mux

mux

SiaMux is a high-performance stream multiplexer. It allows you to operate many distinct bidirectional streams on top of a single underlying connection. We built it for Sia because we weren't satisfied with other multiplexers available at the time.

As a privacy-focused multiplexer, SiaMux behaves differently from other muxes. It transparently encrypts the connection, and supports authentication via Ed25519 public keys. To hinder metadata analysis, it always writes data in fixed-size "packets," inserting padding as necessary. Lastly, SiaMux implements a unique feature known as covert streams. Covert streams hide their data within the padding of other streams, making them completely undetectable to network analysis (at the cost of greatly reduced throughput). SiaMux can be used for any application in need of a multiplexer, but its privacy features make it a particularly good choice for p2p networks and other distributed systems.

Usage

Dialer:

conn, _ := net.Dial("tcp", addr)
defer conn.Close()
m, _ := mux.DialAnonymous(conn)
defer m.Close()
s := m.DialStream()
defer s.Close()
io.WriteString(s, "hello, world")

Listener:

l, _ := net.Listen("tcp", addr)
conn, _ := l.Accept()
m, _ := mux.AcceptAnonymous(conn)
s, _ := m.AcceptStream()
defer s.Close()
io.Copy(os.Stdout, s)

For authenticated communication, use mux.Dial/mux.Accept with a crypto/ed25519 keypair.

To create a covert stream, use m.DialCovertStream. The accepting peer calls m.AcceptStream as usual.

Benchmarks

SiaMux allocates very little memory (some buffers at startup, plus the Stream objects), does not use channels, and spawns just two goroutines per multiplexer. Despite encrypting the connection, in benchmarks SiaMux is competitive with alternatives such as yamux, muxado, and smux.

BenchmarkMux/1            5156 ns/op     830.90 MB/s      193977 frames/sec        0 allocs/op
BenchmarkMux/2            9550 ns/op     897.21 MB/s      210562 frames/sec        0 allocs/op
BenchmarkMux/10          50567 ns/op     847.19 MB/s      198281 frames/sec        0 allocs/op
BenchmarkMux/100        449494 ns/op     953.07 MB/s      223966 frames/sec        0 allocs/op
BenchmarkMux/500       2184647 ns/op     980.48 MB/s      229271 frames/sec        7 allocs/op
BenchmarkMux/1000      4548476 ns/op     941.85 MB/s      221448 frames/sec       28 allocs/op
BenchmarkCovertStream    45805 ns/op      93.53 MB/s       21833 frames/sec        2 allocs/op