Hyperconverged Infrastructure (HCI) is one of those technologies that look great on the whiteboard. The idea of collapsing the compute, storage and networking tiers into one, and creating an infrastructure that automatically scales as each node is added to the cluster sounds perfect. For the most part, HCI does live up to its promise of simplification, at least initially. However, as the environment scales with the addition of more and more nodes, complexity sets in, especially regarding the network.
One big challenge is operational. The scaling by “just adding another node” claim that some HCI vendors make is complicated by the realities of the data center. Adding another node means making sure it is networked into the infrastructure, which means finding a network administrator with the time to ensure the network is provisioned correctly for that new node. While it may take less than an hour to rack the new node, it can take days for the network administrator to configure the node’s networking. The same realities of siloed job responsibilities also apply to configuration changes. Moving or changing a node will likely involve the networking team.
As the HCI environment grows, the additional nodes also consume switch capacity. Each node typically requires four or more network switch ports. Given that most data centers use brand name switches that carry a premium price tag, the network consumption costs of these nodes adds up. As ports are consumed, new switches need to be added, further increasing costs and slowing IT down with their deployments.
A final challenge in HCI is troubleshooting. In a shared everything environment like HCI, it is difficult to get visibility into which VM and which network path is causing or being impacted by a problem.
Big Switch Cloud Fabric for HCI
Big Switch Networks is a leader in Software Defined Networking (SDN). Its Big Cloud Fabric (BCF) leverages commodity third-party switches available from Dell, HPE and other white box network switch manufacturers to create a very cost effective network infrastructure ideal for HCI. The solution enables HCI environments from Nutanix, Dell-EMC, VMware vSAN, Simplivity, and Red Hat Ceph to reduce the scaling friction caused by the network infrastructure.
Armed with BCF, organizations can have the network automatically configure itself as new nodes are added, or configurations change. They can also reduce costs while keeping up with the port consumption that HCI scale causes by using third party off-the-shelf switches from a variety of switch manufacturers. These switches are meshed together and appear, to administrators, as one fabric. Lastly, using Big Switch in an HCI environment allows for deeper visibility into the virtual machines’ use of the network infrastructure, enabling IT to identify hot spots and take corrective measures easily.
A concern for an HCI administrator though, is supportability of a SDN solution that runs within the infrastructure. Big Switch, in a recent announcement, took several steps to ease this concern. The Big Switch Cloud Fabric is Nutanix Ready Core validated for networking and is a member of the Nutanix technology alliance partner program, Nutanix Elevate. Also, Big Switch and Dell EMC have built a reference architecture for the VMware SDDC, based off of Big Cloud Fabric, ScaleIO, VMware vSphere and Dell EMC Open Networking hardware.
Big Switch Cloud Fabric also has direct integration with VMware vSAN. It enables rapid vSAN deployments with automatic configuration of cluster networking and one-click multicast deployment. BCF’s vSAN-aware deployment tools simplify rollout and management, while a dedicated vSAN dashboard in fabric analytics provides real-time and historical views of cluster events for rapid troubleshooting.
Finally, BCF also works with Red Hat Ceph. Red Hat Ceph architecture is an open source SDS solution; deployments are made up of storage clusters across a mix of disk types (SSD, SATA, SAS, etc.). Clusters make multiple copies of the objects for redundancy, which increases the east-west traffic on the storage network.
Typically, Ceph has two storage networks – one is the front-end application-facing public network with a separate back-end cluster network. The back-end cluster network is how the Ceph-based object management instances communicate with each other to write, access, duplicate and manage the data stored in the objects. BCF supports both Ceph networks within a single SDN fabric. Leveraging its SDN capability, BCF can automate the proper network configuration operations of Ceph software instances for scale-out storage capacity.
As HCI scales, the network becomes more critical having to deal with adding additional nodes and increased east-west traffic to ensure data is shared and protected. The network challenges threaten to slow down an organization’s adoption of HCI. SDN solutions like Big Switch BCF allow HCI environments to scale automatically, on less expensive network equipment while providing deep visibility into each VM’s utilization of the network. For environments looking to scale their HCI environment, a SDN solution like Big Switch BCF should be a must-have add-on.