Scaling and capacity verification¶
CNs can spread sessions and run suitable distributed plans. TN, log quorum, object storage, network and background work can remain bottlenecks; separate deployment does not imply arbitrary linear scaling. Do not use the historical 0.8 roadmap to describe current TN behavior or promise arbitrary horizontal growth for every node.
Deployment and observation¶
Configure local/object storage explicitly. Public/private/cross-cloud deployments require network, residency, storage-interface and log-reliability checks. Proxy labels/CN groups allocate compute; permissions and physical isolation need separate validation.
Hold data, plans and concurrency constant. Record migration, active transactions, remote operators, warming, p50/p95, throughput and errors. Validate correct results and recovery targets with performance.
See scaling, CN groups and architecture.