Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Containers are available in public beta for simple, and programmable compute

blog.cloudflare.com

Post reply on HN
Print view View original

11–20 of 23 posts

Re: Containers are available in public beta for simple, and programmable compute

#11

Post by blixt »

NathanFlurry wrote: post #10
develatio wrote: post #3

If my math is not wrong, running a single “standard” container during 1 month (non-stop) would cost ~55$. This looks extremely expensive for anything other than a simple demo/toy project. I can’t think of a reason I’d use this for heavy services instead of using [anything else]. Maybe I’m not seeing the use-case?

A 1 CPU + 2 GB of RAM + 50 GB ephemeral storage on Cloudflare Containers is $74.90. The same on Fly Machines is $31.00 (performance-1x, varies by region). Fly Machines has the same sleeping functionality as Cloudflare. Rivet Containers also has a similar price point of $29.40, but takes a different approach to sleeping (opts for optimizing coldstarts + autoscaling over snapshotting). (I work at Rivet)

I think Modal, which AFAIK has a similar feature set to Cloudflare Containers, also works out very favorably price wise compared to Cloudflare Containers.

Re: Containers are available in public beta for simple, and programmable compute

#15

Post by sofixa »

develatio wrote: post #6
aiisahik wrote: post #4

Earlier quoted context omitted.

I don't think you can calculate the cost of serverless compute this way. What containers do you have that run "non-stop"? If the container doesn't run any workloads, it doesn't cost you anything. Most of the compute i pay for sit idle most of the time. This is amazing pricing.

Say I want to deploy a service that is currently receiving 1rps at a constant rate, no upticks, no gaps. Wouldn’t that be the cost? If the answer is “yes”, then no, that is a terrible pricing.

Realistically, almost nobody has this type of usage. And for those that do, yes, serverless autoscaling up from zero is not appropriate.

Re: Containers are available in public beta for simple, and programmable compute

#16

Post by 0xy »

develatio wrote: post #3

If my math is not wrong, running a single “standard” container during 1 month (non-stop) would cost ~55$. This looks extremely expensive for anything other than a simple demo/toy project. I can’t think of a reason I’d use this for heavy services instead of using [anything else]. Maybe I’m not seeing the use-case?

And the gigantic AWS-tier bandwidth costs. This misses the mark by a lot. Classic example of pricing ruining a launch of decent technology.

It seems like always-on containers are not viable on this, so what's the point?

Re: Containers are available in public beta for simple, and programmable compute

#17

Post by develatio »

sofixa wrote: post #15
develatio wrote: post #6

Earlier quoted context omitted.

Say I want to deploy a service that is currently receiving 1rps at a constant rate, no upticks, no gaps. Wouldn’t that be the cost? If the answer is “yes”, then no, that is a terrible pricing.

Realistically, almost nobody has this type of usage. And for those that do, yes, serverless autoscaling up from zero is not appropriate.

True. Most services (unless your service is a demo/toy project, as I stated earlier) have way more traffic. It might not be evenly distributed though the day, but if you add all the CPU time in a day, I’m sure it will exceed 24h. So that leaves me with the question for who is this? Developers that are starting and want to deploy something small?

Re: Containers are available in public beta for simple, and programmable compute

#18

Post by Havoc »

develatio wrote: post #3

If my math is not wrong, running a single “standard” container during 1 month (non-stop) would cost ~55$. This looks extremely expensive for anything other than a simple demo/toy project. I can’t think of a reason I’d use this for heavy services instead of using [anything else]. Maybe I’m not seeing the use-case?

It’s likely aimed at bursty workloads. ie not one instance but a use case that fluctuates between 1 and 100 instances.
Post reply on HN