4 ms·

Curious to know if anyone is running real production workloads with Cloudflare pages + workers. I'm asking because every time I've tried serverless offerings I'

Curious to know if anyone is running real production workloads with Cloudflare pages + workers. I'm asking because every time I've tried serverless offerings I've always been disappointed with the performance and debugging experience.

WORMS_EAT_WORMS 5y ago
> I've always been disappointed with the performance and debugging experience I think a comment like this deserves some details!
- holonomically 5y ago
  Very often I'll have some function that works fine locally but breaks when deployed to the serverless environment and figuring out why it's broken takes a lot more time than expected. Often it's because I was using some feature that for some reason or another is not supported in the serverless sandbox but other times it's memory or CPU limitations that are completely opaque and impossible to debug. For example, is it possible to use libraries written in C. Node.js has a bunch of libraries like that and it's never clear if those are supported in any given serverless environment. Most of these platforms also don't have actual hard limits so it's possible to use up whatever budget is allotted to a project and then get throttled randomly without any indication of what exactly happened or why exactly the limits where reached. Maybe things have improved since last time I tried these platforms but I'm doubtful. It's very hard to design easily debuggable systems when there are so many layers of abstraction involved and there is no way to actually set breakpoints and see what is going on in the live process.
  - kall 5y ago
    Well for workers these questions can be answered more straightforward than for lambda: Will a node/C thing work? No. Only third party packages that are cleanly isomorphic will work. The most exciting thing you get is fetch. You can compile things to WebAssembly I guess. is performance good? For HTTP response times it‘s top notch. For compute, if it matters then it‘s not a good use case imho. If you‘ve had performance issues with workers (not compute bound) I would like to hear about them, because I have put it into a "way to go when I want every answer to be reliably fast" bucket. Debugging? Well… it‘s a matter taste if what they provide is enough. Better than lambda but worse then a local server.

WORMS_EAT_WORMS 5y ago

> I've always been disappointed with the performance and debugging experience I think a comment like this deserves some details!

holonomically 5y ago
Very often I'll have some function that works fine locally but breaks when deployed to the serverless environment and figuring out why it's broken takes a lot more time than expected. Often it's because I was using some feature that for some reason or another is not supported in the serverless sandbox but other times it's memory or CPU limitations that are completely opaque and impossible to debug. For example, is it possible to use libraries written in C. Node.js has a bunch of libraries like that and it's never clear if those are supported in any given serverless environment. Most of these platforms also don't have actual hard limits so it's possible to use up whatever budget is allotted to a project and then get throttled randomly without any indication of what exactly happened or why exactly the limits where reached. Maybe things have improved since last time I tried these platforms but I'm doubtful. It's very hard to design easily debuggable systems when there are so many layers of abstraction involved and there is no way to actually set breakpoints and see what is going on in the live process.
- kall 5y ago
  Well for workers these questions can be answered more straightforward than for lambda: Will a node/C thing work? No. Only third party packages that are cleanly isomorphic will work. The most exciting thing you get is fetch. You can compile things to WebAssembly I guess. is performance good? For HTTP response times it‘s top notch. For compute, if it matters then it‘s not a good use case imho. If you‘ve had performance issues with workers (not compute bound) I would like to hear about them, because I have put it into a "way to go when I want every answer to be reliably fast" bucket. Debugging? Well… it‘s a matter taste if what they provide is enough. Better than lambda but worse then a local server.