Equipment fails and workloads crash -- it's an unavoidable reality for even the most reputable systems. However,...
long-term network performance can be enhanced by making architectural design choices that identify and eliminate single points of failure in the private cloud network. It's important to note that the entire corporate network does not need to be re-architected for resiliency. In many cases, only the private cloud portion of the enterprise network needs such scrutiny.
There are countless approaches and design philosophies for resilient networks, but most involve establishing redundant network pathways from servers to switches to storage. For example, each server might use a minimum of two network interface card (NIC) ports -- perhaps as many as four or even eight ports per server -- and each NIC port would lead to a corresponding port on a different switch. The idea is to create a network fabric where alternative paths are always available and operated with failover, trunking and load balancing capabilities.
Network resiliency is usually complemented with workload redundancy tactics like server clustering, duplicate workload tools like Stratus Technologies' EverRUN MX software, or even high-availability capabilities of virtualization platforms like VMware HA, which can restart crashed workloads on another suitable server.
Prepare alternative or backup connectivity
Whether your private cloud is supporting external users or allowing workloads to burst into the public cloud to meet added user demand, WAN access is critical for any modern cloud network. However, a single connectivity provider is a potential single point of failure for the enterprise. A business that demands resilient cloud network performance should consider integrating a backup WAN provider at the facility -- preferably through a completely redundant fiber or other physical infrastructure. When trouble strikes one provider or their fiber, a business can enable the backup provider (on a separate fiber) with minimal service disruption.
About the author:
Stephen J. Bigelow is the senior technology editor of the Data Center and Virtualization Media Group. He can be reached at email@example.com
Designing a private cloud superior to AWS
Reconfiguring your private cloud to keep users happy
Application resiliency vital for elastic infrastructure