← privacyprophet.xyz

Devcon 8

Scaling to 10GB Blocks: FRIVail, a Data Availability Scheme

Rachit Anand Srivastava · Sr. R&D Engineer, Avail · @privacy_prophet

▤ Download the slides PowerPoint · full deck · 3.5 MB ⌥ Read the code GitHub · rac-sri/FRIVail

tl;dr

FRIVail is a data availability scheme built on FRI over binary fields (Binius). A single FRI commitment shows the erasure coding is correct, so there is zero overhead. It is hash-based, so it is post-quantum. Each shard proves its own row, which lets the network scale horizontally toward 10 GB blocks.

how it works

  1. Encode: bytes → multilinear polynomial → Reed–Solomon codeword (2× blow-up)
  2. Commit: Merkle roots of the codeword and every FRI fold layer
  3. Prove: FRI folds the codeword down to a terminal codeword; L random queries check each fold
  4. Verify: the codeword is within the unique-decoding radius of one RS codeword
  5. Sample: each light-client sample is checked as one extra FRI query at index i (ρr+1 = QSelect(i)), the fix from FRIDA
  6. Recover: any 50% of the codeword rebuilds the data

benchmarks

datacommit timeproof timeproof size
4 MiB39.1 ms28.5 ms221.3 KiB
8 MiB74.2 ms50.6 ms244.6 KiB
16 MiB149.1 ms89.8 ms275.4 KiB
32 MiB296.1 ms165.6 ms301.8 KiB
1 GiB11.26 s10.66 s475.2 KiB

Proof size grows about 2.1× for 256× more data (221 KiB at 4 MiB → 475 KiB at 1 GiB).

papers

related