High level summary​

Most of this cycle went into Leios and the block header abstraction that supports it. The EraBlockHeader interface was reshaped into a separate type class per protocol that can be specified on a per-era basis, which required a dedicated BbodySignal type and a redefined BBODY rule for Babbage, where the TPraos to Praos switch happened. Related to this, the block header now uses a more lax VersionInfo in place of ProtVer for Dijkstra and no longer relies on a self-reported version when constructing headers. On Leios itself we now select the Leios committee at the epoch boundary, expire BLS keys after KES expiry, check the BLS proof of possession in Dijkstra's POOL rule, added a lens for the announcement block header reference on LeiosEraBlockHeader, and added Dijkstra forecasting for Leios on the prototype branch. We also began laying groundwork for Peras by adding its protocol parameters, while keeping Peras certificates out of the Dijkstra block body for now.

Work on the Plutus V4 context and sub-transactions continued, with the PlutusV4 context implemented for the GuardingPurpose and the sub-transaction index now available in the Plutus context. On the Dijkstra era itself we added a DijkstraEraTx type class with a hashTx function, fixed a superclass constraint in DijkstraEraPParams, added reference-input cost protocol parameters (ppRefInputsCostPerMultiAssetPolicyL and ppRefInputsCostPerDatumByteL) starting at Dijkstra, and fixed conservation of value for phase-2-invalid transactions. Elsewhere we introduced a distinct type for the pool distribution used for governance, fixed VRF reuse when re-registering a pool, and removed some unnecessary memory overhead in native scripts.

On the testing side we continued filling in predicate-failure coverage for the sub-rules and related rules (MEMPOOL, SUBGOV, SUBLEDGER), enabled tests in CertSpec, added a BBODY ImpSpec, added regression tests for conformance, LeiosCert serialization tests, and more withdrawal tests. For infrastructure and releasing we updated the Allegra changelog, bumped the CHaP index-state, and dropped cardano-crypto-leios as a source-repository-package from the cabal project.

Low level summary​

Features​

• PR-6054 - Use a distinct type for the pool distribution used for governance
• PR-6062 - Fix VRF reuse when re-registering pool
• PR-6052 - Select the Leios committee at the epoch boundary
• PR-6047 - Expire BLS keys after KES expiry
• PR-6089 - Fix superclass constraint in DijkstraEraPParams
• PR-6077 - Fix conservation of value for phase2- invalid transactions
• PR-6099 - Add DijkstraEraTx typeclass with hashTx function
• PR-6088 - BlockHeader cross era interface
• PR-6084 - Use more lax VersionInfo in place of ProtVer in BlockHeader for Dijkstra
• PR-6101 - Add ebReferencesAnnouncementBlockHeaderL to LeiosEraBlockHeader
• PR-6067 - Add Peras protocol parameters
• PR-6068 - PlutusV4 context for GuardingPurpose
• PR-6117 - Do not use self-reported version when constructing block headers
• PR-6064 - [Leios Prototype] Add Dijkstra forecasting for Leios
• PR-6110 - Add ppRefInputsCostPerMultiAssetPolicyL and ppRefInputsCostPerDatumByteL protocol parameters starting at DijkstraEra
• PR-6085 - Check BLS proof of possession in Dijkstra's POOL rule
• PR-6121 - Sub Tx index in plutus context
• PR-6122 - Remove unnecessary memory overhead in NativeScripts

Testing​

• PR-6082 - Add regression tests for conformance
• PR-6091 - Add LeiosCert serialization tests
• PR-6095 - MEMPOOL predicate failures
• PR-6083 - SUBGOV predicate failures
• PR-6115 - Enable tests in CertSpec
• PR-6116 - BBODY Impspec
• PR-6111 - Additional withdrawal tests
• PR-6112 - SUBLEDGER predicate-failure tests

Infrastructure and releasing​

• PR-6087 - Update Allegra changelog
• PR-6107 - Update CHaP index-state and remove cardano-crypto-leios as SRP from cabal project

High level overview​

This week, the Mithril team completed the support for an IVC follower aggregator joining the network at any time, the circuit verification key whitelist/revoke mechanism, the scaling tests of the SNARK infrastructure to full security parameters, the finalization of the Mithril SNARK Book, and the activation of the dual genesis keys. They also completed renaming the circuit test assets and production keys, and identifying the circuit verification keys by their transcript representation. They worked on adding a Proof of Bound Possession for the SNARK verification keys, renaming the future_snark feature to snark, hardening the SNARK primitives and adding the missing conversion and error-path tests in ivc_halo2_snark.

The team completed leveraging the certificate chain cache with the EarlyStopVerification mode, adding a file-system backend for the cache and enabling it in the client CLI. They also started adding an IndexedDB backend for the cache in the WASM client and documenting the certificate chain cache.

Finally, they completed the upgrade to Cardano node 11.1 and kept investigating the blocks and transactions verification failure on the testing-preview network.

