Hacker News new | past | comments | ask | show | jobs | submit login

> Hosts on Sia do not have to worry about building and maintaining enormous datacenters, employing thousands of employees, and marketing their services. They only need to worry about providing reliable storage capacity to renters on Sia.

I would expect that building large datacenters and employing people to manage them is exactly how you go about providing reliable storage capacity.

And while it's true that providers don't have to worry about marketing or branding, well... the flip side is that they can't do any marketing or branding and need to compete in a market where the product is completely commoditized.




Sia is built on a completely different assumption than most large datacenters - uptime requirements for hosts are around 95%, which equates to about 36 hours of downtime per month.

This means that you can run a much leaner database and require substantially less expertise to keep things going. If a rack goes down, you don't need someone on-site to bring it up ASAP, you can fix it the next workday without upsetting your customers.

95% uptime for hosts translates to 99.99+% uptime for Sia users, because users store data across hosts in a 10-of-30 scheme. As long as 10 out of 30 of the hosts are online, you will be able to access your data. The probability of losing 20 hosts when each of your 30 hosts have 95% uptime is exceedingly small; your practical uptime depends more on the reliability of the software than it does the reliability of the hosts you use.

As for commoditization: that's the goal! If we can completely commoditize data storage, prices should come down dramatically, and the market should be a lot more efficient. You won't need to be an Amazon or Google to have a stable and competitive data offering.


> This means that you can run a much leaner database and require substantially less expertise to keep things going. If a rack goes down, you don't need someone on-site to bring it up ASAP, you can fix it the next workday without upsetting your customers.

This is already the case for cluster storage systems like Ceph.


This extends to the whole site. You don't need backup power. At all. Maybe no redundant network equipment, if you get replacements on speed dial. Likely no redundant fiber, just don't take your time to arrange a replacement if someone cuts it (use a different, pre-scheduled conduit you can rent and fill with fiber in a day or two).

If you get unexpected "bus factor", you can handle replacement combing back early from holiday.

Get creative.


> I would expect that building large datacenters and employing people to manage them is exactly how you go about providing reliable storage capacity.

Further, I imagine that if there is money to be made providing reliable storage to Sia users, then those with experience of reliably providing low-cost, high-volume storage will corner the market pretty fast!




Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: