Percona's Quick Test Finds MongoDB 9.0.4 About 6% Faster Than 8.0 and 8.3 When CPU-Bound; MongoDB Cites Up to 20% Higher Transactional Throughput

Overview

Percona has published what it calls “a first, quick test” of MongoDB 9.0, and its headline result is smaller than the best-case figures in MongoDB’s launch. In its November 6 post, Percona says MongoDB 9.0 “used less CPU per operation than 8.0 and 8.3 with every data size,” which made it “about 6% faster when the CPU was the bottleneck, but only about 2% faster when the data fit comfortably in memory or when the disk was the limit.” MongoDB’s own figures, as Blocks & Files reported them, include up to 20% higher throughput for transactional workloads compared with 8.0. The two sets of figures come from different tests, and the cited sources do not make them directly comparable.

What MongoDB Claimed

MongoDB 9.0 was announced on September 29, 2026. TechTarget described it that day as “now generally available.” According to Blocks & Files, MongoDB listed these gains over v8.0: up to 35% faster find-one queries, up to 30% faster update-one queries, up to 2x throughput on large instances, and up to 20% higher throughput for transactional workloads. TechTarget attributed the performance gains to “tests conducted by MongoDB” and gave no figures. Blocks & Files and TechTarget both report MongoDB’s claims; neither says it tested them itself.

On October 9, Percona listed the same four figures and wrote that “at the time of this writing I could not find any details about the methodology used or the raw data of the tests performed.” That is Percona’s statement about what it could find on that date, not a finding about what MongoDB has published since.

What Percona Measured

The test, written up by Percona’s Ivan Groenewold, compared MongoDB Community 9.0.4 with Percona Server for MongoDB 8.0.34-15 and 8.3.13-4. According to Percona, the Percona builds come from the same upstream code and “none of the PSMDB-specific features were enabled in order to keep the configuration as similar as possible.” Each version ran on its own three-node replica set deployed on Google Cloud.

The workload was the built-in “flights” workload of the PLGM load generator: 60% finds, 20% updates, 10% deletes and 10% inserts, driven by 16 workers with no think time and the default write concern. Percona ran it with 1 million, 4 million and 8 million documents. Its reported results:

• 9.0 used less CPU per operation than both older versions at every data size, and the query plans were the same on all three.
• Percona says 8.3 “was the slowest of the three whenever the CPU mattered.”
• In its per-operation comparison of average latency, with 4 million documents “finds show the biggest difference (+12%).”
• With 8 million documents, single-document updates and deletes by flight_id were “the same on all three versions or even slightly faster on 8.0.” In that run the servers did not finish a full pass of the workload in 30 minutes, which Percona says the load generator flags as incomplete.

Percona describes the test as “one workload, one hardware configuration for three different data sizes” and recommends treating 9.0 “as a release to test, not one you have to deploy urgently.” It says it will publish a follow-up.

Why the Figures Are Not Like-for-Like

This comparison is the Herald’s own reading of the cited sources. MongoDB’s figures are stated as “up to” gains over 8.0, and none of the cited sources says which workload they came from. Percona’s figures describe a single mixed workload and compare 9.0.4 with two Percona builds, including 8.3, which does not appear in MongoDB’s list as reported. Percona used one hardware configuration, so its test does not address MongoDB’s large-instance figure. A smaller number in Percona’s test therefore does not by itself contradict MongoDB’s ceiling figures.

Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget, said in the TechTarget article that 9.0 and the Atlas Infinite architecture announced alongside it are both significant additions, with 9.0 delivering substantial performance improvements.

Who Ran the Test and Their Interest

Percona makes its own MongoDB distribution, Percona Server for MongoDB (PSMDB). In its October 9 post it wrote that it had “already started working on Percona Server for MongoDB 9.0” and that it “continues to make enterprise-grade capabilities available without an Enterprise Edition license.” The test itself used no Percona-specific features, according to Percona, and it published its configuration so others can repeat it. In that post its author called 9.0 “useful” without seeing changes that warrant rushing to upgrade, which is an opinion, not a measurement.

What We Don’t Know

• Which workload and settings produced MongoDB’s “up to” figures; none of the cited sources says.
• How 9.0 compares with 8.0 on larger hardware, since Percona tested one configuration.
• What Percona’s promised follow-up will show; Percona says it plans to repeat the test, try different levels of concurrency and faster disks, and look at why 8.3 needs more CPU.