I’ve set up a very simple app where I can create and list rooms for bookings.
I have a test demo with 3 rooms in my collections.
Most of it, setup through a GraphQL schema upload.
It works great but I’m worried about response times I’ve been getting with very basic operations.
I have a allRooms query that returns a paginated set of rooms and I’m getting sometimes 4 sec request times. Seems like once it’s hot it gets faster and reached 800ms range but still that’s pretty darn slow.
Is it just what to expect with fauna, or is it because I’m using free tier?
It is not expected to see such high latencies. We have had some issues over the weekend which could have contributed to this. Can you please try again to see if you get same latencies ? Happy to troubleshoot more if you could share email id on the account.
I’m in a situation that sounds very similar to Christophe’s. I have some straightforward, paginated graphQL resolvers that use simple indexes. Straight-up FQL based endpoints are not faster. I’m accessing the Fauna graphQL service via a Vercel serverless function that my Vercel-hosted frontend makes calls to. My email is “k at blwsk.com”.
As you can see below, response times average >1s, which is concerning to me as I validate Fauna for my use case.
@Christophe_Verbinnen we have identified couple of places where these additional latency is coming from and will be working through the improvements in next couple of months. Till then, ~400ms to ~1s would be a norm on the GraphQL API side. This behaviour is not related to the size of collection.
On the drivers users can specify a timeout but that is yet to come on GraphQL API.
@skris GraphQL API internally calls FQL and one of the identified issue is the inconsistency in latencies on the FQL side. We are planning to fix this in next few months.
@Artemis69 Did you get feedback from Fauna on this?
@Jay-Fauna Jay, could you shed some extra light on this issue? I’m still experiencing latency sometimes longer than a second a simple write. I’m curious if this is something that will improve, or not.
The “identified issue” noted above was resolved by the end of October. That issue affected all requests, not just GraphQL ones. Remember that the GraphQL API is its own service, so it adds its own overhead to always run your query against the latest schema in a stateless way.
Since then, several additional performance improvements have been released and we do see a good trend in the average latency as well as the 95th and 99th percentile latencies. The remaining latency spikes are related to the consistency guarantees that Fauna provides. That said, it is understandable that it can still be surprising and “weird”.
Our engineers are always on the lookout for improvements, both in the core of the database and for the GraphQL service. We already have some on the roadmap, but we can’t comment on timelines to release them.
The original post concerned the event from last September/October, so I will close this particular topic to avoid confusion.