Thread by @epolynya

Seeing FUD that "rollups only scale compute". Not true, rollups and DAS scale compute, storage IOPS, storage size and bandwidth - the whole deal.

Compute: this one is established and undisputed. How much depends on the rollup's design, but let's say ~100x vs L1s.

(contd.)

The greatest trick rollups pull off is actually data:

- Converting complex state (SSD) to sequential DATA (HDD).

- Compressing this sequential data heavily.

As an example, the baseline transaction on L1s is ~128 bytes, and complex DeFi transactions can be hundreds of bytes.

Meanwhile, dYdX transactions are only 5.35 bytes, and the baseline is 16 bytes. Furthermore, instead of XXX bytes on expensive SSDs, you only need X or XX bytes on very cheap HDDs. Altogether, this is a 100x to 1,000x boost in data efficiency.

But wait! There's data sharding!

With the magic of data availability sampling data, the more nodes you add, the more data capacity you get! This is the very opposite of monolithic blockchains. You also alleviate the bandwidth bottleneck as each node only needs to process a small chunk of the whole.

Summing it up:

Rollups scale compute 100x

Scale data 100x to 1000x

DA layers take this already 1,000x more efficient data and compounds on it by offering massive capacity that scales with decentralization, thus also scaling bandwidth.

Moore's/Nielsen's laws apply over time.

To be clear, the rollup sequencers may need to be expensive, and each rollup works differently, but the important distinction is rollup sequencers are only responsible for ephemeral liveness and CR, while monoliths need to offer far harder permanent liveness, safety and CR.

So, at the execution layer, rollups can foster rapid innovation, with techniques like novel consensus systems, high frequency state expiry, parallelization of compute etc.