How should hot-gate latency be compared with operational latency? #6
-
|
The benchmark artifacts include isolated hot-gate load-test results, while the README also documents broader operational pipeline latency. What is the correct way to compare these two latency profiles without treating them as the same measurement? |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment
-
|
They should be interpreted as different measurement scopes. The hot-gate load test measures only the synchronous decision path:
It excludes RPC, payload lookup, relay submission, persistence, dashboard IO, receipt polling, and realized PnL accounting. This makes it useful for measuring CPU-side decision latency and adversarial rejection behavior. The broader operational latency includes runtime overhead such as network IO, async scheduling, cache/pool lookup, payload construction, submission, persistence, and finalization observation. So the comparison should be:
The two latency profiles should be tracked together, but not compared as equivalent metrics. |
Beta Was this translation helpful? Give feedback.
They should be interpreted as different measurement scopes.
The hot-gate load test measures only the synchronous decision path:
It excludes RPC, payload lookup, relay submission, persistence, dashboard IO, receipt polling, and realized PnL accounting. This makes it useful for measuring CPU-side decision latency and adversarial rejection behavior.
The broader operational latency includes runtime overhead such as network IO, async scheduling, cache/pool lookup, payload construction, submission, persistence, and finalization observation.
So the comparison should be: