#Dagger Cloud pricing 🚲🏠

1 messages · Page 1 of 1 (latest)

still jacinth
#

🧵

#

@idle grotto @inner rain @solar gale 👋

#

I don't expect this thread to ever be "resolved", from my experience on pricing discussions... 🙂

inner rain
#

haha, truth

still jacinth
#

A few general thoughts:

  • If you're running 6000 minutes / month of CI and you're still on the cheapest GHA runner, you're an exception. Much more likely that your pipelines started getting super slow around that scale, and you're solving hte problem by throwing compute at the problem.
#
  • In that more typical situation (5k - 10k minutes / month, overpaying for compute, CI generally broken and painful) then not only is $250 - $500 / absolutely a no-brainer to make the pain go away, it may actually pay for itself by chopping your compute spend by a lot
idle grotto
#

There will more likely be way less dagger cloud minutes once we move to dagger in gha since we'll be billed by dagger engine minutes not by job execution minutes

still jacinth
#

Yeah it doesn't map perfectly to machine minutes. But in practice it converges. The difference has never been an issue in practice for our early customers (the largest is around 100k minutes/ month)

still jacinth
#

Of course over time, we hope you expand to daggerizing more and more - but that only happens once you've decided you're getting enough value for the price 🙂

idle grotto
#

Understand I'm coming from a place of pitching dagger to my managers, so pricing more concerns me in the way of convincing them of the value that dagger adds and if the additional value of dagger cloud is worth the price for it. In most cases, they will consider that value of dagger to be separate to the value of dagger cloud.

still jacinth
#

Dagger Engine is not all of Dagger

#

That framing will help a lot

#

cc @dusk ore @whole rampart

idle grotto
#

I think I'm going to need to do some more thinking and experimenting before I can talk further about what I think about the pricing.

inner rain
#

it's very hard to pitch a price that is 10x the underlying hardware, we could throw a lot more metal at the problem for that price, also knowing that we are already highly parallel, so 10x is the lower bound, likely much higher in practice

idle grotto
#

^ This use case seems very different from mine, probably why pricing is so hard 😛

still jacinth
#

(and/or using the Daggernaut plan to get telemetry)

inner rain
#

we threw CUE at it to build a a really nice caching like thing (at a higher level than container layers)

whole rampart
mortal ether
#

As another data point for pricing, anecdotally, I really just want visualization for my pipelines.

My current experience with dagger and CI runners is that all the output of a dagger run is dumped into one step in a CI pipeline - unless I'm doing something wrong as I am a beginner.

I currently use dagger for personal use and want to setup automated CI/CD with my homelab and Gitea actions or Drone and with Dagger cloud serving as the visualization of the pipeline.

I don't know what pricing structure is best suited to that use case but it's where I'm at.

inner rain
#

yeah, totally, if there was a visualization only plan on the cheap, that's what we are mainly missing to consider adoption, something on a per seat basis

#

saving a dev even 15-30-60 minutes a month dealing with build logs when failures happen would be the ROI point ideally

whole rampart
#

I think for your particular case @mortal ether the upcoming single-player "Daggernaut" plan might make more sense.

Having a viz-only team plan makes sense, appreciate this feedback @inner rain

still jacinth
#

One complicating factor for a per-seat plan is that we don’t know where to draw boundaries between visualization features that require a paying seat, and those that don’t. If your product is open source and you want all open-source devs to visualize pipelines, do they all have to signup to see anything? Do they each need a paid seat? And if the answer is “no that should be free” then how do you determine who needs a paid seat and who doesn’t?

#

On top of that our web UI is still young and evolving fast. Not having a per-seat pricing allows us to not worry about what a user can or cannot do or see

#

Lastly per-seat pricing can cause friction: “I want to invite my team to use Dagger with me but that would cost X more, maybe I only invite one person for now”

inner rain
#

one benefit of a viz only plan is that you can get people on the platform, then show them where they can save money (self service upselling?)

#

I got an "that's insane" response to the pricing, so we cannot even consider Dagger with this client. No sense in doing the experiment if the logs are going to be difficult and the price to have them nice is out of reach

rich silo
inner rain
#

I guess when it comes to having nice logs, maybe bytes processed is a better metric

idle grotto
inner rain
#

Love Dagger and the possibilities of the DAG, but logs is a definitely blocker to wider adoption for any of my use-cases, and if that is the only price to get nice logs, it's gonna be sad

still jacinth
still jacinth
inner rain
#

one thing to note about me, there is the hof project, and then client projects, and I tend to talk about both as if they are "my" projects

#

up to now, it's mainly talking from my main client perspective, where we have talked about adopting Dagger, we have a need to refactor a bunch of builds and haven't decided how yet. So here, we are talking 10+ devs

#

for hof, I want to build dagger into some of the CLI functions, but the logs in their current state are not really presentable to users, I'm not sure how this one would work out, real edge casey for a Dagger pricing bikeshed

mortal ether
#

I don't know how I feel about the Daggernaut pricing. I don't know what the appropriate price level is for it. Paying $60 a year for just better visualization is a little steep for a hobbyist toying around in a homelab environment.

inner rain
#

I'm happy to pay the price of a coffee a month for nice logs personally, it will quickly pay for itself compared to searching for the fails on GHA

still jacinth
#

It would probably be $42 for the year if you pay upfront but I hear you @mortal ether

idle grotto
inner rain
#

The Dagger TUI is fine and all for local dev work, the problem is CI

#

for our hof CLI, there are places where we want to run Dagger behind the scene, but will need to present logs (basically think of running tasks in a container via dagger, rather than directly on the host)

mortal ether
# still jacinth It would probably be $42 for the year if you pay upfront but I hear you <@140339...

