The Blockchain Trilemma is Real in the Real World

Numerous projects claim to have overcome the blockchain trilemma. I find that claim dubious at best. I’m also deeply suspicious that they’ve solved it “academically.” I’m not going to explain all of the ways that they can mislead you – but I am going to apply some cold hard reality of the actual world on various project’s transactions per second (TPS) claims.

The Trilemma

The Trilema states that a blockchain can be two of secure, decentralized, or high TPS. Chia chose the first two because we believe that losing either of the first two means you have a really crappy database and not a blockchain. We also believe that the trilemma is really a dilemma as a centralized “secure” blockchain is an oxymoron – but that is for another day.

Here is one of the reasons why. In the US the average upload speed for a fixed broadband internet connection was 35 Mb/s as of April 2023. Wireless is less than half that.

I found three different numbers when attempting to estimate the average size of a Solana transaction – a blockchain known for high TPS. The highest was from a TPS study that assumed “no more than 176 bytes,” and the lowest I computed was 99 from a post on how to increase the number of accounts that can be sent to in one transaction. A random developer implied the average was 137. Since the average of all three estimates rounds to 137, we will use that.

Mathing It Out

Now, let’s do some math. The speed in bytes of the average internet link is 35 × 1000 × 1000/8 or 4,375,000 Bytes/s. That means that 4,375,000/137 leads to a maximum one peer upload TPS of 31,934. When you divide this by the number of peers in a Bitcoin node that’s 266 TPS – and if you use the current Chia default of 40 it’s only 798 TPS. These are also wildly unrealistic numbers because they ignore protocol overhead, TCP overhead, and any of your kids trying to watch Netflix or you being on a Zoom call.

But that’s not all!

The expensive computer you need to keep up with all that TPS also has to store the historical state of the blockchain. At 798 TPS, that’s adding about 9.5GB per day to that database for a yearly addition of 3.5TB1. And that drive has to be an SSD to keep up with the fast pace of transaction updates. As of this writing, a Samsung 8.0TB SSD will set you back $530 but won’t even get you three years of operation. Don’t even think about what the EBS bill would be if you chose to host in AWS.

High TPS “blockchains” will be Proof of Existing Data Center for the foreseeable future. These nodes can not be run anywhere but on a fiber backbone hosted in the data centers that already control our lives. If your definition of decentralized is “fine with the cost of running a node being prohibitively expensive and only located in data centers,” you’re not going to make it.

If any blockchain claims it has a high TPS and is decentralized and secure – at least for now, we can assume it’s snake oil.

1.(796 TPS) × (137 Bytes) × (24 hr) × (60 min) × (60 sec) ≈ 9.5 GB/day ↩︎

original source – https://www.chia.net/2024/03/01/the-blockchain-trilemma-is-real-in-the-real-world/