An often overlooked yet crucial component of an HCI environment is the network, especially as the environment scales. Each additional node consumes network ports and needs IT to logically connect it to the HCI cluster. In many HCI environments, the steps required to connect a new node to the network are entirely manual, which often take days to perform. As IT considers prospective HCI solutions, they should spend time to understand how it manages the network. The more automated the process is, the more responsive IT can be to organizational demands.
The Challenges of HCI Networking
Data centers traditionally create a silo of north-south traffic (server-to-storage) and east-west traffic (server-to-server) when designing their network infrastructure. East-west traffic in a non-virtualized environment is rare. Virtualization introduced the concept of virtual machine migration and live virtual machine migration, making server to server (east-west) networking more critical. Storage however, typically remained north-south on a dedicated network. The independence of server and storage networking made it more complicated to implement, operate and manage but it performed well between configuration changes.
HCI attempted to simplify data center operations by collapsing computing, storage, and networking into a single tier. It also, as we’ve explained in previous blogs, put more pressure on the computing and storage components. Networking also carries a more significant burden in the HCI construct. In HCI, both server-to-server and storage I/O traffic is east-west because it collapses north-south storage IO into the nodes. In HCI storage, IO traffic is typically the replication of data to other nodes in the cluster to support VM migration and protect against media or node failure. The advantage of a replication model is reads are local to the VM accessing the data, which means that if the node contains high-performance NVMe-flash the VM receives maximum benefit from the investment.
Software-defined networking (SDN) is a perfect match for HCI environments, but surprisingly most hypervisors have relatively weak built-in SDN capabilities. Most either sell an expensive add-on or partner with third parties to deliver the capabilities.
HCI and SDN – What to Look For
IT needs an HCI solution that provides full SDN functionality. They should look for the ability to dynamically create and secure the network to meet the evolving needs of the organization and its applications. The SDN should enable IT to speed deployment of new HCI nodes and workloads within the HCI environment automatically and non-disruptively. Common tasks, like adding a node, for example, should be automated so that when IT adds a new node, one command can correctly configure the network. The HCI solution should also enable the easy creation of virtual networks to maintain performance levels and contain security vulnerabilities from spreading across the organization’s networks. Finally, the network should be policy driven which governs both physical and virtual networks even as they scale.
SDN is not only an essential element of scaling the HCI environment; it is critical for the entire data center and makes it easier for the organization to extend into the cloud. Most organizations and HCI vendors position HCI as ideal for a small set of use cases like virtual desktop infrastructure (VDI). The point solution approach leads to multiple HCI solutions or HCI coexisting with bare metal and traditional virtual infrastructures. If the organization takes a more strategic approach to HCI with the eventual goal of running most data center applications within the environment, they will see a far greater ROI and see HCI deliver on its promise of a simple cloud-like architecture. Strategic HCI means reconsidering hypervisor, storage software, server (node) hardware and networking. IT needs to look for enterprise-class solutions instead of the more typical point solutions.