Yral: Cleaning up canisters left over from previous per-user multi-subnet multi-canister architecture

[saikatdas0790]

At Yral, over the last couple of months, we have progressively moved over most of our app metadata over from our previous architecture of a per-user-canister model to a microservices architecture where a single canister would specialize in a particular aspect of metadata storage, for example, user profile metadata, video post metadata, rate limits and so on.

As part of that, we are going to be cleaning up infrastructure and removing canisters that are no longer actively used. We had primarily held off on doing this and focused on building user facing features to focus on product market fit.

However, this is technical debt that needs to be repaid and we’re choosing to allocate bandwidth to this effort this week onwards.

It makes sense to do this now owing to gradually rising cycle costs and concerns posted on the NNS and the Dfinity forums

Proposal: 141656

https://forum.dfinity.org/t/hey-this-is-just-like-a-feeler-ok/68218

To summarize, our current canister infrastructure that will remain:

• user_info_service
• user_post_service
• rate_limits

What we’re removing is:

• platform_orchestrator - top level canister that controls orchestrators on individual subnets
• subnet_orchestrator - children of the platform orchestrator that help it operate individual canisters on each subnet
• individual_user_canisters - children of the subnet orchestrators where each canister was intended to hold data for a single user on the app

[Hephalump]

@rishichadha30

As you might have noticed, there’s been a forum post circulating about YRAL’s investments into FuelEV and Estate. We’d like to take this opportunity to clarify a few things for the community and share our side of the story—why we invested in FuelEV and Estate

https://forum.dfinity.org/t/how-yral-ai-exploited-the-sns-fuelev-sale/42600?u=hephalump

https://forum.dfinity.org/t/yral-response-to-fuel-ev-sns-investment/42668?u=hephalump

https://forum.dfinity.org/t/remove-4-node-providers-from-the-nns/62191/27?u=hephalump

• Delete All their SNS
• It’s a Bad Idea

I wonder what @Bitwalker would say about this?.

[Bitwalker]

What I didn’t say about them yet ? There’s no reason to repeat myself here, so my posts gets flagged as spam.

[Nate]

Batman thinks the joker and his crew are extremely dumb

[Nate]

I will take a deeper look. So far all I know is Adam got his money taken and is blaming everyone else for his bad investments.

Dirty Harry Do You Feel Lucky Punk

[DirtyHarry]

[saikatdas0790]

If that’s what the NNS decides, we will happily comply. I doubt the NNS will choose to do this however

[saikatdas0790]

They will ideally go back to canisters belonging to the DOLR ai DAO

[Hephalump]

@saikatdas0790

I doubt the NNS will choose to do this

[saikatdas0790]

We never “took”.

We have always chosen to double down on the IC

We focused on wherever we could help the IC the most.

We did what we did because we felt it would be in the best long term interest of the entire ecosystem and DAOs that we steward

Back then, the SNS was a new approach to funding software that could eventually lead to software becoming sustainable via self-functioning communities and ecosystems.

We chose to share our knowledge and wisdom of running and operating canisters at scale with other well meaning long term partners in the ICP ecosystem

When we were given harsh feedback on our approach, we took corrective action and reversed and course corrected to the best of our ability.

Same for the canisters we spawned. Since the very beginning, we have focused on acquiring users instead of sweating the minor details.
We have been building on this network since its infancy. We’ve tried every model and every mechanism of scaling on the IC. We know its ins and outs.

We didn’t magically waltz into the ecosystem and weren’t showered with gifts as narratives spun by motivated individuals would have you believe. We did the best we could with the resources we were provided.

We spawned hundreds of thousands of canisters because that was the only way the IC could scale back then to serve quickly signing up users. “without real users” → We see very differently on what a “real user” is. If an individual is trying to access our app, we did whatever it took to provide them the best experience, even if it meant eating the infrastructure cost at that point. That’s how every software at scale operates. You think the GPTs and the Claudes you are using are not doing the same?

“saturate subnets, bottlenecking performance, and creating artificial demand” → If a paying ecosystem project that pays for its cycles is spinning up canisters, isn’t the network fulfilling its intended purpose?

Also, this is an engineering problem if the impact is as large as you say. If subnets are so easily overwhelmed, they need to be better designed. IF the design is working as intended, and if you need to hinder this behaviour, then cycle costs should be hiked.

Large ecosystem whales shouldn’t need to randomly start witch hunts to create false narratives if the canister economic model is flawed.

“Catastrophic damage” is such a silly take. If we have “catastrophic damage” from an ecosystem project spending raised funding on infrastructure that runs the software that their DAO governs, then you have a lot of introspection to do.

[saikatdas0790]

I don’t consider myself “lucky”.

I try to be hardworking though.

[saikatdas0790]

Do you ever provide sources for any of the things you post?

We did have an employee who helped build our frontend from back then named Harsh. It is possible that you had your GPT look at parts of this code that calls into the canister code. And that is what you copy pasted here as an image

[Hephalump]

@saikatdas0790

I try to be hardworking though.

[saikatdas0790]

We’ll limit this thread to Yral and related conversation instead of starting a completely unrelated topic here. Thanks

[Hephalump]