Hacker News new | ask | show | jobs
by mrj 1106 days ago
Porter looks cool and I would love to not hand roll kubernetes stuff.

But the pricing looks curious. Their "default infrastructure" is $300 a month set up in AWS plus usage costs on Porter. It's already 6x what I'm spending on Heroku, why would I pay usage costs per CPU to Porter if they're not running any CPUs?

Honest question really because I'm interested, it just seems far out of line price-wise. I would totally pay for help managing infrastructure, just not server usage. It would make more sense to pay for the things I'm actually using or a fee per developer.

4 comments

Co-founder here. Thanks for asking this - we get this question a lot and have tried our best to address it in the FAQ section of our pricing page (https://porter.run/pricing). Infrastructure provisioned by Porter is constantly managed by Porter so that the end user does not have to worry about the details of maintaining a Kubernetes cluster. In this regard, our pricing structure is identical to traditional platform as a services like Heroku, Render, Fly.io, or Vercel, where the respective PaaS provider is effectively charging a margin on top of their own cloud resources on AWS/GCP. This form of leasing out the PaaS provider's AWS/GCP infrastructure to end users also often results in rapidly increasing costs especially as you start hitting scale on hosted platforms like Heroku.

The value you get from Porter is the same as these platforms: with the clusters plugged into our internal system, our team monitors your infrastructure and owns its reliability so the end user can just focus on the application. Other platforms that also run in the customer's own cloud, such as Red Hat OpenShift, has similar resource based pricing models.

On this note, we offer an option to bring your own Kubernetes cluster and connect Porter to it. In this case, the end user manages their own infrastructure and Porter purely acts as a UI on top of your own Kubernetes cluster. Accordingly, for this use case, pricing is based on the number of user seats instead of resource usage. Many of our customers who use Porter in this capacity often have their own DevOps teams that manage the infrastructure and use Porter purely as a means to simplify deployment for their developers.

It's also worth mentioning that Porter is not meant for small workloads or hobbyists. If your spend on Heroku is less than $300, we actually do not recommend using Porter and have advised many users who were interested in Porter to use platforms like Render and Fly.io. Porter is designed for companies - mostly startups - who actually need to scale their applications and have sufficient budget for their cloud infrastructure.

It’s not super important to your point (altho it goes towards pricing I suppose) but Fly doesn’t resell a public cloud, but instead owns/rents hardware racked in a bunch of regions.
> our pricing structure is identical to traditional platform as a services like Heroku ... often results in rapidly increasing costs especially as you start hitting scale

Hi, thanks for the reply and congrats on everything you've built. I am your target demo I think, an early stage startup that's just starting to get traction. I'm expecting costs to increase soon and that's a big motivating factor to getting off Heroku.

So I'm stuck with either doing all this yaml junk myself or something like your solution. But it is definitely not something I would consider if I'm going to run into those escalating costs that you mention.

A gui / guided kubernetes could be just the thing though.

If you're an early stage startup, you certainly are our target demographic - for startups with less than 5M in funding, we also offer a startup deal where you can get $10k in credits from us (you can apply from our pricing page). If you have any cloud credits on AWS or GCP, which is fairly easy to get as a startup, along with this deal you can effectively use a PaaS for free because the infra runs in your own cloud account.

Also just to clarify, our pricing structure is designed to increase more slowly as you scale. We offer volume discounts above a certain resource usage, so that your cost increases more logarithmically rather than exponentially.

Not sure about Heroku, Render, and Vercel, but Fly.io doesn't resell AWS/GCP.
Founder of Argonaut (yc backed as well) here.

That's the model we use at argonaut.dev. We charge a flat fee per developer and you pay the AWS bills to AWS or GCP (using credits ideally). We settled on this because it is transparent and does not lead to any surprises in billing, especially for startups which are our core target.

We have customers who have evaluated and chosen us for our flexible CI/CD pipelines, transparency, and pricing. We also help setup other infra dependencies you might have like managed databases, redis on aws and gcp.

> Porter looks cool and I would love to not hand roll kubernetes stuff.

    cat ./yaml/applications/kubernetes-dashboard.yaml | ./scripts/kubectl-apply.sh && ./scripts/run-argocd-sync-and-wait-pipeline.sh "kubernetes-dashboard"
What am I missing?
Hi, CEO of Qovery here - the same business model for us as well - per active user using the product.