SunisDown
SunisDown
here's the results ``` Result with WILLNEED enabled: Time elapsed for get_links_range: 17250nanos Time elapsed for get_links_range: 18250nanos in cache ``` The performance is better while I disable the WILLNEED...
It's weird that when I put the test code in the same UT, then the result changed. I will dig it and update here if I have new updates.
I have retested it and it shows same performance with `WILLNEED` comparing to without `WILLNEED`. how I tested it: 1 create a collection and insert data in the collection 2...
collection config: ``` curl -X 'GET' \ 'http://localhost:6333/collections/benchmark' \ -H 'accept: application/json' { "result": { "status": "green", "optimizer_status": "ok", "vectors_count": 1000000, "indexed_vectors_count": 1000000, "points_count": 1000000, "segments_count": 2, "config": { "params":...
@generall > "vectors_count": 0, sorry, the incorrect data was pasted and I corrected it.
[OpenDAL](https://github.com/apache/incubator-opendal) project seems to fit this issue, which can help the system integrate with object store.
is `payload-index-config-persistence` in this repo? can't find it.
I submitted a PR that validates the collection name when creating the collection.
> Though I do like to support a / in collection names, it may be better to just ban this (and some other) characters all together. Agree. It would be...
I'm not sure if this is a breaking change since the collection name like "test/demo" does not work as expected. > I'm unsure at this time if we should consider...