FullDAS: towards massive scalability with 32MB blocks and beyond

[cskiraly]

@cskiraly

Duplicate reduction techniques

…

• We might also use techniques based on diffusion state, such as Push-Pull phase transition: in the first steps of the diffusion of a segment, there are almost no duplicates. Most of the duplicate load is happening in the last steps of the diffusion. One can emphasize Push at the beginning (even with an increased degree, sending to more nodes than without this technique), and change to gossip-and-pull towards the end of diffusion if the diffusion state can be estimated.

I’ve published a more detailed writeup with measurements on this technique. See it here:

PPPT: Fighting the GossipSub Overhead with Push-Pull Phase Transition

Is it a typo here? Did you mean ‘sending IHAVE to the rest’? Yes, it is a typo. I check if I can still correct in the writeup. Thanks! Yes, I did. I’m preparing new figures with these.