| Vitalik actually has an excellent blog post on rollups. [1]
As well as the Rollup Centric Roadmap posted last year which is a little more technically specific[2]. All the evidence/math is there to read, or in the developer forums. Currently Ethereum has 15 transactions per second. Optimistic rollups can rollup 200 or so transactions per second into each ethereum transaction. Optimistic rollups are launching onto mainnet this month after successful testing for the past few months[3]. So straight away with ORs we can hit 3000tps. Then on top of that general purpose zk-rollups aren't far behind [4][5] and they can get into the thousands of tsp themselves. Once we get data sharding on layer 1 this amplifies the number of rollup transactions we can handle pushing us up to ~100000tps (a combination of increasing how much the main layer 1 chain can handle, and this multiplying with the rollups). And then there is computation sharding further down the line which will push it up again. This has taken so long and hasn't been seen before simply because nobody knew how to do it! The researchers have spent the past 5 years intensely working on it. But its actually starting to happen, and this month too! Which is pretty wild for those of use who've been working on it along the way, lots of previously unfeasible and expensive applications are suddenly going to be possible and cheap. We scale up with rollups this year, then move onto a full Proof of Stake network (not DelegatedPos)early next year, finally ditching energy inefficient PoW to everyones relief, and we have a much greener, better distributed, easy to participate in, scaleable for mainstream use, beautiful ledger. [1] https://vitalik.ca/general/2021/01/05/rollup.html
[2] https://ethereum-magicians.org/t/a-rollup-centric-ethereum-r...
[3] https://optimismpbc.medium.com/
[4] https://aztec.network/
[5] https://ethereum.org/en/developers/docs/layer-2-scaling/ |
"A simple Ethereum transaction (to send ETH) takes ~110 bytes. An ETH transfer on a rollup, however, takes only ~12 bytes"
But the 12 byte number makes some assumptions that will probably not be true or only be true if many transactions are made by the same addresses.
Computationally there didn't appear to be any scaling since checks still need to be made by someone and verified by other nodes that fraudulent transactions aren't in a roll up.