[noise] Wiki? (was: Re: Echo server and benchmarks)

Trevor Perrin trevp at trevp.net
Thu Apr 21 16:21:37 PDT 2016


On Thu, Apr 21, 2016 at 12:36 AM, Alex <alex at centromere.net> wrote:
> I'm currently hosting an "echo server" which will echo back to you any
> payloads received. The server supports all combinations of handshake
> patterns, ciphers, DH schemes, hashes, etc.

That's awesome - you're producing a lot of good resources (vectors,
benchmarks, test servers), I'd like to get this linked from the
website, but it might take me awhile, and then I'm a bottleneck.

What about having a developer wiki, linked from the website, where we
can collect useful resources?  Would this work? -

https://github.com/noiseprotocol/noise_wiki/wiki


> I also ran benchmarks[2] on all 512 combinations of handshakes
> [2] http://centromere.github.io/cacophony/benchmarks.html

That's beautiful, really nice charts:  it's good to see visuals on
what affects performance:

For example, hash choice:

https://centromere.github.io/cacophony/benchmarks/XX.html

Noise_XX_25519_ChaChaPoly_SHA256 = 1.15
Noise_XX_25519_ChaChaPoly_SHA512 = 1.16
Noise_XX_25519_ChaChaPoly_BLAKE2s = 1.12
Noise_XX_25519_ChaChaPoly_BLAKE2b = 1.15

So the difference between slowest (SHA512) and fastest (BLAKE2s) is
only 3-4%.  And BLAKE2b is about equivalent to SHA256.

The 448 DH looks 5-6x slower than 25519, whereas with equivalent
optimization it should only be 3-4x slower.

The PSK penalty is larger than I'd expect, around 10%?

NoisePSK_XX_25519_ChaChaPoly_SHA256 = 1.28
NoisePSK_XX_25519_ChaChaPoly_SHA512 = 1.32

Hmm... I wonder if your hashing is relatively slow compared to 25519?

Trevor


More information about the Noise mailing list