OcaltQL Real Tough Benchmark: four parallel full-stack executions in the ~460 ms class
A production stress test across database, string, mathematical, array, temporal and parallel-control domains, repeated four times concurrently. All gates passed; network remained the dominant cost.
On 21 August 2026 the OcaltQL runtime was put through a production stress test on the public STARTER endpoint. The workload was defined once as a user-defined operation, then invoked four times concurrently and collapsed into a single result.
What was exercised
Three tables were created and seeded with eighteen rows, then queried with two inner joins, a left join, aggregate counts and a bulk status transition. Alongside that: a string pipeline through uppercase, reversal, two replacements, length, split, join and search; power, square-root and factorial mathematics; array construction with deduplication and reversal; live temporal binding including age-in-years and a UTC conversion; a six-node parallel fabric collapsed and indexed; and five levels of nested conditional gates.
Every phase passed. The final composite signature was 311394323.
What the numbers say
Runs of this class complete in roughly 460 milliseconds end to end. That figure includes the network round trip, TLS, request parsing, four parallel full-stack executions and the response transfer. The engine and database portion is a fraction of it.
That is the result worth stating plainly: with schema creation, eighteen inserts, multiple joins, a bulk update, factorial-level arithmetic, string pipelines and four concurrent invocations, the dominant cost is still the network. Not the engine, not the storage.
Why the signature matters
The signature folds the mathematics, the database cardinalities, the string metrics, the fabric size, the computed age and live temporal components into one value. It could not have been produced from cached or mocked state — the clock was read at execution time, and a substituted clock would have produced a different number.
Method
Executed against console.ocalt.com on the STARTER plan. Internal phase timings are not exposed by the runtime, so no per-phase breakdown is claimed. The language reference is at ql.ocalt.com.