What's New in Paradigm's Reth 2.0

The Ethereum L1 has two kinds of clients, consensus and execution.

Consensus clients handle the proof-of-stake (PoS) layer, while execution clients manage Ethereum Virtual Machine (EVM) state, i.e. facilitating transactions and executing smart contracts.

These infra layers aren't always all that flashy, but Reth's reveal of their 2.0 execution client is interesting for a few reasons.

Reth is notably a Paradigm Research project, built and led by many of the same people behind Tempo, the permissioned, payments-optimized EVM L1 that has stirred plenty of tension in the Ethereum community in recent months. As an EVM solution, Reth supports node operators on both Ethereum and Tempo.

Reth, which debuted two years ago, is also currently the fourth-largest Ethereum execution client by adoption, lagging behind Geth, Nethermind, and Go-ethereum. Are the technical upgrades in 2.0 poised to help the client rise further in popularity?

Let's take a deeper look.

Gigagas Gains

Gas is the unit Ethereum uses to measure computational work. For context, the Ethereum L1 currently has a gas limit around 60 million gas per block, and the average Ethereum block takes ~12 seconds.

By extension, gigagas is the technical measurement for 1 billion gas units. This is why one of the main headliner advances in Reth 2.0 is that the client can now process 1.7 gigagas/s on the right hardware. 1.7 billion gas units of work every second!

While Ethereum researcher Justin Drake has discussed (on Bankless!) the path to 1 gigagas per second on the Ethereum L1, the reality is that in today's numbers, Reth 2.0 can already process the equivalent of roughly 28 Ethereum mainnet blocks per second. So, the performance gains here likely have more to do with Paradigm's interest in building for future scaling needs of Tempo's permissioned validator set.

Smarter State & Storage

Every time a block is processed, Ethereum needs to compute something called the state root, which is basically a fingerprint of the entire state of the chain at that time.

Clients have needed to wait for all transactions in a block to finish completing before computations could begin, which created chokepoints.

Reth 2.0 solves this with two improvements working together: a Sparse Trie Cache and Partial Proofs. The former takes snapshots of relevant state between blocks, while the latter essentially halves the amount of data a node has to pull from disk to verify changes. Together these resources bring state root calculation at the end of blocks down to 2 milliseconds or less.

Moreover, Reth 2.0 also overhauls how data is stored. A new "tiered storage" system separates commonly used data from historical records, so that a full Ethereum mainnet node can now fit in less than 300GB, with snapshot downloads as low as 170GB supported.

Instead of taking days, spinning up a Reth node from scratch is streamlined to a couple of terminal commands plus waiting ~10 minutes.

A Tale of Two Nodes

Paradigm was long seen as an indomitable resource for the Ethereum community, and seeing their leadership and research shift much of their time and energy (while also hiring away key open-source talent) to pushing out Stripe's vision of a more corporate-aligned ecosystem has been a bummer.

For their part, the Tempo team has repeatedly maintained that their work will continue to benefit the broader crypto ecosystem, including Ethereum. Reth 2.0 is at least one concrete data point in favor of Paradigm's position. Tempo node operators can run it just as Ethereum node operators can run it, and this is undoubtedly a significant net benefit to the Ethereum ecosystem in its own right.