Thomas Cameron
Thomas Cameron
I have done something similar in the past. Here is my take: - zstd should be faster based on some benchmarks - Also, the official [wiki](https://github.com/facebook/rocksdb/wiki/Compression) recommends zstd Let me...
So I did some benchmarking using scripts I found on `bench/services`; I used both `upsert` and `search` respectively. It turns out that Lz4, snappy and zstd makes almost no difference...
So I tried to benchmark it with `cargo bench`. I tried every single one of the bench mark target that I could find on `lib/segment/Cargo.toml`, but I always end up...
Thanks for the tip. Let me try it out.
Hey @agourlay Thanks, it worked. Yes, i totally forgot to put `--bench` in there 😢
Hey @generall, @agourlay I measured the time required to start up the server and I was able to observe a noticeable difference this time. - Time required to start the...
So here is the test result. It's a copy and paste of my shell. https://github.com/thomas-k-cameron/qdrant/blob/zstd/examples.md You can change the compression algorithm by changing the env variable e.g. `ROCKSDB_COMPRESSION="lz4" cargo run...
I think the easiest would be to go through a snapshot. e.g. Take a snapshot of the storage and recover the snapshot the next time you start up the database....
@agourlay How is it going?
> @agourlay @thomas-k-cameron > > i dont have anything to add here, just trying to hunt down serious memory issues we are having, seemingly related to payloads. > > Not...