Q: How does OpenMetal’s dedicated networking compare to shared public cloud bandwidth?
The fundamental difference is physical dedication.
On AWS, Azure, or similar platforms, the NIC and network fabric behind your instances are shared across many tenants. Oversubscription is standard practice: raw physical capacity is divided among more workloads than can simultaneously use it at full speed. This introduces variability: throughput is often acceptable under light load but degrades under contention, particularly for latency-sensitive workloads like distributed databases, financial processing, and real-time data pipelines.
On OpenMetal, each customer’s infrastructure runs on hardware allocated exclusively to them. The network ports, switching fabric, and VLAN allocation serve a single tenant. Within-VLAN traffic runs at up to 10 Gbps and is not metered or rate-limited based on neighbor activity, because there are no neighbors sharing the physical path. The throughput a deployment is configured for is the throughput it receives, under load and at rest.
The cost model also differs. OpenMetal includes all private VLAN traffic in the base cloud cost. Major public cloud providers bill intra-region and cross-AZ traffic separately, which adds cost that scales with internal workload communication, often invisibly until the invoice arrives. For architectures with high internal data movement, the difference in networking cost alone can be significant over time.