hi

Did

Cloudfare

changed its name to

Cloudflare

?

is it possible to host a forum software like wordpress, nodebb , phpbb or something custom lightweight on cloudfare worker

it will use routes,

post.js

,

login.js

all will be hosted on different worker and connected by routers.

Using durable objects I guess you can probably achieve some form of cms/forum

spikes are usually from cold starts when the Worker script is being parsed. Cloudflare allows these kinds of CPU spikes, so it's not something to immediately worry about

300ms does sound a bit weird though

CPU time spikes aren't abnormal, no, though 300ms sounds extreme. The usual culprit is garbage collection -- you might have a dozen requests fill up memory, then another one comes in, tries to allocate memory, and triggers a GC pass. Cold starts aren't directly reflected in CPU time analytics, but sometimes they can be reflected indirectly, as the first request might need to initialize some state.

I thought all requests were isolated? So using the full 50ms and 128MB on one request doesn't affect the next which takes 10ms and 30MB?

Each Isolate isn't thrown away after each request, but it can handle more than one, just not concurrently. That's why the global scope can be used for some analytics collection to send batches of data. it won't persist long, nor any guarantees are made, though.

Just a small clarification: multiple requests can't run in parallel within an isolate, but they can indeed run concurrently. That is, they simply cannot execute JS at the exact same time within the same isolate.

Yeah yeah, obviously. Thanks for specifying though... I assumed it was known, but it might not be 🙂

Thanks for the clarification. For example if I was sending ~1000 requests per second, each around 30ms/50MB. Some would be queued until a previous request was completed? Would that trigger the 50ms/memory limits? Obviously a lot of those requests would be put into there own isolate and is scaled up etc.

Seeing the latency number when doing tests I would expect it to scale pretty much instantly to handle the full load in parallel. They just don't restart each Isolate between each request.

good morning friends 👋

Good afternoon! ☕ ☕ ☕

Ok thanks, makes sense and what I would expect. I was concerned the 50ms/128MB limits could be hit by just having multiple requests and they are somehow sharing the resources. But if thats not the case, its all good 🙂

btw, I'm going to start writing the January issue of the Serverlist newsletter (https://blog.cloudflare.com/serverlist/) today and going to start digging around the internet for content -- if you've read any interesting serverless/jamstack articles and want to share them here, I'd love to check them out 🙂

I think "The usual culprit is garbage collection -- you might have a dozen requests fill up memory, then another one comes in, tries to allocate memory, and triggers a GC pass." is what had be confused. Makes it sound like multiple requests to the same isolate would trigger the exception.

Thanks for your time! 🙂

Have you noticed a problem with network routing? - throwing around the world

@User For Cloudflare in general? Yes today I have actually noticed weird PoPs are being used when testing from https://tools.keycdn.com/performance

Seems to be changing quite frequently too

And with https://speedy.pages.dev/

London routed to Rio de janeiro and Belgium routed to New jersey

it never was cloudfare, probably a typo

Has anyone attempted tracing/zipkin in Workers?