Modify Python & performance Locust library to use post-quantum algorithms

[davidgca]

Hi everyone,

I've been working in the performance field for a while, and I've always had a keen interest in quantum computing, particularly quantum and post-quantum algorithms. Merging these interests, I've developed a modified version of Locust that integrates the OQS-modified version of OpenSSL, allowing me to make both pre- and post-quantum requests. Would you be interested in me sharing this modification?
The goal is to create Python scenarios using Locust, that handle both pre-quantum and post-quantum algorithms for comparison. Here’s a quick rundown of what I did:

Recompile OpenSSL using the OQS Dockerfile as a base.
Recompile curl using the OQS Dockerfile as a base.
Compile liboqs.
Recompile Python to utilize the OpenSSL version from the first step.
Set up a performance scenario in Locust where the post-quantum requests are made using Python's subprocess library and OpenSSL's s_client (since the current Python libraries don’t support post-quantum algorithms yet AFAIK).

What do you think? Do you see value in this? I share some results from one of my tests using the kyber768 and X25519 algorithms with a server that supports these curves.:

[SWilson4]

Thanks for sharing, @davidgca. This is certainly a nice presentation. Please correct me if I'm wrong, but I'm assuming this is based on our OpenSSL fork? I expect we'd want it to use our OpenSSL 3 provider for the OpenSSL integration, since the fork is no longer maintained. Do you think it would be possible to obtain a similar result using the provider instead of the fork?

[davidgca]

Alright, in that case, I think I can add all this code to the project. Let me know where it should go (I assume I just need to create a new directory and share a PR with you?).

On another note, I've completed an initial performance test using an OQS-enabled Nginx, and I noticed a significant difference between the post-quantum-supported server and a classical Nginx server. Would you like me to share this with you? Where would be the best place to publish the performance report?

[davidgca]

I think the best option is adding this to oqs-demos as @SWilson4 propose, WDYT?

[SWilson4]

I think that's the best plan. We can always move it to a dedicated profiling repo if we want to use it specifically for that purpose. As I understand it Locust is more powerful than just a benchmarking tool and perhaps deserves its own demo. I'd suggest opening a PR there creating a new "locust" directory to contain your Dockerfile / documentation / etc.

Thanks for being willing to contribute your work to the community!

[davidgca]

Here my little contribution:
open-quantum-safe/oqs-demos#304