Skip to main content

Hydra Team Update

· 4 min read
Noon van der Silk
Noon van der Silk
Software Engineering Lead

High-level Summary​

We've published the 2.2.0 release! It bundles up the partial fanout work and the hydra-tui revamp from the last couple of updates, together with a much richer benchmark suite (real-world TPS metrics, a Mixed UTxO generator, and a new matrix sub-command that sweeps cluster sizes and UTxO shapes, published to a scenario benchmarks page). On top of that we unified the on-chain membership checks around the BLS accumulator, which shrinks the snapshot-signing tuple and removes a pile of now redundant hashing, and started optimising snapshot processing so script-heavy heads don't pay to re-evaluate Plutus they've already validated.

What did the team achieve?​

  • Released 2.2.0, packaging partial fanout, the hydra-tui revamp, and the new benchmarking work, with the changelog and network config updated accordingly
  • Unified the membership checks: the three separate UTxO hash fields (utxoHash, alphaUTxOHash, omegaUTxOHash) collapse into a single accumulatorHash derived from the BLS accumulator, shrinking the snapshot signing tuple from 7 fields to 4, and full fanout now verifies outputs via a KZG membership proof (the same path as partial fanout) rather than a hash comparison #2619
  • Started a snapshot-side throughput optimisation: transactions already validated on receipt are now re-applied via reapplyTx, skipping redundant Plutus script evaluation on the hot path, with a fallback to full application when a commit or decommit changes the active UTxO #2717
  • Extended PR benchmarking: a "TPS" count (both waiting for TxValid and firing as fast as possible then counting confirmed-snapshot txns) #2635, and a benchmark comparison comment showing the PR-vs-master difference #2721
  • Added a head-state visualisation tool for inspecting single states and diagnosing multi-party scenarios #2719
  • Added a mustNotMintOrBurn guard to the Increment and Decrement validator transitions, consistent with all the other head transitions #2718, #2697
  • Further validator fixes: CloseInitial now constrains the accumulator commitment, and the HeadStatus schema and Haskell types agree on Final #2712
  • Removed leftover CollectCom/Abort handling missed in the original "directly open head" work #2709
  • Switched to makeIsDataIndexed instead of unstableMakeIsData for stable on-chain constructor indices #2724, #2713
  • TUI fixes: the increment dialog no longer re-queries L1 UTxO state every time (it reuses the cached value and lets you refresh if you want) #2726, plus more small fixes #2710 and a fix for flakey TUI tests #2708
  • Merged a handful of simple fixes across smaller issues #2715
  • Added deposit and recover transition tests and tidied up some others #2711
  • CI and tooling: more nix caching by building derivations with -Werror to match nix flake check #2720, added etcd to the exes shell so smoke tests can run a system etcd #2722, and skipped the preprod-with-mithril path while it's offline #2723

What's next?​

  • Add a test that probes how large a UTxO set we can actually (partially) fan out #2616
  • Continue the still-open node-observing-unrelated-Heads deposit issue #2606
  • Finish the user-friendly hydra-node configuration file #2581
  • Accept PaymentExtendedKey (HD wallet keys) #2583