ICYMI: in about 3 weeks, Ethereum will enable PeerDAS - a novel data storage method increasing max network throughput by a whopping 8x times! it is the biggest scalability improvement since the introduction of Layer-2 rollups - and a major acceleration towards Ethereum’s singularity let’s break down the tech behind PeerDAS in the simplest way possible. 1. as more consumer apps and institutions choose Ethereum because of credible neutrality and 100% uptime, the volume of transactions grows dramatically 2. keeping up with this demand requires validators to install ever larger (and faster) storage and bandwidth - still manageable today but progressively out of reach for independent validators like home stakers 3. two solutions: abandon home stakers who are the source of Ethereum’s decentralization (bad), or find ways to make transaction data storage & validation way more efficient so that home stakers could keep up (good) 4. enter PeerDAS - a method of splitting & distributing transaction data in pieces, where individual validators only need to verify and store 1/16th of the original data for it to be preserved in full - a massive efficiency improvement 5. how it works: instead of sending the whole transaction data to every single node in the network, PeerDAS partitions it into 128 pieces and distributes a random sequence of at least 8 pieces to every node for verification and storage 6. the network is then split into 128 groups of nodes - “gossip networks” - each responsible for storing & verifying their own data piece. every node belongs to at least 8 of these groups. collectively, they store all the data, but individually, they store & process significantly less, eliminating resource constraints as a bottleneck to scalability improvements 7. it's like breaking down the verification & storage of the whole data into smaller chunks that can be done independently and in parallel, with results then pooled together, reducing the effort required from each participant 8. the trick is that as long as >=50% of the pieces have been verified and stored by different nodes, the original data can be restored in full from these fragments, like a scratched DVD that would still play the movie if not too badly damaged. 9. the name of this trick is “erasure coding”. kinda like a multisig approval requires k/N threshold to push a transaction, erasure coding allows to split data into N pieces and choose the threshold k that’s required to restore the original data. in PeerDAS, N pieces is 128 and k is 64 - a 50% threshold 10. so while every node gets to participate in verification & storage, the network actually needs only 50% of the data to recover the original, with some caveats. this is a high enough hurdle to withstand coordinated attacks on the chain, and low enough to enable efficiency improvements. 11. mechanically, every block, the minimum number of data pieces a single validator node must download, store and serve to peers is 8 out of 128 ...