AccretionDB — Engine Introspector

User Bytes Written

—

lifetime

GET Operations

—

total get() calls

Bloom Effectiveness

—

SST scans skipped

VLog Reads

—

value fetches from disk

Write Amplification WiscKey advantage

—

× amplification

Bounded at ~2.0× (WAL + VLog ingestion). Compaction shuffles only 20-byte key pointers, not values.

Read Amplification Bloom-filtered

—

× amplification

1.0× = 1 SST search + 1 VLog read per get(). Bloom filters eliminate false binary search overhead.

LSM Tree State —

Memtable

—

↓ flush when full

L0 SSTables

—

↓ compact when L0 ≥ 15

L1 SSTables

—

↓ value lookup

Value Log (VLog)

Append-only · key-value separated · WiscKey architecture

Bloom Filter Detail

SSTables considered —

Bloom skips —

Binary searches executed —

Skip rate —

0%Skip rate100%

Engine Health

WAL status —

Memtable fill —

L0 compaction pressure —

Memtable entries —

Initializing...

Interactive Console live · zero-overhead

accretiondb> Engine introspector ready. Commands: put <key> <value> · get <key> · del <key>

put <key> <value> get <key> del <key>

Benchmark Runner synchronous · isolated

Workload

Operations

Results will appear here after running a benchmark.

Benchmarks run against an isolated acdb_bench_dir and clean up after themselves. Engine production data is unaffected.