- SECP benchmarking: we concluded our benchmarking runs and analyses of the new SECP primitives for the Valentine hard-fork.
- Release benchmarking: we performed a round of benchmarks for the 1.35.6 release.
- UTxO-HD benchmarking: we performed first runs for UTxO-HD and are currently refining the benchmarking setup.
- New tracing: for better accessibility, the new tracing system is being outfitted with introspective capabilities.
- Infrastructure: with the Nomad cloud workbench backend we were able to perform our first test cluster runs successfully on SRE infrastructure.
- Infrastructure: the initial NixOps workbench backend has been completed; a PR containing this work, along with many quality-of-life improvements of our tooling, got merged.
- For SECP, we settled on a fixed tx count per block, while simultaneously spending as much as possible of the block budget. Thus we were able to minimize the impact of per-SC-call overhead.
- The final runs were performed with various fractions, e.g. half, of the current block budget to ascertain how these workloads would fare compared to a value-only run.
- The SECP machinery and profiles are currently being generalized into an approach to aim for very specific aspects of a smart contract for benchmarking.
- After analyzing initial UTxO-HD runs, it turned out that mempool snapshotting had to be throttled for benchmarking; it affects a lock that UTxO-HD had to introduce into the forging loop.
- We're currently adapting the benchmark setup to that, and will then perform a new combination of baseline and UTxO-HD runs.
Benchmarking the 1.35.6 release candidate could attest to a perfectly clean bill of health.
Work on the new tracing system's introspective capabilites is ongoing: Immediate use cases of the new API include being able to statically validate generated tracer documentation, as well as providing information of a specific tracing setup in the node via traces themselves. These features will make the new system both more robust, and more accessible.
- Work on the cloud deployment capability of the Nomad workbench backend continued; for testing we can automate multiple Nomad clients.
- Locality assumptions were removed and job monitoring was refactored.
- To facilitate directly-executable derivations, Nomad Job specification files are now self contained with GitHub references and configs needed to run a cluster.
- We're currently evaluating different options for genesis distribution in said cluster.
The NixOps workbench backend has reached an initial functional stage. Consequently, the relevant PR was merged. It also contained many improvements to our analysis tooling, as well as a structural overhaul of workbench itself. We consider this an important step of future-proofing our benchmarking machinery.