Low level overview​

Features​

• Completed the issue Support an IVC follower aggregator joining the network at any time #3446
• Completed the issue Implement circuit verification key whitelist/revoke mechanism #3148
• Completed the issue Test scaling SNARK infrastructure to full security parameters #3153
• Completed the issue Finalize the Mithril SNARK Book #3383
• Completed the issue Activate dual genesis keys #3388
• Completed the issue Rename the .bin circuit test assets and production keys #3552
• Completed the issue Identify circuit verification keys by transcript representation instead of bytes #3556
• Completed the issue Leverage the certificate chain cache with the EarlyStopVerification mode #3521
• Completed the issue Add a file-system backend for the certificate chain cache #3522
• Completed the issue Enable the certificate chain cache in the client CLI #3523
• Worked on the issue Add a SNARK verification key Proof of Bound Possession #3537
• Worked on the issue Rename future_snark feature to snark #2814
• Worked on the issue SNARK primitives hardening #3557
• Worked on the issue Add the missing conversion and error-path tests in ivc_halo2_snark #3470
• Worked on the issue Add an IndexedDB backend for the certificate chain cache in the WASM client #3524
• Worked on the issue Document the certificate chain cache #3525

Protocol maintenance​

• Completed the issue Upgrade to Cardano 11.1 #3346
• Worked on the issue Blocks and transaction verification fails on testing-preview #3526

High level summary​

• Benchmarking: Node 11.1.1 release benchmarks; LSM-trees (on-disk LedgerDB) benchmarks.
• Development: Benchmarking workload submission full Dijkstra support - as well as on-disk serialization for db-synthesizer.
• Tracing: Full switch to Hermod Tracing in planning; Alert manager for cardano-tracer is ongoing work; Native span tracing for Hermod.
• Infrastructure: cardano-smartbench: A detached performance workbench - feasibility.
• Leios: On-disk LedgerDB tx validation benchmark shipped to SPOs. First Leios tests on the performance cluster, including tooling updates.
• Organizational: Performance & Tracing Team Meetup in Helsinki.

Low level overview​

Benchmarking​

The P&T team has performed, analyzed and published benchmarks for the Node 11.1.1 release. While the CPU usage improvement persists, and the memory increase is smaller than with 11.1.0, it's still measurable. It was mentioned as a known issue for this release. It seems to correlate with cyclic events such as ledger snapshotting, and while it has been observed on Mainnet as well, it did not represent an operational risk to the network.

Furthermore, we re-ran the 11.0 baseline, as network latencies between cluster machines had changed since May 2026 and were confounding network-related performance metrics. We additionally ran benchmarks with a new Ledger feature that loads huge datasets injected into ledger state in a streaming fashion, and we could validate it works as intended. This, however, is only relevant for testnets; the feature is blocked for Mainnet, as no data may be injected at all.

Last but not least, for Node 11.1 we repeated benchmarks of on-disk LedgerDB stressing the UTxO set - both with and without artificial memory constraints on the process. Those benchmarks are currently under analysis.

Development​

tx-generator, the workload-submission tool behind our benchmarking profiles, gained two new pieces of functionality (cardano-node PR#6695). It can now dump its generated transaction stream to disk - as plain text or as raw CBOR - instead of always submitting it live against a running cluster, so other tools can inspect or validate a workload without needing a full cluster to generate against. At the same time, its era-specific code paths were extended to cover the Dijkstra era, so we can generate and submit Dijkstra-era transactions for benchmarking purposes.

The longer-term goal of writing an entire transaction workload to disk is having db-synthesizer as a consumer. It enables creation of fully valid Praos or Leios chain fragments in a short time - and those chain fragments will exhibit all the properties defined in our benchmarking profiles.

Tracing​

In one of the upcoming node versions after the Dijkstra hard fork, we plan on replacing the new tracing system (trace-dispatcher) with Hermod Tracing. While Hermod currently is at feature and API parity with trace-dispatcher, it makes a sensible split between API and core packages, allowing for improved development in the future. But more importantly, it will enable some new tracing system features which have been deferred up to now, such as live reconfiguration of trace filter rules, and single-source definition of tracers for all purposes - consistency checks, auto-documentation and of course trace emission. We're currently integrating Hermod into the Haskell node component stack and preparing for a smooth transition when the time comes.

The implementation of the monitoring/alarm system is progressing (cardano-node PR#6664). Currently we're working on a robust test harness; a feature like this requires extremely rigorous coverage. Additionally, we're working on interfacing the system with escalation channels, such as pagers, pager app APIs, email and so forth.

Native support for nested spans in Hermod, which we've mentioned as exploratory work in earlier posts, has taken its first concrete shape: a new module implements Loki-style begin/end span tracing, pairing an opening and a closing trace message under a shared span identifier (hermod-tracing PR#19). The PR is still fresh and under review.

