I'm not looking for a benchmark or something like ...
# spicedb
l
I'm not looking for a benchmark or something like that (would be good, but not a necessity). It's more like some real example (use case) or someone that have experience with a large number of resources (10 million+ for example).
j
Zanzibar itself (on which SpiceDB is based) has numbers well, well, well in excess of that
it should also be noted that the number of relationships isn't as important as the shape of the data
if you have 10M relationships representing 10M documents, each with one user, the checks will be insanely fast
if you have 10M relationships representing 1 document, it'll be a bit slower (not much, mind you)
the real performance cost is due to fanout and depth
l
great! I understand... and is there any "magic number" of depth that would be interesting not to exceed? of course the less the better, but... is there any number that would be a "warning"?
j
SpiceDB by default simply prevents beyond 50
that's configurable, of course
l
nice, thanks for the quick reply!
j
anytime!