#Latency on worker and Durable object
11 messages · Page 1 of 1 (latest)
There is no guarantee that the DO is in the same location as the worker instance
Not every datacenter is capable of holding them
Depending on what you use the DO for, latency can also vary as it might have to do a global lookup first
I think this is especially the case when first creating a new instance using an ID derived from a name
How can we ensure consistent sub-100ms response times when using Durable Objects, given that both the client and the Durable Object are located in the same region? We've observed response times exceeding 100ms and are seeking ways to optimize performance.
@pseudo iris we are new to DOs. We’ve never seen worker-to-DO latency be less than 120ms no matter where or how we create them. Is your experience similar to ours?
DOs are geographically pinned to where theyre created, so if two users from different parts of the world want to use the exact same DO instance they will have to travel to the one instance to use it, which will result in different and sometimes long latencies
I personally have never measured it because DOs are expected to have more latency to them and I have never found it to be noticable
DOs and KV always have the tradeoff of consistency vs latency
You will never achieve the same latency with DOs as you do with KV