On reads, of course LMDB blows all the other away, as usual. Interestingly, they test with 4, 8, 16, and 32 concurrent reader threads on a Ryzen 5900X CPPU which only has 12 cores / 24 hardware threads.
Again, as expected, the LMDB read rate scales linearly from 1 / 4 / 8 threads, at 556ms / 141ms / 72ms. Essentially perfect scaling. The jump to 16 threads and 47ms isn't linear because the CPU has run out of cores, likewise for 32 threads at 44ms.
Howard Chu @ Symas
npub1tt…ex9ud
2024-02-07 01:26:28 UTC
in reply to nevent1q…p4yh
Author Public Key
npub1tt4j2zeswksjh5z7zmy283qd4yd920afy9j28xg45wsxzhl9rpjqrex9udPublished at
2024-02-07 01:26:28 UTCEvent JSON
{
"id": "0e319c78620c710d1f6a16637764e451a7f44e3bf8b805de15f80c7c7ccdc3df",
"pubkey": "5aeb250b3075a12bd05e16c8a3c40da91a553fa92164a39915a3a0615fe51864",
"created_at": 1707269188,
"kind": 1,
"tags": [
[
"e",
"2e9c5e2fe66c892b48e7308769dd9b0e2e3a34d69d31f09ed21f36ff4dbe285c",
"wss://relay.mostr.pub",
"reply"
],
[
"proxy",
"https://mastodon.social/users/hyc/statuses/111887593549586249",
"activitypub"
]
],
"content": "On reads, of course LMDB blows all the other away, as usual. Interestingly, they test with 4, 8, 16, and 32 concurrent reader threads on a Ryzen 5900X CPPU which only has 12 cores / 24 hardware threads.\n\nAgain, as expected, the LMDB read rate scales linearly from 1 / 4 / 8 threads, at 556ms / 141ms / 72ms. Essentially perfect scaling. The jump to 16 threads and 47ms isn't linear because the CPU has run out of cores, likewise for 32 threads at 44ms.",
"sig": "d95f8033bb0378ff8adf673849d783f2901478b4f66434695d517db13f6714f5d7806bc5b7f1314d6bae38d5947613cc9f39d5480cf1d87201c08b2d4ac215c5"
}