I would expect 10 people in the same building to have 1 miss and 9 hits, but it seems like that woul
I would expect 10 people in the same building to have 1 miss and 9 hits, but it seems like that wouldn't really be the case here?
DIf there’s two nearby PoPs, you would have 2 misses and 8 hits probably.
You may want to enable tiered cache if you’re hoping to reduce requests to the same origin.
You may want to enable tiered cache if you’re hoping to reduce requests to the same origin.
Iooh interesting. that can't happen at the worker level, right?
Dyou enable it on your dashboard. https://developers.cloudflare.com/cache/how-to/tiered-cache/
The When you do your
I don't think a Worker can force in or out of tiered caching (it can only force in or out of caching entirely).
The
fetch in your Worker will automatically use tiered cache if you have it enabled. When you do your fetch calls, you can opt to bypass CF's cache completely. if you want.I don't think a Worker can force in or out of tiered caching (it can only force in or out of caching entirely).
Tiered Cache uses the size of Cloudflare’s network to reduce requests to customer origins by dramatically increasing cache hit ratios. With data …
Ithank you!
DWe use tiered caching on Worker subrequests to Backblaze B2; saves a ton in terms of latency and cost.
Ibut you have to have that enabled at the "website" level first, right?
Iit's not a worker-level config?
CI never messed with workers/pages yet, almost curious how they could help with a broadcasting service like Discord, Zoom, etc.
DIty!
Dyou will also need to have the Worker listening on a route (or custom domain) for the same domain that you have tiered caching enabled
Di.e. not workers.dev
CWhat are you guys using workers for that substantially helped you? I mean for people with more than 20K unique users daily.
Dwe have less users than that and Workers are still insanely useful.
CI'm kicking around 4K unique daily, only about 1% CPU on the main single core HTTPS server ( I run many services though as users navigate)
CBeen eyeing some of the CF toys you feel?
De.g. what I was working on today is our Worker that acts like a middleware for Azure AI Document Intelligence. The Worker fixes a lot of annoyances/missing features, like storing of results and caching.
CHmm more detail? Unsure of azure AI stuff.
Dit's Microsoft's OCR solution
CI'm mostly running many-to-many broadcasting services like Zoom/etc conferencing stuff.
Coh okay
CI'd love to find a cool way to include these serverless functions, still kind of on that stateful side of things.
COnly thing I can use is a CDN for most part. :X
Dif our python scripts talked directly to Microsoft, we would be paying $10 per 1,000 requests, even if we had previously OCR'd the same file. It would also take several seconds (because it would process the same file again from scratch).
CDisgusting pricing...
CI hope this would be ten folds better for you then and realistic.
C1,000 requests I rip through in minutes if it were me..
Deh, it's kinda fair when you consider how good it is. It's just stupid that Microsoft doesn't serve you the same result when you make the same request lol. So we built a Worker that fixes that for us.
CWell how many requests you really blasting?
Dhttps://developers.cloudflare.com/durable-objects/reference/websockets/
^ This might interest you then.
^ This might interest you then.
WebSockets allow real time communication between a client and server. Both Cloudflare Durable Objects and Workers can act as WebSocket endpoints – …
CI'm thinking my couple thousands of cocurrent users would blow that to smitherenes.
Clol
CYes my world is WebRTC + WebSockets + HTTPS
Da couple million per month.
CI hope you see a discount at those prices per 1,000 requests
Cor are paid properly for it
CThat's brutal, I almost got smoked with some serious bandwidth pricing starting out with my stuff; literally using over several TB a server a day.
Dit's actually $10 per 1000 pages, so worse than that lol.
CYeah I hope you got the costs sorted out that's for damn sure
Chaha
Dwe have <1,000,000 unique pages. iirc we have like 100,000 unique pages indexed maybe?
Cslowly reading this durable objects with websockets
DI think we've used $2k-$3k of Azure credits on OCRing.
CIf it works it works
CMy prices would had been like 10K+ a month for several hundred users at a time.
Dit's faster and easier to just hit the Worker than maintain a database to track what has been OCR'd vs not OCR'd yet.
CFound good hosting it's like $400CAD for it all
CDatabase is for shared state, so worker can share state for other workers?
Cor not being used that way.
Dwe use Backblaze B2 (could be R2 if you prefer) for this specific dataset; we kinda use it like a key/value store.
There's Cloudflare D1 too that we do use in some of our other Workers
There's Cloudflare D1 too that we do use in some of our other Workers