Hyperconverged Infrastructure (HCI) vendors often talk about their software; in fact, an increasing number of them are claiming to be software vendors more so than hardware vendors. The HCI software stack is primarily made up of two components; a hypervisor like Microsoft’s Hyper-V or VMware’s ESX and storage software like Microsoft’s Windows Server Software-Defined which makes improvements to Storage Space Direct and Windows Server 2016. The quality of the HCI software is continually improving, but each of these stacks relies on commodity hardware which limits the overall effectiveness of the solution.
The appeal of commodity hardware is its low cost, but the shortcomings of commodity hardware quickly erode that cost advantage and lead to a much more expensive total cost of ownership.
The Problem with Commodity Hardware
Commodity server hardware typically uses a storage controller to interface between the PCIe bus and the physical storage devices installed in the unit. Most HCI nodes on the market today support a maximum of 24 drives per node. The PCIe architecture of four PCIe slots with 16 data lanes per slot of the typical HCI node causes the 24 drive limit. If the server uses flash-based drives, the server doesn’t have enough bandwidth to derive optimal performance from the drives. The PCIe bus and the single storage controller create a bottleneck. Some vendors try to leverage a few native NVMe drives as a cache to the SAS connected SSD or hard drives, but this technique creates performance inconsistencies.
The hardware configuration of the typical HCI node is problematic for several reasons. First, since the systems only support 24 drives, the customer is often forced to add additional nodes to meet capacity demands. HCI is known for its scalability, adding nodes to meet capacity and performance needs but any clustered system becomes more complex as node count grows. HCI should indeed allow for scale-out growth by adding nodes, but it should do so sparingly.
The other challenge is performance. The typical HCI node bottlenecks at two points, the number of available PCIe lanes (typically 4 slots with 16 lanes each) and the number of storage controllers (typically one), computing power, however, is not typically a problem. If multiple storage-intensive workloads on the same node start to demand rapid IO response, then the node may not have enough IO lanes to support the demand.
If a storage performance situation arises, adding more nodes will not typically improve performance. In HCI, a workload is tied to a specific node while it is on that node. The customer’s only recourse to improve performance is to move other workloads off of that node so that the storage-intensive workload has access to more of the resources, which of course reduces the efficiency of that node and further encourages node sprawl.
Axellio Solving the Node Sprawl Problem
Instead of creating HCI software, Axellio designs hardware ideal for HCI environments. If HCI software is the fuel, then Axellio is the engine. Axellio’s hardware is certified on Windows Hyper-V with Windows Software-Defined. It expects to achieve certification with VMware ESX in the very near future and is even pursuing certification with Nutanix Acropolis Hypervisor as Nutanix evolves into more of a software company.
Axellio nodes support up to 72 drives per node instead of the more standard 24. The increase in drives per node solves the number one cause of node sprawl, a need for additional capacity. All those drives however also improve performance because the internal architecture of each Axellio creates a non-blocking, high bandwidth architecture that allows full utilization of its All-NVMe flash architecture. Each node has 6 PCIe slots with 32 data lanes each.
The combination of an all-NVMe Flash architecture plus the internal bandwidth to take advantage of the drives’ performance means that each node can potentially support dozens if not hundreds of VMs, which also means that a 4-node Axellio cluster can support more workloads and more storage-intensive workloads than a 16-node cluster from a competitor.
Our next blog examines the Return on Investment (ROI) of an HCI solution built on premium node hardware versus the typical commodity approach.
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.
Key Takeaways:
- 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?”.
George Crump is the Chief Marketing Officer at VergeIO, the leader in Ultraconverged Infrastructure. Prior to VergeIO he was Chief Product Strategist at StorONE. Before assuming roles with innovative technology vendors, George spent almost 14 years as the founder and lead analyst at Storage Switzerland. In his spare time, he continues to write blogs on Storage Switzerland to educate IT professionals on all aspects of data center storage. He is the primary contributor to Storage Switzerland and is a heavily sought-after public speaker. With over 30 years of experience designing storage solutions for data centers across the US, he has seen the birth of such technologies as RAID, NAS, SAN, Virtualization, Cloud, and Enterprise Flash. Before founding Storage Switzerland, he was CTO at one of the nation's largest storage integrators, where he was in charge of technology testing, integration, and product selection.
StorageSwiss.com - The Home of Storage Switzerland 