DO/D1 smart placement
If it is 1 -10 ms latency, thats wonderful. Based on the earlier conversations we have had regarding about DO locations (where do DO live links), DO's are available only in a few DC (For example SGA for an Indian audience). And the minimal latency for DO location from MUmbai would be 200 ms according to the web site. I am making some assumptions here a. That the D1 replicas when they come into pic after Q1 would still live near SGA like locations due to colo reqs 2. Not just initial connections to DO, but subsequent latency for all communications would be the same. In that case, how can the latency be lower (like 10 ms) without smart placement? I would say having these many pops actually is a menace for performance as compared to bettering it. Anothe rpoint to ponder and I am thinking out aloud -> about a use case where I depend on an R2/KV and a DO ( i have such a use case) . I am going every where to collect my objects and the initial request would take multiple latency hits, I cache them using Cache API after the initial, but again theres no guarantee my next customer would hit the same pop.
7 Replies
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
@Max (@rozenmd) - inviting you here, so you can spread your expertise 🙂
?
The latency related concerns, are they warranted? Is the understanding about the worker connections to D1(thru DO) - is the understanding incorrect?
I don't understand what you're asking tbh - the comment you're replying to was about the difference between a pure db provider and Workers - Workers has overhead that the pure db provider doesn't
Unknown User•5mo ago
Message Not Public
Sign In & Join Server To View
For now they do, yes