This is an implementation of the IndraDB datastore for sled.
The sled datastore is not production-ready yet. sled itself is pre-1.0, and makes no guarantees about on-disk format stability. Upgrading IndraDB may require you to manually migrate the sled datastore. Additionally, there is a standing issue that prevents the sled datastore from having the same level of safety as the RocksDB datastore. Performance of sled in two-hop-queries is better than rocksdb though. We measured about 40% faster query times.
- This version does index all properties by default (and therefore duplicates the data). This is adding the benefit of faster queries but also increases the storage requirements.
- Batch operations are not optimized yet,