One of the promised benefits of hyperconverged infrastructure (HCI) is simplicity coming from the convergence of computing, networking, and storage into a single tier and from the ease of scale by “just adding a node.” Half of the promise is true, having a single interface and hardware tier is easier to manage but “just adding a node” creates complexity over time. The primary problem is HCI adds nodes too quickly because each node is often underpowered and can’t scale beyond its initial configuration. The result is node sprawl which decreases per node performance and increases complexity.
Node Sprawl Equals Complexity
“Just add a node” is almost a cruel joke played on unsuspecting IT professionals. Each time a node is added to the HCI cluster IT needs to complete many tasks. First, there is the physical racking of the node. In some data centers just finding space for the next node is a significant task. For practical purposes, it makes sense to place the new node near the current members of the HCI cluster, which may require IT to shuffle other hardware to make space. Second, the node has to be physically racked, powered on and physically connected to the network, which means the data center needs to have the necessary power and available network ports. In most data centers physically connecting a node can take days if not weeks.
Once IT has the node racked, the networking team needs to get involved to make the logical connection to the HCI cluster. In many data centers, the networking team is the busiest of any group and often has the longest queue for task completion. Setting up the logical connections for the nodes can take an additional few days or weeks depending on the networking team’s backlog. Combined with the physical setup, it is not uncommon for HCI administrators to wait more than a month before they can prepare the node for use in the cluster.
Preparation of the node for use in the cluster requires adding the node’s IP address and installing the HCI’s storage software on the nodes. If the node’s storage is of different types, then the administrator also needs to allocate the storage into various pools. Once the essential preparation is complete, the administrator needs to move virtual machines to the node. If the HCI’s storage software stripes data across the nodes it needs to take time to aggregate the new node’s storage into that stripe, which can negatively impact performance.
The Cost of Complexity
The above steps all have a cost associated with them, but IT planners may, incorrectly, assume these are all soft costs, but there is also a physical cost. Each node costs money. HCI nodes are typically “maxed out,” regarding storage and computing power. Initially, much of these resources may go unused. Each node also requires power and cooling. Each network cable consumes a network port potentially forcing the purchase of new or expanding current switches, which also require more network administration time to configure. In addition to the above costs, there is also the cost of something going wrong, requiring IT time to troubleshoot, diagnose and fix.
Stopping Node Sprawl is Job 1
The primary goal of HCI 2.0 is to stop node sprawl. It does this by making sure each node has the maximum performance possible. Instead of buying dozens of “cheap” nodes, HCI 2.0 intelligently utilizes a handful of quality nodes capable of internal expansion before another node is required. HCI 2.0 can still scale-out like HCI 1.0, but the goal is to minimize the number of times that has to occur.
Where HCI 1.0 was all about scaling-out, HCI 2.0 is about scaling-in before scaling-out. The scale-in technique enables organizations to optimize the power of Intel CPUs, and NVMe Flash drives to support dozens of virtual machines per node. Reducing the number of nodes from dozens to a half dozen significantly lowers networking complexities and costs.
Our next blog details the HCI 2.0 solution from Axellio. The company’s high-performance nodes are ideal for Tier 1, storage-intensive workloads. In the meantime, click here to watch Storage Switzerland and Axellio Inc., in our on demand webinar “How to Put an End to Hyperconverged Silos.” In this on demand webinar, you’ll learn why current generation HCI solutions fall short and the essential requirements for HCI’s next generation.
- Learn Why HCI 1.0 Shortcomings Are Costing You Money and Adding Complexity
- Learn Why Hardware Matters in HCI Solutions
- Learn How HCI 2.0 is ideal for Tier 1 Workloads
Register and watch now to receive a free copy of Storage Switzerland’s latest eBook, “What is HCI 2.0?”.