The Challenge
What DATwise Was Facing
DATwise sells analytics tooling to enterprise customers, each of whom brings their own data volumes, query complexity, and SLA expectations. The original single-tenant architecture could not be economically multi-tenanted — shared query infrastructure caused noisy-neighbour problems where one heavy analytical query degraded performance for all other tenants on the same cluster.
The Solution
What We Built
We redesigned the platform around a resource-isolated multi-tenant architecture. Each tenant was assigned a dedicated query worker pool with configurable CPU and memory limits, provisioned via Kubernetes namespaces and enforced through resource quotas. The data layer used ClickHouse for columnar analytical storage, partitioned per tenant. A query routing service distributed workloads and enforced per-tenant query concurrency caps. Infrastructure-as-code in Pulumi made new tenant provisioning a single pipeline run taking under 3 minutes.
Results