That is a little more appetizing. Really, my usage would be really slim - I have no idea how much I'd cost Dagger the company as a single user that is just using visualization for a couple of home projects. My fly.io bill before self-hosting was $0 for Discord bots and small web projects. I mean, $42 is very affordable but there is a point where I'm like "eh, it'd be neat to have a nice UI for my CI runner Dagger pipelines but I can afford one small lunch in San Francisco for that price." I don't know what the right price point is for hobbyists though because Dagger should get revenue.

#

I am literally "engineer who uses product and then brings that to work" tho. I dogfood most technology choices at home before I bring them up professionally.

still jacinth
#

One option in the future might be to price telemetry and distribute caching separately. How would you feel about that option?

inner rain
#

telemetry == pretty logs + insights?

mortal ether
#

"bring your own telemetry" when lol

#

And, yeah, that would be cool. Distributed caching is really cool for companies but there are engineers who will use this for small projects and they are a pipeline for corporate adoption if they like the sauce.

idle grotto
#

My hope is that dagger becomes successful enough to have a free tier with just pretty logs 😛

#

That's my selfish "I want everything free" side coming out, though

still jacinth
inner rain
#

by insights, we're looking at DORA metric like things too, so related things Dagger might show us from the telemetry (logs + metadata)

idle grotto
#

Dagger Cloud is very much giving off Pulumi SaaS vibes though in being a luxury for home users.

still jacinth
#

For reference Datadog CI (a CI pipeline insights product) is $12/committer/month

inner rain
#

omg, their pricing is complex

idle grotto
#

Pretty common to have monthly meetings with your datadog rep because of how complex their pricing is 😄

mortal ether
#

It's datadog. You need to deliver your firstborn in order to just sign the contract.

#

Hashicorp is kind of similar.

still jacinth
#

One customer is paying 5 figures for just that feature and planning to cut it and just use insights from Dagger Cloud

idle grotto
#

Ooh, I think I needed to see more of that in the demo

still jacinth
#

on top of spending less on AWS for CI compute

inner rain
idle grotto
simple mica
# still jacinth For reference Datadog CI (a CI pipeline insights product) is $12/committer/month

This didn't fly with my company when we looked at this. We noped out real fast and strung together some DD dashboards that kind of gave some insights into our kube Jenkins runners and some pipeline stage level metrics from Jenkins. Not nearly as nice as directly integrating with DD CI insights.

To this day we don't have good visualization of our CI. We are building a Jenkins plugin to scrape data from Jenkins into a datalake but that product probably won't be as nice as what Dagger could provide. It's not there yet though.

#

DD is super expensive. Not just for CI but in general

inner rain
still jacinth
simple mica
fallow niche
#

What would be a good guidance for how to estimate "dagger engine minutes" ? Just the sum of current CI compute minutes? Does it matter how many dagger engines you run? E.g. if you run the engine just-in-time with each CI run vs hosting a couple of engines continuously with N>1 clients connecting to them?

still jacinth
#

@fallow niche it is the amount of time each engine has spent actively computing while under the management of Dagger Cloud (for telemetry and/or distributed caching), rounded up to the nearest minute.

#

So in a CI context, it will be the slice of your CI compute minutes spent running the dagger engine.

fallow niche
#

So theoretically, let's say I have CI (assume all Dagger) just basically running nonstop 24/7. Would it matter in terms of Dagger Cloud costs if I have just one beefy engine vs 3 smaller ones ?

still jacinth
#

In practice, what happens is that you have a fixed workload to run, and that's roughly the same number of minutes total. Running it on a beefier machine will make it less minutes. Spreading it across multiple machines will divide the minutes. So in practice, you don't have to worry too much about the impact of your CI architecture on Dagger Cloud cost. You simply choose the architecture that makes things faster, and that will also make Dagger Cloud cost cheaper

#

The optimal architecture (small vs large machines, long running vs ephemeral, many machines vs. few) really depends on your workload

#

How CPU-intensive it is, how cacheable, etc

fallow niche
#

Yeah, will have to test but I see how it might not be a factor (hopefully). Would be great if I can just go ephemeral and forget about running engines as services (assuming this setup performs well), without having to worry about the cost implications.

still jacinth
#

This is why the free trial is so important, it's a chance to estimate your cost in a real world setup

idle grotto
still jacinth
hidden steeple