Infrastructure​

We've begun a feasibility study for detaching our performance workbench (the benchmarking automation framework) from the main Haskell node project - working title cardano-smartbench. It would let the workbench pick custom inputs, tooling and analysis targets rather than being tied to the Haskell node project's own layout. This addresses issues for benchmarking branches that have diverged a long distance from the project master (such as Leios), and would also generalize the workbench to work with clients other than the Haskell node in the future. The current challenge is keeping the existing CI guarantees alive on the Haskell node's master branch, without risking all downstream tooling silently going out of sync. As it's a feasibility study, there's no PR for it (yet).

Leios​

The transaction validation time benchmark has been turned into a self-contained, distributable executable and shipped to SPOs. It can run on their own hardware without any nix or Haskell/cabal stack. We're specifically interested in the on-disk measurements for LedgerDB on a wide variety of configurations - as being I/O bound (vs. CPU and RAM only) implies being dependent on a much larger array of parameters, such as kernel version, SSD hardware and connection, file system choice and even mount options.

We're currently testing our benchmarking automation on the performance cluster with the Leios prototype. While the numbers obtained from that won't be representative, we're preparing for the Leios Release Candidate: our infrastructure needs to support everything Leios adds - from config values and protocol parameters to understanding Leios-specific timestamped events for performance analysis.

Organizational​

This month, the Performance & Tracing team had a three-day in-person meetup in Helsinki. We had excellent discussions on focus shifts that come with our transition to ICAN Group. Additionally, we ideated means of making the team's delivery more accessible to the wider Cardano community, especially non-technical people - as our direct audience is no longer only IOG engineers. Thank you to everyone who contributed their views and ideas!

High level overview​

This week, the Mithril team completed the implementation of the Relation trait for the IVC circuit and the removal of the HTTP client and TLS features from STM. They continued implementing the circuit verification key whitelist/revoke mechanism, supporting an IVC follower aggregator joining the network at any time, and adding a Proof of Bound Possession for the SNARK verification keys. They also started activating the dual genesis keys and the future_snark feature by default, testing the scaling of the SNARK infrastructure to full security parameters, and finalizing the Mithril SNARK Book.

The team completed the certificate chain cache committed after chain verification in the client library and started leveraging it with the EarlyStopVerification mode.

Finally, they released the new 2630.1-hotfix distribution which fixes a stack overflow in the aggregator and signer on the release-preprod network when decoding a block, kept upgrading to Cardano node 11.1, and kept investigating the blocks and transactions verification failure on the testing-preview network.

Low level overview​

Features​

• Completed the issue Implement Relation trait for the IVC circuit #3428
• Completed the issue Remove the HTTP client and TLS features from STM #3558
• Completed the issue Add a certificate chain cache committed after chain verification in the client library #3520
• Worked on the issue Implement circuit verification key whitelist/revoke mechanism #3148
• Worked on the issue Support an IVC follower aggregator joining the network at any time #3446
• Worked on the issue Add a SNARK verification key Proof of Bound Possession #3537
• Worked on the issue Activate dual genesis keys #3388
• Worked on the issue Activate future_snark feature by default #2814
• Worked on the issue Test scaling SNARK infrastructure to full security parameters #3153
• Worked on the issue Finalize the Mithril SNARK Book #3383
• Worked on the issue Leverage the certificate chain cache with the EarlyStopVerification mode #3521

Protocol maintenance​

• Released the new distribution 2630.1-hotfix
• Published a dev blog post Distribution 2630.1 hotfix is now available
• Completed the issue Stack overflow in aggregator and signer on release-preprod when decoding a block #3549
• Worked on the issue Upgrade to Cardano 11.1 #3346
• Worked on the issue Blocks and transaction verification fails on testing-preview #3526

High level summary​

The Plutus team has delivered a number of September milestones, and is putting the finishing touches on the remaining ones.

A new with-crypto cabal flag makes it possible to build Plinth and Plutus without the system crypto C libraries. We've also further improved the usability of the certifier in the uplc cli tool.

Work in progress includes the cost models for keepPolicies and dropPolicies (CIP-0168), and a CIP for removing the scope check during script deserialization.

Key Pull Requests Merged​

• Encode single constructor ledger types using Data.List
• Add some V4 helper functions
• Improve efficiency and consistency of Data.AssocMap
• Deserialiser optimizations
• Add a with-crypto cabal flag to build Plinth/Plutus scripts without system crypto C libraries
• Improve uplc/plc certifier usability
• Emit checkpoints in UPLC inliner
• Add more test cases for Plutus Ledger Api V4 types

Pull Requests In Progress​

• Cost models for keepPolicies and dropPolicies (CIP-0168)
• Make Data AssocMap FromData shape-safe
• Require exact product arity in generated FromData
• Evaluate Scott caseInteger scrutinee once
• Specify casing on built-in types
• Builtin Pattern Matching (CIP-0194)
• Compile BuiltinData transparently in the plugin