eugeneia
commented
1 month ago Hey, designs like Snabb (and Rush) only really make sense when you are centrally routing traffic at data center scale, say 100G network cards in servers that are dedicated to the workload.
Since you are talking about P2P networks and portability requirements, I don't think Snabb/Rush's high-performance strategies (bypassing the OS, driving a NIC directly, doing privileged DMA) make much sense.
That being said, it is possible and possibly very pleasant (depending on your programming tastes) to use Rush to build networking applications in userspace without using exclusive access to networking hardware (see synthetic network https://github.com/daily-co/synthetic-network/blob/main/rush/src/synthetic_network.rs). It's not faster than what you can do with standard operating system network APIs, but you can mess around with raw packets etc.
Hello,
I hope that you are doing well today.
I am on the quest to build an ultra-fast routing mesh distributed P2P VPN approach for a Yggdrasil VPN type network but that is significantly faster and can massively scale since it will run a number of P2P project (i.e. IPFS, eCommerce, graphics gamming, ChatGPT-like AI, etc.)
I need to be able to compile it into a simple and efficient single application that can be run as a Daemon/Service on native Windows (no wsl), Linux, and MacOS.
To that end, I had been looking at various vSwitch and vRouter implantations and projects like Snabb Switch, Open vSwitch, and many others that might be viable as a core project to build on and develop, if at all possible.
Due to the nature of the applications that will be running on this, I am wondering if Rush might be a viable solution to get started and would like to discuss it more with you, if possible.
Thanks in advance and have a great day