Update README.md

README.md

@@ -8,11 +8,12 @@ A K.I.S.S. implementation of log-structured merge trees (LSM-trees/LSMTs).
 
 This is the fastest and most feature-rich LSM-tree implementation in Rust! It features, among other things:
 
-- Size-tiered or Levelled compaction
-- Partitioned block index to reduce memory footprint [1]
+- Size-tiered or Levelled compaction with concurrency support
+- Partitioned block index to reduce memory footprint and keep startup time minimal [1]
 - Block caching to keep hot data in memory
 - Bloom filters to avoid expensive disk access for non-existing items
-- Sharded log & memtable for concurrent writes
+- Sharded journal & memtable for concurrent writes
+- Journal truncation on recovery for consistency
 - Atomic batch operations
 - Automatic background compaction & tombstone eviction
   - Does not spawn background threads unless actually needed
@@ -21,14 +22,6 @@ This is the fastest and most feature-rich LSM-tree implementation in Rust! It fe
 - CRChecks data blocks
 - 100% safe Rust
 
-### Performance to expect roughly:
-
-- Super quick startup, with low memory overhead (<10 MB for 500M+ objects)
-- ~400k+ writes per second for a single writer thread
-- ??? MB/s throughput
-- <1μs reads for hot (cached) data (2M+ reads per second) for a single reader thread
-- <10µs reads for cold data (500k+ reads per second) for a single reader thread
-
 ## Benchmarks
 
 Testing system: