LMDB Library Performance Analysis

Performance Analysis

The following tables show each benchmark ranked by latency (smaller is better), with percentage difference from the fastest LMDB library build. The latest tag (LMDB_0.9.33) is highlighted in bold.

Read by Key

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.29`	44.066	baseline
2	`LMDB_0.9.19`	44.080	+0.0%
3	`LMDB_0.9.21`	44.103	+0.1%
4	`LMDB_0.9.17`	44.104	+0.1%
5	`LMDB_0.9.33`	44.129	+0.1%
6	`LMDB_0.9.28`	44.131	+0.1%
7	`LMDB_0.9.27`	44.177	+0.3%
8	`LMDB_0.9.22`	44.217	+0.3%
9	`LMDB_0.9.30`	44.264	+0.4%
10	`LMDB_0.9.31`	44.296	+0.5%
11	`LMDB_0.9.20`	44.422	+0.8%
12	`LMDB_0.9.18`	44.681	+1.4%
13	`LMDB_0.9.24`	45.623	+3.5%
14	`LMDB_0.9.23`	56.618	+28.5%

Write Entry

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.18`	74.909	baseline
2	`LMDB_0.9.17`	75.790	+1.2%
3	`LMDB_0.9.19`	81.404	+8.7%
4	`LMDB_0.9.23`	82.870	+10.6%
5	`LMDB_0.9.24`	82.952	+10.7%
6	`LMDB_0.9.27`	82.989	+10.8%
7	`LMDB_0.9.28`	83.025	+10.8%
8	`LMDB_0.9.22`	83.027	+10.8%
9	`LMDB_0.9.29`	83.063	+10.9%
10	`LMDB_0.9.30`	83.237	+11.1%
11	`LMDB_0.9.21`	83.237	+11.1%
12	`LMDB_0.9.33`	83.304	+11.2%
13	`LMDB_0.9.20`	83.459	+11.4%
14	`LMDB_0.9.31`	88.419	+18.0%

Calculate xxHash64

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.27`	39.087	baseline
2	`LMDB_0.9.31`	39.224	+0.3%
3	`LMDB_0.9.28`	39.251	+0.4%
4	`LMDB_0.9.29`	39.269	+0.5%
5	`LMDB_0.9.21`	39.350	+0.7%
6	`LMDB_0.9.17`	39.389	+0.8%
7	`LMDB_0.9.20`	39.450	+0.9%
8	`LMDB_0.9.24`	39.482	+1.0%
9	`LMDB_0.9.22`	39.637	+1.4%
10	`LMDB_0.9.30`	39.681	+1.5%
11	`LMDB_0.9.18`	39.685	+1.5%
12	`LMDB_0.9.33`	40.183	+2.8%
13	`LMDB_0.9.19`	43.163	+10.4%
14	`LMDB_0.9.23`	44.051	+12.7%

Iterate Sequentially

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.27`	19.865	baseline
2	`LMDB_0.9.29`	19.952	+0.4%
3	`LMDB_0.9.31`	19.994	+0.6%
4	`LMDB_0.9.30`	20.082	+1.1%
5	`LMDB_0.9.33`	20.154	+1.5%
6	`LMDB_0.9.28`	20.256	+2.0%
7	`LMDB_0.9.18`	20.516	+3.3%
8	`LMDB_0.9.21`	20.546	+3.4%
9	`LMDB_0.9.20`	20.621	+3.8%
10	`LMDB_0.9.24`	20.742	+4.4%
11	`LMDB_0.9.23`	20.768	+4.5%
12	`LMDB_0.9.22`	20.788	+4.6%
13	`LMDB_0.9.19`	20.863	+5.0%
14	`LMDB_0.9.17`	20.912	+5.3%

Iterate Reverse

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.29`	19.300	baseline
2	`LMDB_0.9.21`	19.371	+0.4%
3	`LMDB_0.9.31`	19.377	+0.4%
4	`LMDB_0.9.19`	19.395	+0.5%
5	`LMDB_0.9.22`	19.449	+0.8%
6	`LMDB_0.9.30`	19.461	+0.8%
7	`LMDB_0.9.24`	19.475	+0.9%
8	`LMDB_0.9.28`	19.486	+1.0%
9	`LMDB_0.9.33`	19.516	+1.1%
10	`LMDB_0.9.20`	19.541	+1.3%
11	`LMDB_0.9.17`	19.591	+1.5%
12	`LMDB_0.9.23`	19.598	+1.5%
13	`LMDB_0.9.27`	19.656	+1.8%
14	`LMDB_0.9.18`	19.662	+1.9%

Calculate CRC32

Rank	LMDB Tag	ms/op	vs Fastest
1	`LMDB_0.9.23`	43.095	baseline
2	`LMDB_0.9.22`	43.234	+0.3%
3	`LMDB_0.9.24`	43.294	+0.5%
4	`LMDB_0.9.20`	43.306	+0.5%
5	`LMDB_0.9.31`	43.333	+0.6%
6	`LMDB_0.9.28`	43.343	+0.6%
7	`LMDB_0.9.27`	43.384	+0.7%
8	`LMDB_0.9.33`	43.429	+0.8%
9	`LMDB_0.9.30`	43.446	+0.8%
10	`LMDB_0.9.29`	43.453	+0.8%
11	`LMDB_0.9.17`	43.531	+1.0%
12	`LMDB_0.9.18`	43.543	+1.0%
13	`LMDB_0.9.19`	43.800	+1.6%
14	`LMDB_0.9.21`	43.857	+1.8%

Test Configuration

The benchmark was executed on 2025-11-06 using LmdbJava Benchmarks.

Each test run uses -Dlmdbjava.native.lib to ensure use of a specific LMDB library built from the identified LMDB tag. This isolates LmdbJava wrapper code during each benchmark, ensuring the only variation is the actual LMDB library.

All tests use the LmdbJava Agrona implementation with the following configuration:

Test Environment

CPU	AMD Ryzen 9 7900 12-Core Processor (24 cores)
RAM	62 GiB
OS	Linux 6.17.7-arch1-1 (x86_64)
Java	25.0.1
LmdbJava	`master#65df2ee`

Benchmark Configuration

Implementation: LmdbJava Agrona only
Test Profile: Run 4 sequential integer configuration (100-byte values)
Key Type: Sequential 32-bit integers
Value Size: 100 bytes
Access Pattern: Sequential
Benchmarks: All 6 operations (readCrc, readKey, readRev, readSeq, readXxh64, write)