September 2026 Summaries

Cockroach Continuum is presented as an Agentic Database Cloud designed to manage rapidly growing database estates created by AI-driven applications, replacing cluster-by-cluster administration with centralized, policy-based estate operations. The platform provisions and scales databases according to workload requirements, pools resources through virtual clusters to reduce idle capacity and noisy-neighbor issues, and uses Cockroach Aegis agents to monitor, diagnose, recommend, and help execute operational actions within human-defined guardrails. Its Cockroach Plenum infrastructure separates compute from storage for elastic scaling and usage-based or provisioned storage options, while serializable isolation across regions aims to preserve transactional correctness under high concurrency. Cockroach Continuum also emphasizes PostgreSQL compatibility, AI-assisted migration, and CockroachDB’s production maturity and resilience, citing deployments at substantial scale. The company argues that consolidated capacity management and minute-based resource billing can lower total database costs as estates expand, with Standard and Mission Critical editions planned for launch on Continuum Cloud on September 15, 2026, alongside preview availability of Aegis and usage-based Plenum storage.

CockroachDB BYOC is now generally available on AWS, Azure, and Google Cloud, offering a managed database service that runs within customers’ own cloud accounts and VPCs to address data residency, procurement, and cloud-commitment requirements. Cockroach Labs manages provisioning, patching, upgrades, backups, monitoring, and incident response under the same uptime SLAs as CockroachDB Advanced, while customers retain control of compute, networking, storage, backup buckets, row data, query results, and unredacted logs. The service uses customer-controlled cross-account identities rather than VPNs or inbound VPC access, with applications connecting through endpoints in their own accounts and only telemetry and redacted logs leaving the environment. Engineers receive no standing account access, and temporary access is auditable and revocable, although revoking it ends Cockroach Labs’ operational responsibilities without disrupting the database’s existing traffic handling. The offering cites ISO 27001-related certifications, SOC 2 reporting efforts, PCI DSS Level 1 certification, and annual third-party penetration testing, and customers can begin setup through their Cockroach Labs account team.

Cockroach Labs presents Virtual Clusters, a public-preview feature of Cockroach Continuum, as a way for platform teams to provide application-specific, isolated database environments while operating a single Dedicated Host Cluster rather than a separate physical cluster for each workload. Each virtual cluster has distinct connections, identities, schemas, data boundaries, and resource ceilings, while sharing underlying compute within one organization, allowing platform teams to centralize upgrades, monitoring, incident response, capacity planning, and governance while enabling self-service provisioning for application teams. The company argues that this approach can reduce the operational burden of cluster fleets, particularly as AI-driven applications create more rapidly changing demands for isolated environments. Cockroach Labs says the model is already used in its serverless infrastructure, supporting more than 85,000 virtual clusters across 15 host clusters with median provisioning times of two seconds, and reports that its internal AI assistant Mica has provisioned hundreds of isolated workloads on one Dedicated Host Cluster.

Cockroach Labs will stop publishing source code for new CockroachDB and Pebble releases in its public repository, while retaining the CockroachDB Software License and leaving the existing repository available but unmaintained. The company says AI-assisted code analysis has increased security risks by making it easier to identify vulnerabilities and reproduce protected software, while legal protections such as copyright have become less certain in an AI-driven environment. Cockroach Labs argues that privatizing development will better protect its intellectual property, customers, and ability to continue product innovation, while offering FAQ, support, and account-team resources for users who relied on source access. Despite this shift, the company says it remains committed to open source through its maintenance of several Go libraries, upstream contributions to projects such as Go, gRPC, and Bazel, and public sharing of technical ideas through blogs, papers, and conferences.

Cockroach Labs has introduced Cockroach Continuum, a database platform designed to replace traditional, ticket-based provisioning and fixed-capacity planning with elastic private virtual clusters that can scale from fractions of a CPU to large distributed deployments. The company says its internal agentic application platform illustrated the need for this approach, as more than 500 users launched over 1,000 applications in two months and one group’s active databases grew from 13 to 392 while costing about $250. Continuum combines cluster virtualization, which enables many isolated databases to share physical infrastructure and aggregate fluctuating demand, with Cockroach Plenum for independently scalable storage and compute, and Aegis, an advisory AI operations layer that analyzes clusters and recommends actions. Built on CockroachDB’s distributed architecture, the offering targets enterprises managing rapidly expanding fleets of applications and workloads across public, private, BYOC, and self-hosted environments, emphasizing operational efficiency alongside resilience, SQL compatibility, and multi-region support.

Cockroach Labs introduces Cockroach Continuum, an elastic database platform designed to support large, bursty agentic and enterprise workloads by allowing thousands of isolated virtual CockroachDB clusters to share private physical infrastructure. Building on CockroachDB’s distributed SQL, KV, and storage architecture, Continuum uses Plenum, a disaggregated multi-tenant storage layer, to separate durable storage from compute, enabling rapid node creation, independent scaling, and reduced storage costs through shared immutable data. Its KV and SQL layers support multi-tenant virtual clusters and SQL Pods that can scale from zero to meet demand, while admission control and resource management aim to prevent noisy-neighbor effects and enforce fair resource allocation. The platform also includes RoachMgr, a specialized control plane for managing deployments and upgrades across database estates, Aegis, a read-only AI DBA and SRE service that analyzes cluster conditions and recommends improvements, and an AI-powered Migration Assistant based on MOLT tooling for moving existing PostgreSQL workloads, with further database support planned. Cockroach Labs positions Continuum as a way to lower database estate costs, simplify operations, enable database-per-tenant and ephemeral testing models, and retain CockroachDB’s consistency, resilience, scale, and multi-region capabilities.