First, I think what they are doing is conceptually kind of cool. You build a kubernetes cluster and then submit it to the network and people pick your cluster from a marketplace to run their apps on. They pay the network in the native ATK token and then you get paid for the resources you provide to the network. Part of me thinks this is kind of cool, and the purpose is clearly to break the duopoly that is forming in cloud compute between Amazon and Microsoft. But there is a but here. A big one.
Kubernetes is not an easy service to run publicly. Breaking out of containers is considered a sport in the infosec community and that is against very hardened containers on the latest, most modern installations with teams of people handling security. I highly doubt the clusters running in peoples basements connected to this network will count. The documentation provided does not include a detailed hardening guide. It includes some basic security features, like keys for ssh instead of passwords, but nothing too special. I wouldn’t want to run any sensitive workloads there.
That’s where workloads like Chia plotting come in. From the kubernetes owner’s perspective its being deployed like an appliance and probably not going to be a massive security risk. From the plotter’s perspective the only thing at risk is their public keys, which are public. So it seems like a pretty good use case on both fronts. It also defaults to Mad Max but seems to support Bladebit as well if you pick a big enough instance.
The real concern I have is cost. This seems to me like an expensive cloud platform, but its hard to check on pricing without going through the entire process. If this is something that interests the community I can look into it, but from my perspective you are better off plotting with an old machine laying around than trying to download terabytes of plots off a cloud service somewhere.
Article taken from thechiaplot.net