Sorry if this is a dumb question - but just starte...
# spicedb
j
Sorry if this is a dumb question - but just started playing around with Authzed With a very small set of relationships (10's) - permission checks on minimizeLatency are showing up at 30ms for checkLatency for joseph_example. 1. Is this graph e2e latency as observed from the client or just server side? I'm calling Authzed serverless from CA. 2. What is the graphed latency: avg, p99, max? 3. If this is just server-side latency, is this latency expected?
d
Hey @Jospeh - today SpiceDB Serverless only supports one region (Iowa) . AuthZed has another product called SpiceDB Dedicated that can get you a permissions system wherever you need it (today we support GCP and AWS). Feel free to DM me if you’d like to learn more.
j
Oh yah sounds good - I remember we discussed this yesterday. But I was curious - even if I ignore network latency - are there any performance SLA's on SpiceDB?
For this graph: 1. Is this just the server processing time? (I suspect so as SF -> Chicago is usually ~60ms RTT) 2. Is this avg, p99, max latency? 3. I had set the consistency to minimize_latency - is there anything else I can do to speed it up?
d
Hey @Jospeh sorry for the late reply! It depends on what the client side sets as their consistency requirement and on the cache hit rate - the range I’ve seen in SpiceDB Serverless and SpiceDB Decidated is 3ms to 25ms.
That’s a good question on the graph - @vroldanbet do you know what we’re using in the above screenshot for serverless?
Minimize latency is the way to go @Jospeh - next up would be deploying SpiceDB Dedicated which would interconnect privately with your environment and sit adjacent to it. Later this year we’ll introduce SpiceDB Cloud which will allow you to deploy closer to your application as well.
v
@dsieczko that'd be a question for @Joey . I think those are server-side metrics from SpiceDB prometheus endpoint.
j
yes, it is server processing time
v
is that a p99?
we should probably make it clearer what percentile it is
j
P95
@Jospeh as for speeding up checks, it really depends on your schema and how cacheable things are