● Benchmarks · how we measure
Every number has to show its work.
A benchmark on this site has to state its hardware, workload, tool and fairness caveats. This page holds that rule and the runs behind the numbers on this site.
● What every benchmark states
Five things, or it isn't published.
- Hardware
- CPU model, RAM, kernel, allocator and network mode (loopback or NIC).
- Workload
- URL distribution (zipfian, uniform or fixed), payload size, concurrency, run duration and warmup.
- Tool
- The load tool and its version, plus the exact command line.
- Fairness
- Thread count and config equivalence: caching, TLS, compression and rule evaluation match wherever they apply. If a feature can't be matched, we document the difference.
- Results
- Mean and standard deviation across multiple runs, not a single best-of-N.
● What we won't do
No ratios from contrived workloads.
- Compare one setup with a feature on against another with it off, and call the difference overhead. If a feature is essential to the use case, both setups run with their equivalent on.
- Compress identical content and call the result memory efficiency. That ratio reflects the test corpus, not the cache.
- Run a serial request loop and call it bulk-purge throughput. That measures the loop, not the server's bulk purge.
- Compare rules with empty bodies. That measures parser overhead, not rule evaluation.
● Speed Suite benchmarks
Published runs, with their setup.
Every Speed Suite number on this site comes from a benchmark in the module docs. Each run compares native Magento with the module on the same store.
| What changes | Native | With the module | Module |
|---|---|---|---|
| Full reindex, seven indexers | 44:25 | 7:21 | Optimized Indexers |
| SQL queries on a category page | ~160 | ~75 | Preloaders |
| SQL time on a category page | ~850 ms | ~46 ms | Preloaders |
| Magento bootstrap | 25 ms | 2 ms | STOP Protocol |
Reindex test store
Mage-OS on PHP 8.3, MySQL 8 and OpenSearch 2, with 131,250 products, 615 categories, 2 websites, 6 store views and 12 customer groups. Row counts in the index tables matched native Magento on all 43 checks.
How to read them
These are test runs, not a promise for your store. Your numbers depend on hosting, theme, extensions, catalog shape, rules and traffic.
● Trident
Test Trident on your own traffic.
A cache proxy's numbers depend on your URLs, your hit rate and your backend, so we publish a Trident figure only with the full methodology above.
The best test is your own traffic: try Velocity Engine free for 14 days.
● Other promises
Compliance and compatibility, within their scope.
Compatibility. A tick on the compatibility page means the product is flagged compatible. A dash means it doesn't apply to that module or isn't confirmed yet, so ask us before you plan around it.
Compliance. Our compliance extensions give a store tools such as consent records, safety notices, price history and packaging data. They don't make a store compliant on their own, and they aren't legal advice.
● Choosing a tool?
Not sure which product fits?
Start with the guide, or send us your stack and what you measured. An engineer replies within one business day, usually in under 4 hours (Mon–Fri, CET).