Replying to Avatar ridsdszltdel

Nice benchmark, kamakura-san. Hopefully, you will present this with detail reports on Hackaton Nostrasia Final Presentation.

Only a bit suggestion,

1. Maybe you also need to make sure clearing the I/O cache (/proc/sys/vm/drop_caches, etc) between every test loop to get a fresh state if you haven't done it (yet).

2. I'm assuming this test using 'default' settings for those relay, right? Maybe you can also have the benchmark with some 'optimized' settings if you have the time.
Avatar
kamakura 2y ago

Thanks for the good suggestion. Your feedback is helpful😉 In this test, the cache will be fresh because relays are set up with docker each time🌱

As you noted, the relay settings are in their default. Also, the benchmark load is not maximized. I think trying more combinations of relays and test setups will give us deeper insights.

Reply to this note

Please Login to reply.

Discussion

Avatar
ridsdszltdel 2y ago

Ah, glad to hear that,

Indeed, especially since there are no real formal test between many relays as far as we know except for what you have done. 🙏

Maybe, if you don't mind, you can probably try to test rnostr https://github.com/rnostr/rnostr . Especially, since it is written in Rust and uses LMDB, its performance should be around strfry too.