Changes

Here are some important changes on the benchmark infrastructure and simulations that should be taken into account to interpret performance evolution.

Simulation Changes

2026-03-31 Bench various enrichers (NXP-33520)

Sim30Navigation

The navigation simulation now benchmarks a much wider range of enrichers. Previously, the simulation tested 7 request types; it now tests 20, covering most available document enrichers.

Distribution changes:

Request	Before	After
Get document	30%	13%
Get document folder	10%	10%
Get document dc only	10%	10%
Get document with acl	10%	3%
Get document with breadcrumb	10%	3%
Get document with thumbnail	10%	3%
Get document with properties	10%	10%
Get document with versionLabel	5%	3%
Get document with lock	5%	3%
Get document with audit	—	3%
Get document with collections and favorites	—	3%
Get document with firstAccessibleAncestor	—	3%
Get document folder with hasContent	—	3%
Get document folder with hasFolderishChild	—	3%
Get document with pendingTasks, runnableWorkflows, runningWorkflows	—	3%
Get document with permissions	—	3%
Get document with preview	—	3%
Get document with publications	—	3%
Get document with renditions	—	3%
Get document with subscribedNotifications	—	3%
Get document with subtypes	—	3%
Get document with tags	—	3%
Get document folder with userPreferences	—	3%

Expected metric impact: The overall Sim30Navigation response times will change because the request mix is now more diverse. Enrichers like audit, pendingTasks, runnableWorkflows, or permissions add server-side overhead compared to a plain document fetch. Direct before/after comparison of this simulation’s aggregate metrics with prior benchmarks is therefore not meaningful.

Sim20CreateDocuments

A call to addPreferencesOnParent() is now executed after each document creation. This sets user preferences on the parent folder via PUT /api/v1/{parentPath}/@preferences.

Expected metric impact:

• Creation time roughly doubles (~1min 30s → ~3min).
• Creation rate drops from 1300–1550 docs/s to 900–1150 docs/s.

This additional overhead is expected and reflects a more realistic workload where user preferences are set during content creation.

---

Infrastructure Changes

• 2022-11-21 Use MongoDB 6.0.2 and Kafka 3.3.1 for LTS 2023

• 2022-11-18 Use OpenSearch 1.3.6 instead of Elasticserach 7.9.2 for LTS 2023

• 2022-10-24 Benchmark Nuxeo LTS 2023 with Zulu JDK 17

• 2022-10-03 Change benchmark infrastructure from AWS EC2 to GKE

• 2021-03-22 Use Zulu JDK 11 instead of OpenJDK 11

• 2020-10-29 Upgrade to Elasticsearch 7.9.3

• 2019-10-08 Upgrade to Elasticsearch 6.8.3 OpenJDK 11

• 2018-10-18 Upgrade to Elasticsearch 6.4.2 OpenJDK 8 (previously 6.3.2)

• 2018-08-14 Wait properly on audit writes between simulation

• 2018-08-14 Make sure existing elasticsearch indexes are dropped on start

• 2018-08-13 Elasticsearch 6.3.2 all index in translog async

• 2018-06-22 Upgrade Ubuntu on elasticsearch and nuxeo nodes to 16.04

• 2018-06-22 Switch to Elasticsearch 6.3.0 (18w26)

• 2017-10-07 Use Elasticsearch Rest client instead of Transport client

• 2017-10-01 Update to Elasticsearch 5.6

• 2016-08-30 Upgarde Elasticsearch 2.x Nuxeo 8.4-SNAPSHOT/8.10 requires Elasticsearch 2.x, the cluster is upgrade from a 1.7.3 to 2.3.5

• 2016-05-09 New cluster for Jenkins slave New on demand c3.xlarge slave.

• 2016-04-27 Read simulation do not load external entities This was due to NXP-19581.

• 2016-04-27 Adding a new Search simulation The Navigation simulation is now referred as Read simulation