Launch HN: Tardigrade–Decentralized Cloud Storage with Client-Side Encryption

10 points by jtolds 6y ago ↗ HN

Hi HN! I'm JT Olio. You might know me from previous posts such as "Go channels are bad and you should feel bad" (bad example), "Whiteboard problems in pure lambda Calculus" (good example), or being one of the first to write about switching my company from Python to Go back in 2014.

I've been working with the team here at Storj Labs to build a decentralized Amazon S3-compatible cloud storage service that is affordable, private, and secure. We just launched our new service called Tardigrade (https://tardigrade.io/) and I'd love to get input from more developers. It's open source (check out the GitHub repo at https://github.com/storj/storj) and we're primarily using Go (naturally).

If you haven't heard of Storj, the simplified explanation is that we are the Airbnb of hard drives. We take extra HDD space, aggregate it, incorporate Reed Solomon and other techniques to create resiliency, and sell it as a service to people who need object storage. The end result is a service that is encrypted by default, as fast as the big 3 cloud providers, and half the cost.

We've gone through a bit of a rebirth in the last few years. After hiring Ben Golub (Docker, Gluster, Verisign) as our new CEO, in 2018 we launched a 90 page, very detailed, technical whitepaper (https://storj.io/storjv3.pdf) for our new storage system, in 2019 we launched our alphas and betas, and now we're launching open access to our service! If you'd like to give it a try and share feedback, I can give you some credit to play around with. I'm also happy to answer any questions or share some more details around testing we've completed. Let me know what you think!

4 comments

[ 3.7 ms ] story [ 20.9 ms ] thread

sharemywin 6y ago ↗

can I use this to host images, js files, a database file?

jtolds 6y ago ↗

Yep, though it's worth pointing out that Tardigrade has a per-object fee and has per-account resource limits as circuit breakers. We are currently targeting archival use cases more than content delivery, and while we expect to start edging into content delivery, until we do we recommend that you add a caching layer on top to avoid running into account resource limits.

Any thoughts on some kind of host for usage type system for community type resources. Things like common crawl, creative commons hosting, large data-sets for deep learning.

This is a great idea we are asked about frequently. We've been heads down, focused on getting the core of the product ready for a while now, but expect us to start looking much more earnestly at these types of use cases this year!

Currently, organizations can sign up and serve data out of Tardigrade with their account, but they would ultimately pay for egress. The reason is that storage node operators also like to be paid for egress. We are considering an option where storage node operators can choose to donate their resources to community data sets, which would then allow those community data sets to be hosted with much less cost.