Align read workload GC handling across engines

This commit is contained in:
2026-03-08 21:57:09 +08:00
parent ddc3f8af7e
commit 2c33c45d2c
5 changed files with 37 additions and 5 deletions
+1
View File
@@ -43,6 +43,7 @@ mkdir -p "${KV_BENCH_STORAGE_ROOT}/basic_mace" "${KV_BENCH_STORAGE_ROOT}/basic_r
## What Is Compared
- Comparison unit: rows with identical `workload_id`, `threads`, `key_size`, `value_size`, `durability_mode`, `read_path`
- Fairness rule for read-heavy workloads: `get`, `scan`, `mixed`, and `W1`-`W6` run one GC/compaction pass after prefill and before warmup/measurement, so RocksDB is not compared with GC artificially disabled while reads may have to touch multiple SSTs
- Throughput metric: workload-level `ops_per_sec` (higher is better)
- `W1/W2/W3/W4`: mixed read+update throughput
- `W5`: mixed read+update+scan throughput