A supercomputer manufacturer was delivering a Top50 hybrid-cloud system to their customer, and Broadwing was brought in to aid that delivery. At more than 10,000 nodes, with services spanning on-premises hardware and cloud capacity, the system's scale and complexity were enormous: a parallel Lustre filesystem, Kubernetes-hosted services, and job scheduling all had to work together across two environments.
Bringing a machine of this class to acceptance means meeting critical integration requirements and performance benchmarks on a fixed timeline — a delivery challenge few teams ever face at this scale.
Broadwing engineers embedded with the manufacturer's delivery team and worked the system end to end:
The system met all critical integration requirements and benchmarks within six months. We stayed through user acceptance testing, then handed a stabilized machine — with automated health checks and runbooks matched to how the system actually behaves — off to the teams running it long-term.
