Unclear onboarding, trial and value proposition

[khromov]

I wanted to evalute Ymir for a potential use case. I signed up through the website, installed the CLI, connected AWS, went through the init setup and at the end was greeted with "The "Personal" team doesn't have an active Ymir subscription". Nowhere in the signup or in the dashboard does it say that there is no trial period.

Since the main goal of my signup was to evaluate what kind of performance and ongoing costs Ymir would have (Aurora, ALB, ELB, Secrets manager, etc) I could not accomplish that. Some things can be guessed based on documentation but not all. I would really love to see a trial so that the service can be tested. Many companies require budget approval for expenses and having developers jump through this extra hoop adds unnecessary friction.

In terms of value proposition, it would be really great to see some cost breakdowns of typical low, medium and high traffic sites running on Ymir in the documentation.

My back-of-the-napkin math: The cheapest database costs $15 a month (bump to $45 if you want multi-az redundancy), I assume an API gateway is needed which adds to the cost. This is before any Lambda, bandwidth or other fees. On top of that you have the cost of Ymir, $39.

This will definitely make sense for some clients, but touting serverless I was hoping that the ongoing costs for a small installation would be minimal, which does not seem to be the case.

Having personally experimented with Aurora Serverless I understand why that might not be an option due to its long cold start time.

PS.
I'm a long time newsletter subcriber and have followed this project with great interest, I am very impressed and inspired over what you've built and intend for the above purely as constructive criticism! 🙇🏻

[carlalexander]

Heya Stanislav!

Thanks again for the feedback! I still hope we get to meet in person in Sweden one day. 😄

I wanted to evalute Ymir for a potential use case. I signed up through the website, installed the CLI, connected AWS, went through the init setup and at the end was greeted with "The "Personal" team doesn't have an active Ymir subscription". Nowhere in the signup or in the dashboard does it say that there is no trial period.

The web backend is still in pretty basic shape. I need to do another pass at it now that I put in the billing. There's just still a lot to do everywhere lol.

Since the main goal of my signup was to evaluate what kind of performance and ongoing costs Ymir would have (Aurora, ALB, ELB, Secrets manager, etc) I could not accomplish that. Some things can be guessed based on documentation but not all. I would really love to see a trial so that the service can be tested. Many companies require budget approval for expenses and having developers jump through this extra hoop adds unnecessary friction.

This is actually pretty high on my marketing priorities. I now have about 3 months worth of cost data for my blog so I plan on using that and writing a case study about it. If I'm able to, I'd like to have this out next week or two weeks later.

My back-of-the-napkin math: The cheapest database costs $15 a month (bump to $45 if you want multi-az redundancy), I assume an API gateway is needed which adds to the cost. This is before any Lambda, bandwidth or other fees. On top of that you have the cost of Ymir, $39.

So napkin math on my end too, I'll have something better written up. But like you said, RDS is just expensive sadly, but you can spread it out over multiple sites/projects/environments. For the rest, that's API Gateway, Cloudwatch, S3, CloudFront, Lambda, Secrets, etc., it comes down to roughly $1.20 per 1000 visits.

This is with API Gateway v2 which is more expensive, but it's the only way to get multisite to work. By default, I use API Gateway V2 which is quite a bit cheaper. You might be looking at closer to $1 per 1000 visits with that. I don't have data on that setup unfortunately.

You might also be able to reduce cost if you don't use CloudFront and switch to CloudFlare. I've made sure that's possible, but I haven't documented it yet sorry.

This will definitely make sense for some clients, but touting serverless I was hoping that the ongoing costs for a small installation would be minimal, which does not seem to be the case.

Well, as you can see, it really depends on the total traffic. If you get very little, your cost will be RDS divided by the number of sites that use it. That's it.

I'm a long time newsletter subcriber and have followed this project with great interest, I am very impressed and inspired over what you've built and intend for the above purely as constructive criticism!

Thanks! ❤️ I appreciate it. I know it's still rough even with the year and a half that I've spent on it already. But it gets a bit better every day thanks to feedback like this 😃

[carlalexander]

Published the analysis today :)