Using Durable Objets for Assets - Fast Deployment - Good or bad idea?

So it’s a multi-regional central instance.
Is this central instance hosted on a single continent, or on multi-continents ?

For comparison GCS stores data on a single continent

I suppose you reference this post :


I’ll give a try.

So Workers Sites knows when the assets are reliably available on KV.
And it hits the central instance only one time, implying the central instance is expected to be strongly consistent. (Just like Cloud Storage). Am I right ?

if it is the case then there no need to build asset stuff over DO.

Do you know if there is a source code somewhere so I can check what’s going on during the deployment ? And check what are the assumptions taken by Worker Sites ?

From the post you linked:

“Recently, we have rolled out a major internal change. We have added a second source of truth on the European continent. These two sources of truth will still coordinate between themselves, ensuring that any data you write or update will be available in both places as soon as possible. But latencies from Europe, as well as places closer to Europe than the United States, should be much faster, as they do not have to go the full way to the US.”

First, check the “How it works” section of the blog post you linked to: https://blog.cloudflare.com/extending-the-workers-platform/

If you still want more detail, the implementation of Sites in wrangler is open source: https://github.com/cloudflare/wrangler/tree/master/src/sites

1 Like

Very informative, thanks !

The post dealing on KV latency improvement says :

any data you write or update will be available in both places as soon as possible

I understand it is either a double write on both Masters, or a Master-to-Master replication. Do you have any idea on the delay implied by “as soon as possible” ?

More like 50ms, 500ms, 5s, 50s ?

It’s typically hundreds of milliseconds.

1 Like

Great !

Looks like it’s possible achieve the desired effect with KV.