Re:The Kivaloo Data Store
by morelisp on Monday September 21, 2020 @05:36PM
(#24548752)
Is the one-byte length central to some logic (or some performance concerns), or could one hack the `uint8_t` to a `uint16_t` in kvldskey and so on to get something that could store a bit more? We have some systems that need only 16-32 byte keys but upwards of around 16k values, we're currently using RocksDB but these performance numbers + no compaction process + mux approach are making me interested...
[ Parent
| on HN ]
Interesting question. You would need to adjust kvldskey and all of the places where they are serialized (e.g. in pages and in the network protocol). You would also need a much larger page size -- the maximum key+value pair size has to fit into 1/3 of a page.
But if you're willing to make those adjustments and use 64 kB pages, I imagine it would work just fine. Please stay in touch!
Forgive my complete ignorance/inexperience, but why does the page size need to be increased to 64kB, and why does the key+value length(?) have to fit into 1/3 of a page?
B+Tree leaf nodes contain keys and values. In kvlds (which is the core B+Tree component of kivaloo), the code aims to keep nodes at least 2/3 full, in order to keep the tree approximately balanced. If a key-value pair can take more than 1/3 of a page, you could have a situation where a node is less than 2/3 of a page but adding another key-value pair would take it above the maximum allowed size.
This restriction could be relaxed, e.g. to require only that internal nodes are at least 2/3 full, in which case for the small key / large value case you could have pages which are barely larger than the largest value. I didn't bother doing that since it wasn't relevant to my usage.