Centralization of Decentralized Networks?
We investigated seven of the more mature and developed blockchain networks in web3 and found that the majority are highly reliant on public clouds.
Problem: Many layer-one blockchain nodes are run on public centralized clouds, some as high as 90%.
STORE Research investigated 7 of the more mature and developed blockchain networks in the web3 space and found the majority are highly reliant on public clouds for node hosting leading to a centralized distribution of their underlying network architecture.
General Information
- Of the 7 networks that STORE researched, 5 had >50% of their nodes hosted on public clouds such as Amazon AWS, Microsoft Azure, etc.
- On a few of the networks, over 50% of their nodes were hosted on just 3 of the main public cloud providers
○ Some of the most common being AWS, Hetzner, DigitalOcean, OVH - Roughly 2/3rd of Ethereum and Solana nodes, the two most popular networks to build dApps, are hosted on the few largest centralized public clouds
Why It's Important
From the beginning, one of the core principles of blockchain networks and cryptocurrency has been decentralization. The idea that the currency, applications, communication, etc. being provided by a network are not at the mercy of a single entity who is trusted to not censor, hijack, or mismanage the proceedings. But are we achieving decentralization in Web3 if the majority of network nodes are hosted on the same public cloud services that most Web2 companies use?
Ethereum Mainnet Node Distribution
Fig 1: Ethereum Nodes by Network Type (Data provided by ethernodes.org. Date: 6/17/2022)
Not only are the majority of these network’s nodes hosted on public clouds but oftentimes the same cloud provider accounts for around 1/3rd of all nodes. While this may provide cheap and reliable hosting for the user setting up the node, what occurs if one of these companies (or even all of them) decide they no longer want to participate in the crypto ecosystem and ban these nodes from their clouds?
The same story was found when looking at both the number of validators and the amount of stake weight those validators represent when looking at Solana, who makes this data available. Even when looking at stake weight, public clouds accounted for roughly half the stake weight (as opposed to 2/3rd by count). Clearly, what is needed is a decentralized alternative that provides reliability, and affordable storage and computation while maintaining a censorship-resistant and permission less architecture.
The same story was found when looking at both the number of validators and the amount of stake weight those validators represent when looking at Solana, who makes this data available. Even when looking at stake weight, public clouds accounted for roughly half the stake weight (as opposed to 2/3rd by count). Clearly, what is needed is a decentralized alternative that provides reliability, and affordable storage and computation while maintaining a censorship-resistant and permission less architecture.
Figure 2: Examples of node distribution with Avalanche and Cardano
Appendix
Figure 3: Compilation of research data
