,
6 tweets,
3 min read
Read on Twitter
Erlay, by @pwuille, Gmax, Gleb, et al is a big deal. TL;DR it DRASTICALLY reduces bandwidth requirements of running a full-node compared to the current flooding approach. Bandwidth no longer scales linearly with peer count, decreasing the cost of running well connected nodes. 1/5
Take a look at this figure from their paper (arxiv.org/pdf/1905.10518…). With Erlay, a node with 32 peers uses only marginally more bandwidth than a node with 8 peers. With BTCFlood (the current approach) the node uses 4x the bandwidth! 2/5 
Currently, every time a node receives a new tx, it announces it to all of its peers. This is highly inefficient because it creates a lot of redundant announcements. This chart shows the amount of data currently used for tx announcements by connectivity. 3/5
The insight behind Erlay is that nodes need only announce txs to a small subset of peers and then engage in an efficient set-reconciliation protocol periodically to detect any missed transactions. This set recon. protocol uses the Minisketch library previously announced. 4/5
With Erlay it becomes way less expensive to run very well connected nodes, increasing the connectivity and security of the Bitcoin P2P network. Great work to everyone involved! 5/5
Oh and who knows. Maybe this will be used for the Lightning Network graph p2p protocol at some point? @roasbeef @rusty_twit @bitconner @niftynei 6/5
