Myhnx
Mgoing to make a thread
Lanyone know what’s up with this?
Jwhat wrangler changes are you referring to?
Lhttps://blog.cloudflare.com/pages-workers-integrations-monorepos-nextjs-wrangler the configure pages with wrangler part.
The Cloudflare Blog
Today, we’re launching four improvements to Pages that bring functionality previously restricted to Workers, with the goal of unifying the development experience between the two. Support for monorepos, wrangler.toml, new additions to Next.js support and database integrations!
W30 seconds of CPU time
WAnd you can set a lower limit if you wish on the dashboard
W10ms is I think still the limit for free plan
WOtherwise you're on Standard with 30 seconds
IKeep in mind this is CPU time
WIsaac any idea if free is still 10ms?
I?cpu-time
FCPU Time: the amount of work the CPU needs to do, measured in
Wall Time: the actual amount of time a Worker runs for, as though you measured it with a clock or your browser DevTools
msfetchWall Time: the actual amount of time a Worker runs for, as though you measured it with a clock or your browser DevTools
II believe it is, yeah.
ICPU time is the amount of work the CPU does, it's not the same as duration.
W^
IYou can have a Worker running for 30s but only use 10ms CPU time, or have a Worker running for 100ms but use 2s of CPU time.
IIt just depends on what you're doing in the Worker (or Function) itself.
WDuration is not related. It's simply an amount of time of actual compute
HCPU Time can never be higher than Wall-Time tho? Unless you count calls to other services?
II believe (correct me if I'm wrong) that CPU Time is measured in relation to running a single core at 100%.
I10s of CPU time = 10s of running a full core at 100%
IAnd if you're calling out to stuff in C++ land like hashing I expect it to use multiple cores?
II might be wrong though 
HOh, you meant CPU Time in general. I was referring to specifically on CF, where you only get single thread(unless you call out to a DO or smthn)
IOn multi-processor machines, a computer program can use two or more CPUs for processing using parallel processing scheduling. In such situations, the notion of total CPU time is used, which is the sum of CPU time consumed by all of the CPUs utilized by the computer program.
IYeah your isolate gets single-thread
IBut I expect things in
workerd internally, like calculating hashes and stuff, are multicore themselves?II don't know though.
HI would hope that you aren't billed for multicore ops because they are an implementation detail?
IWell you're billed for the amount of CPU time you use.
HIt would make it pretty difficult to figure out what your pricing might be
IIt uses the same CPU time on a single-core vs multi-core.
IJust multi-core uses less wall time because it's usually faster.
IHOh, I thought you said it multiplies it by number of cores?
INo, so if you cap 4 cores at 100%, then in 1 wall-clock second you'll consume 4 cpu-time seconds.
IBut if you wanted to do the same task on a single core, then you'd cap that single core at 100% for 4 wall-clock seconds and consume the same 4 cpu-time seconds (assuming the scaling was linear)
HI was also just referring to while you can pretty easily figure out single core pricing, by timing total duration length, then subracting I/O, it would be a bit more complex for multicore
IYeah... I'm not sure.
HSince while a request may take 5 cores 10 seconds to process, you don't know if it is utilizing all 5 cores the entire time
ISorry, I kind of went on a tangent - this isn't related to your use-case at all.
