In response to “Is Decentralized Storage Sustainable?”

At Blue Link Labs, we’re creating a peer-to-peer browser named Beaker, and a super-peer service named Hashbase. In this post, we respond to a blogpost that questions the feasibility of decentralized storage due to some economic predictions.

Is decentralized storage sustainable?

TL;DR: Yes — because the goal is to decentralize data control, not disk resources.

Q: Will a decentralized market recentralize?

First, let’s understand the argument being made. From David Rosenthal’s post, “Is Decentralized Storage Sustainable?”:

My main problem is an economic one … I argue that [a decentralized] network is economically unstable and will, over time, become centralized.

David is predicting how the market will act when hosts become interchangeable.

Competition among the peers and decreasing hardware costs will drive down the $/TB/month rent to levels that are uneconomic for Silicon Valley peers, concentrating the storage resource in China (as we see with Bitcoin miners).

He’s arguing a decentralized hosting market will cause the price floor to drop. I think that’s a good prediction if you make two assumptions:

  1. Buying and selling occurs in an automated auction, and
  2. Hosting providers have no way to differentiate.
Lets assume that all the peers in China share the same low cost base. But some will have responded to the increase in demand before others. They will have better economies of scale than the laggards, so they will in turn grow at the laggards’ expense.

This is the Walmart effect: low prices will make hosting unprofitable for anybody that doesn’t operate in an economy of scale.

The result of this process is a network in which the aggregate storage resource is overwhelmingly controlled by a small number of entities, controlling large numbers of large peers in China.

So, do we think David’s prediction is believable?

A: Yes, if you accept the assumptions

I think David’s prediction could be very accurate if you accept the two assumptions I mentioned above.

The price floor will drop if you have a well-functioning market. Networks like Filecoin are designed to match consumers to providers based on automated asks and bids. With no way to differentiate, providers will be competing on price, and the profit margin will get eaten.

However, there are some non-software differentiators. Here are a few:

  • Location. You may want your files to be available at multiple places across the world, to improve performance.
  • Redundancy. You may have strict requirements for keeping backups in multiple locations, to prevent loss.
  • Performance. All disks are not made equal, and you might want a provider that uses the latest whiz-bang tech.
  • Reliability. You may want a good relationship with your provider, so that you have somebody to hold accountable for issues.

It’s possible that these differentiators won’t be important to the average consumer, and so I think David’s prediction could accurate.

This is one of the reasons we launched Hashbase instead of a cryptocurrency network like Filecoin. We agree that Filecoin’s price floor could drop, and that Filecoin services will recentralize, and therefore the added effort and overhead of Filecoin would be fruitless.

Hashbase is a simpler solution. It can be self-deployed and easily replaced, so no autonomy is lost. It also gives users a fixed relationship with their provider, so they know who to hold to account.

Does any of this matter?

We don’t think so. The goal of Beaker and Hashbase is to decentralize control over the data — not over the disks.

We believe peer-to-peer networks will have a huge impact because they solve the walled garden problem on the Web. Data published on Dat remains under the user’s authority. There’s no server setup required. Applications just publish to the user’s local device.

This is a much bigger deal than who runs the hosting provider. Projects like Mastadon or Diaspora are admirable, but they require herculean effort to get right. If we want open source, open data, and privacy in our applications, then they need to be baked into the core technologies.

— prf