OpenDJ vs OpenLDAP

🔬 Benchmark: OpenLDAP vs OpenDJ

Reference links:

• Online benchmark history
• GitHub Actions workflow definition
• JMeter load profile & data

Note

you need to be signed in to GitHub

Versions

Server	Version	Image
OpenLDAP	@(#) $OpenLDAP: slapd 2.6.10+dfsg-1 (May 29 2025 23:41:48) $	vegardit/openldap:latest
OpenDJ	5.1.2.0802aa4081e1c175c40bde1af59439a20fcc5670	openidentityplatform/opendj:latest

Totals (all operations, ADMIN_CONNECT excluded by the plan label)

Server	Throughput (tests/s)	Mean (ms)	Samples	Errors
OpenLDAP	1183.2	168.7	354870	0
OpenDJ	3124.3	63.7	940370	0

Per-operation latency

Operation	mean ms OpenLDAP	mean ms OpenDJ	p99 ms OpenLDAP	p99 ms OpenDJ	err OpenLDAP	err OpenDJ
ADD	52	62.6	350	132	0	0
SEARCH	20.5	55.4	259	123	0	0
COMPARE	12.1	53.8	105	121	0	0
MODIFY	33.2	54.5	129	122	0	0
BIND	973.6	86.1	1144	219	0	0
DELETE	37.9	65.9	254	141	0	0
READD	51.1	67.7	353	138	0	0

Total throughput (tests/s, higher is better)

Per-operation throughput is not charted: every op runs once per loop iteration, so each op's throughput just equals the loop rate. The meaningful throughput is the aggregate.

p99 latency per operation (ms, lower is better)

🟦 OpenLDAP · 🟧 OpenDJ — grouped bars per operation.

Notes

• BIND is the measured user authentication (test=sbind, single bind/unbind on its own connection) as mail=u_<n>,ou=People with the password set by MODIFY. The admin connection bind (ADMIN_CONNECT) is cached once per thread and excluded from these results.
• MODIFY sends the password in cleartext; each server hashes it on write with the same scheme ({SSHA}, Salted SHA-1) — OpenLDAP via the ppolicy hash-cleartext overlay, OpenDJ via its Salted SHA-1 default scheme — so BIND authentication is compared on equal footing.
• Full interactive JMeter HTML dashboards are attached as the jmeter-reports artifact.