Platform Engineering
API Performance Lab
Reproducible before/after benchmarks for indexing, caching, pagination and parallelism.
The Problem
“We optimised it” is not evidence. Performance claims need a baseline, a change and a measurement.
How It Works
- Benchmark a full table scan against an index on 100,000 rows.
- Measure repeated product lookups with and without a cache.
- Compare a 50,000-product list endpoint unpaginated versus paginated, and sequential versus parallel processing.
Technically Interesting
The numbers are real and the weakest one is reported honestly: parallel processing gave only 1.29× because the test ran on a single CPU core.
Why It Matters
The same measure-first discipline drives the data-model, indexing and query work in my enterprise commerce projects.
Scope & honesty. Measured on a small, single-core test environment. Ratios illustrate the mechanism; they are not production benchmarks.