work.
By locking down the control plane it becomes incompatible with classic IPsec, but also avoids most of the security gotchas that plagued IPsec.
Yes, I agree, IPsec's complexity is definitely why client implementations are so universally finicky. That complexity is not needed by Tailscale.
It could have worked in macOS! It implements RFC-2367 (PF_KEY socket type) and I even made it work on my laptop.
I traced the SA database operations on Windows, and it could have been done. Probably. But I abandoned that to preserve my sanity.
I don't think this really follows. Negotiation would be problematic, but you can just version the protocols and do WireGuard v1 and WireGuard v2, with v2 in a PQ configuration. As long as the configuration about which to use is static, you're not running up against the IPSEC problem.
I don't think there's anything necessarily complicated about MLKEM that would make this too difficult.
Switching to IPSEC loses you other WireGuard benefits; the wins don't end at "just one carefully curated set of cryptography primitives", but extend into things like DoS prevention and a design that admits to processing incoming frames without dynamic allocation.
I always thought that just kernel mode makes it better.
Didn't know that WG can't use hardware crypto accelerators. I hope it will someday
It's only half the story. WireGuard uses chacha20-poly1305 and certain implementations can still achieve 100Gbps. Hardware acceleration is the path to 400Gbps and power efficiency though.
It's quantum-resistant if you define a PSK. You still have to distribute the key out of band, but you already have to do that anyways for the public keys of the peers.
Why do public keys need to be distributed out of band?
1) Besides marketing, I see no reason that Tailscale needs to concern itself with the WireGuard protocol spec at all. Tailscale is already incompatible.
2) Tailscale's control plane model for peer distribution avoids every concern affecting IPsec in regard to protocol negotiation security and compatibility.
TL;DR, diverging from WireGuard & supporting AES (via protocol negotiation with hardware acceleration), would be relatively painless.
Any chance the Apple TV can get a performance boost? Specially as an exit node. It always gives very little bandwidth. Thank you!