Why a 40G internal network matters between compute and storage

[rychy]

Прочети на български

When the VPS workload and the storage layer are separated, the internal network stops being a secondary detail. It becomes part of how the storage behaves.

The public VPSStorages architecture has a 40G internal link between the compute and storage servers. The idea here is not just "more speed", but a more predictable environment for storage operations, backup jobs and the VM workload.

Where people often go wrong

• they look only at the internet uplink and ignore the east-west traffic inside the infrastructure
• they confuse a latency problem with a "slow disk" without checking the network path
• they use storage over the network without a clear idea of what happens under congestion

If storage is separate, you have to think about bandwidth, latency, failure domain and operational visibility as one whole. Otherwise it is very easy to start chasing the wrong bottleneck.

This topic is a basic starting point. There will also be a separate post about how we look at the edge/firewall layer before traffic even reaches the servers.

Related topics

• Why a 40G internal network matters
• A VM latency problem
• The troubleshooting process

---

Want to see these settings in action?
All our VPS services are built around the same engineering approach. If you are looking for real infrastructure rather than just a reseller service, see the VPSStorages plans and infrastructure.