A private cloud's success is only as strong as its network performance. Before walking through how to maintain a private cloud network, you need to know all of the network's ins and outs.
Private clouds have gained ground with data centers looking to scale certain applications while still maintaining control over others. But relying too heavily on provisioning self-service and reduced oversight can cause workloads to spiral out of control. Private cloud networks with traditional architectures and equipment aren't optimized for cloud operations, which causes critical bandwidth shortages and performance problems. Admins can take a few steps to preserve private cloud network performance.
Review the current network infrastructure
If you don't document your current network architecture, you run the risk of slow application performance down the road. The enterprise may require physical equipment upgrades, so document each network interface card (NIC) or adapter, Ethernet cable, switch port, and router. For example, a multi-port NIC adapter with TCP chimney offload capabilities can replace an ordinary single-port NIC; each NIC port can connect through Category 6a cables or ports on a switch with enhanced backplane performance.
Other data centers might adopt 10 GigE backbones for extremely high-bandwidth areas. The setup or configuration of each device can also have a measurable effect on network traffic and congestion.
Brace for disaster with failover technologies
Resilience is critical to any private cloud, but traditional single-port connections among servers, switches and other network devices can lead to network failure. This includes NIC and switch-port failures, congested network segments, and disrupted cabling.
Once you've documented and reviewed the current network infrastructure, evaluate how multi-port NIC adapters improve network reliability. The simplest approach is to replace or augment existing single-port NIC adapters on each private cloud server with multi-port adapters and connect each NIC port to a different physical switch.
When properly installed and configured, a failed NIC port, switch port or loose cable won't cut off the server's connectivity. Instead, network traffic fails over to the remaining connections. Some workloads will experience a slight dip in performance, but admins can shift those workloads to other servers while the failed server or switch is repaired or replaced.
Stephen J. Bigelow is the senior technology editor of the Data Center and Virtualization Media Group. He can be reached at email@example.com.