I plan an app that will include some location based live features. “Live” in this case does not mean true realtime, so Workers KV’s eventual consistency is a perfectly acceptable trade-off for the ease of scaling it provides.
Yet it would help a lot to get a feel for the “usual” technical characteristics of the list() operation of Workers KV to properly plan for user expectations and interactions.
I already have consulted how KV workers as well as the rest of the documentation provided.
In particular three questions remain regarding Workers KV list().
- What is the expected P50 / P90 / P99 time an individual
list()operation usually takes until it returns? Assume that thelist()has a prefix and will yield a result set of 4k5-5k entries, which only have metadata associated with them. - Suppose a key is written into KV with
put(). Now the only claim in the documentation is that that key will be updated eventually within 60s across the whole
CF network. But if a list()were to run in the same datacenter as theput()was performed, what would be expected P50 / P90 / P99 delays for that key to show up in the results of thelist()call? - Is there a way to get an in-advance count of the expected size of the result set for a certain
list()operation? That way I could adjust certain parameters beforehand in case of expecting many tens of thousands of entries over dozens oflist()calls. Such a result set would be too large to be helpful to any user, so it would make sense in such a case to apply a more narrow prefix straight away.
Furthermore, I’d be very interested to see the option to filter by age (key updated since timestamp T) and by metadata (metadata has key K / metadata key K equals value V).
TY for reading.
