search

ipfs-rust / quinn-noise

Noise handshake for the quinn quic implementation.
20 stars 5 forks source link

Update to latest quinn #6

Closed rklaehn closed 1 year ago

rklaehn commented 1 year ago

Basically just box all the things, and move some things around a bit...

rklaehn commented 1 year ago

I validated this by hacking it into the quinn bench at https://github.com/rklaehn/quinn/tree/main/bench

It seems to work, but is a tad slower than TLS:

This:

Transferred 1073741824 bytes on 1 streams in 4.05s (252.54 MiB/s)

Stream download metrics:

      │  Throughput   │ Duration 
──────┼───────────────┼──────────
 AVG  │  252.56 MiB/s │     4.06s
 P0   │  252.50 MiB/s │     4.05s
 P10  │  252.62 MiB/s │     4.06s
 P50  │  252.62 MiB/s │     4.06s
 P90  │  252.62 MiB/s │     4.06s
 P100 │  252.62 MiB/s │     4.06s

TLS:

Transferred 1073741824 bytes on 1 streams in 2.68s (381.66 MiB/s)

Stream download metrics:

      │  Throughput   │ Duration 
──────┼───────────────┼──────────
 AVG  │  381.62 MiB/s │     2.68s
 P0   │  381.50 MiB/s │     2.68s
 P10  │  381.75 MiB/s │     2.68s
 P50  │  381.75 MiB/s │     2.68s
 P90  │  381.75 MiB/s │     2.68s
 P100 │  381.75 MiB/s │     2.68s
dvc94ch commented 1 year ago

probably due to using chacha poly instead of aes gcm

dvc94ch commented 1 year ago

Thanks!