The goal of this milestone is to increase the bandwidth of the DAL network on mainnet. It covers both the modifications required in the DAL node and the amendments to the protocol.

Work breakdown

Parallel Computation

• Use Eio_posix for parallel computation, as Eio_main does not seem to work here. (!19586 (merged) )
• Refactor the amplificator to replace the use of Lwt_io.fork by a domain. (!19625 (merged))
• Replace the iter in the validation of shards by parallel computations. (!19646 (merged))

Experiments

• Run an experiment with mainnet conditions (stake repartition, time betwwen blocks...) but with DAL bandwidth increased, to show that the computations in the DAL nodes are able to sustain claimed bandwidth.
• Run experiments with specific stake fractions to identify the needs for bakers depending of their stake.

Increase of the size of slots and the number of slots

• support constants change in the economic protocol
  • protocol stitching code
  • deal with entrapment accusation at protocol border (If the cryptobox parameters are used in the protocol to check if a shard really belongs to a commitment, then the right cryptobox parameter has to be used by the protocol when checking denunciation)
  • update refutation game logic (at protocol border)

Deliverables

• The DAL node is able to use several threads to validate faster.
• An experiment as been conducted, showcasing the ability to sustain claimed bandwidth (8 MB/s ?).
• Hardware requirements are updated (for producers, rollup operators, and bakers with various stake)
• Protocol is ready for new bandwidth (constants have be updated, stitching has been done, refutation games can be performed, even across the protocol change).

Follow-up questions

• Optimization of the load between threads
• Optimization of batch sizes
• Identify the bottleneck when the increase of the number of threads does not lead to a linear increase of the performance.
• Re-run the experiments for block time reduction.