What are the recommended specs for a Kubernetes/Helm on-prem production deployment?

Refer to the Infrastructure-Spec-For-Tazama.md, in the tazama-lf/docs repo (dev branch) - with Kubernetes sizing tables at different FTPS (financial transactions per second) scales, based on Azure examples. It excludes the BIAR stack, which should be sized separately on comparable hardware alongside the specified nodes. As of July 20, 2026, some details were still being clarified in #get-help: the document’s examples reference Redis and ArangoDB, while the current On-Prem-Helm charts use Valkey and PostgreSQL, so it is not yet confirmed whether the sizing figures still apply as-is; the meaning of “CRSP” in the scaling table and whether the per-service counts (TMS, TP, TADP, Redis, NATS) represent replica counts were also unconfirmed; and whether BIAR should run in the same cluster, on dedicated nodes, or in a separate cluster was still open. A Tazama Technical Steering Committee Helm Working Group is separately active on Helm deployment practices - see https://github.com/tazama-lf/technical-steering-committee/discussions/21 - and is a good place to raise these questions directly. Source: #get-help, July 2026.