Multi-tenant SaaS platform moving off colocation
A multi-tenant application running on owned hardware in a colocation facility, where capacity has to be bought months ahead of demand.
This is an illustrative reference design describing how Protogenies approaches this class of problem. It is not a named client engagement, and it contains no performance claims.
What has to be solved
Onboarding peaks exceed available capacity, environments cannot be reproduced, and hardware refresh is approaching.
Technical constraints
- Tenant data must remain logically separated
- No extended maintenance window acceptable
- Small operations team with no cloud platform experience
Approach
A landing zone with separate accounts per environment, containerised services on managed Kubernetes, managed PostgreSQL with replication, and a wave-based migration ordered by dependency depth.
Implementation
- 01Dependency mapping and workload disposition per service
- 02Terraform modules for network, cluster, database and observability
- 03Replication-based data migration rehearsed in a non-production copy
- 04Traffic cutover behind DNS and load balancer weighting with a documented rollback
- 05Post-cutover validation checklist and decommission plan
Technology stack
Expected qualitative outcomes
- Environments reproducible from source control rather than rebuilt by hand
- Capacity follows demand instead of being purchased in advance
- Cutovers rehearsed and reversible per wave
- Per-environment cost visible to the engineering team
Things to weigh up
- Stateful services usually set the pace of the plan
- Licensing and data egress from the incumbent provider need early attention
Talk to the engineers who would do the work
Bring your current architecture, constraints and the problem you are trying to solve. We will tell you what we would change first, what it depends on, and where we would start.