As first attempted in #304 (before that turned into a CI-focussed PR), the scanner server would benefit from using prefetch for incoming pixel recos (fewer http calls to mq broker, more messages per call).
Prefetching would likely not decrease the runtime of most scans, but it would speed up the initial ~80% (hand-wavy estimate) of the scan. Toward the end of every scan, the server is idle (waiting for messages) more often than it is actively fetching -- AKA there is a long tail.
See b250d43
Will need testing in the dev-production environment.
As first attempted in #304 (before that turned into a CI-focussed PR), the scanner server would benefit from using prefetch for incoming pixel recos (fewer http calls to mq broker, more messages per call).
Prefetching would likely not decrease the runtime of most scans, but it would speed up the initial ~80% (hand-wavy estimate) of the scan. Toward the end of every scan, the server is idle (waiting for messages) more often than it is actively fetching -- AKA there is a long tail.
See b250d43
Will need testing in the dev-production environment.