Abstract

This is a follow up of %(OKR 2022Q3 - 1.1) Pipelining where main goal was to ready the Tezos protocol with the capabilities required to implement pipelining in the Octez node and baker. Hence, the goal of this milestone is to implement said optimizations in the node and baker. Doing so will allow us to reach sufficient level of confidence that is required in order to reduce time between blocks. Moreover, an effort will be made on providing benchmarks and convincing evidences that global TPS are increased.

Backlog

General:

• Address the %Pipelining - Backlog and backport issues in this milestone

Mempool:

• (critical) use new data structure from protocol

Baker and node:

• (critical) new RPCs for the two new notifications !7513 (merged)
• (critical) decide how the baker should retrieve the set of valid operations in complicated cases
• (critical) modify baker automata to start endorsing as soon as possible (before apply) !7516 (merged)
• (critical) do not apply blocks in the baker anymore (only at round 0) !7490 (merged)
• design how to handle locally-prioritized operations (external mempool)
• Modify the protocol's validation mempool mode to handle future (pre)endorsements !7815 (merged)

Mempool cleanup (optional but would lead to technical debt):

• move part that bounds number of operations from plugin into shell !6787 (merged)
• review error types in mempool
• remove parts that are no longer needed
  • remove legacy mempool for Kathmandu !7379 (merged)

Node / P2P:

• only download operations that are not already in the mempool