What do you mean by “using cache?” Do you mean in-memory?

I mean using the cache worker api

cache.put(request, response)

method

Ahh.. DOs don't run in every colo. So obviously depending on a million other things, there could be scenarios where you want to cache the output of a DO in the cache of the Worker's colo.

If we assume the user calling that DO will always be nearest to the same colo, and the DO just returns something from persistent storage, is there any performance benefits of using the cache api?

basically what is the added latency of spinning up a DO + routing the message to it + persistent storage versus cache

Like I'm achieving 50-80ms requests for this particular DO for example

Cold start is 300ms+

That depends greatly on how close the nearest colo with a DO is

And what do you mean with cold start? The first time you create a DO for a particular user?

Yeah

or even if the DO has been evicted, there's a slight delay, no?

Yeah.. if you use

ifFromName

that can take a while because of the global coordination required

Of all the overhead that come with DO, by far the most impactful one will be latency between the worker colo and the DO colo

(Besides the initial creation based on a name that is)

Right now, most DO clusters are in the US and Europe.. so average latency will be low there..

Where as in the Southern Hemisphere, it gets to be more

I see, that makes sense. The distance between the worker and DO seems to be the deciding factor

so that's where the cache comes into play

Yeah.. with the speed of light still being fast, but also rather finite 😉

So yeah.. depending on your use-case, you could end up wanting to use the cache

I think I do want to use cache, but there's just one problem. It's critical in my particular use-case that the first load is fast regardless of location. I need to cache and potentially refresh/purge data in the cache

I haven't been able to find a suitable solution to this

I need to pre-populate the cache and then refresh/purge when changes are detected on the backend

The initial load needs to be sub-50ms

that is my main requirement

I would need to know a bit more of your requirements to give you a decent answer. But it gets to be a bit much to do that here. Can you DM me a paragraph or two what you need to do? Where the data is coming from?

Unless you can wait little over a week and want to work with @User to present this as a case to me on our next Workers for Fun and Profit episode 😄

Yeah I might be able to drop in! Thanks for the help!