On 2025-11-22 12:15, Mathieu Desnoyers wrote:
...
The fact that the max inaccuracy is bounded means that we can perform
approximate comparisons in a fast path (which is enough if the values
to compare greatly vary), and only have to do the more expensive
"precise" comparison when values are close to each other and we
actually care about their relative order.
I'm thinking memory usage tracking, runtime usage (scheduler ?), cgroups
ressource accounting.
Another use-case worth explicitly mentioning is accounting the
resources (memory, CPU, I/O) used by a given entity (process,
cgroup) and enforce quota validation based on a configured
capacity limit.
This is a comparison with a limit which can be much larger than
the current resource usage value. Therefore, because the inaccuracy
is known, it is possible to enforce the quota limit with an
approximate comparison, and only use a precise sum when the current
value is near the limit, and rate-limit the frequency at which precise
limit validation is done.
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com
