Breaking the single-host bottleneck

A PropTech business working with Cloud Combinator on AWS. The client is anonymised at their request.

The challenge had four focus areas, each stemming from the concentration of the whole platform on a single instance.

Rather than move the whole platform at once, Cloud Combinator built a repeatable AWS staging foundation as infrastructure as code and introduced services one at a time, validating each before the next, so migration risk stayed low and rollback paths remained open throughout.

The engagement is measured against clear success criteria: the target architecture removes the single-instance bottleneck, matches AWS services to workload type rather than selecting them generically, supports phased validation and rollback instead of a high-risk cutover, and delivers clearer operations, scaling boundaries and security control points than the current state. The figures below capture the shift the modernisation delivers.

• The bottleneck removed — Web, data and processing tiers are separated across a multi-AZ VPC, so the platform no longer depends on a single host and each workload scales on its own terms.
• S3 as the source of truth — Scan data moves off the portal's EBS volume to S3, unlocking stateless workers, horizontal scaling and stronger backup and recovery, with the daily sync retired only once S3 is authoritative.
• Efficient, queue-driven processing — CPU, GPU and ROS2 workloads run as decoupled workers fed by SQS and AWS Batch, replacing the polling GPU instance and letting GPU capacity scale to zero when idle.
• Resilience and operational clarity — Multi-AZ RDS, autoscaling, centralised CloudWatch logging and alarms, and AWS-native secrets and certificate management replace host-level tooling and single points of failure.
• Handover — A target-state architecture narrative, a service-by-service AWS mapping and a phased migration plan are delivered, with the legacy device path preserved until replacement flows are validated.

With a validated, infrastructure-as-code foundation and a proven service-by-service migration path, the client can move from a fragile single-host platform to a scalable, resilient AWS architecture, and expand its device fleet without the constraints of the original design.