Why Hyperconverged (HCI) is the Future of Enterprise Video?

The Traditional 3-2-1 Architecture is The Inverted Pyramid of Doom

Introduction

The term hyperconverged infrastructure (HCI) has become an industry buzzword that has been applied to a number of different new computing technologies. The misuse of the term has caused confusion for many IT professionals looking at HCI as an infrastructure solution. In this article we will shed some light on what HCI really means and why it might be the right IT solution for you.

The Inverted Pyramid of Doom

When virtualization first arrived in the market, the physical server model was dominating IT infrastructure. As a software solution, virtualization required the existing physical servers as well as shared storage technologies like SAN and NAS to survive and thrive. The 3-2-1 architecture was the result of combining these existing hardware components into clusters. Unfortunately, these hardware components were never designed for virtualization and were typically from different vendors. The 3-2-1 architecture has led to a number of challenges, the most obvious being the complexity. Not only do these various layers each have their own management systems, but they each have their own individual support services. Each vendor component solution seemed to require its own training and certifications, and many IT departments have found themselves needing to hire multiple specialized experts either internally or as external consultants to cover these varying components. Dealing with compatibility issues between different vendor solutions such as ensuring vendor X solution’s update level is compatible with vendor Y solution’s update level can be challenging for even the highest priced experts. Another issue in the 3-2-1 architecture is expandability, both in terms of capacity and performance. Shared storage appliances tend to be monolithic and only be designed to scale up by filling empty drive bays. When the system requires bigger or faster storage in a 3-2-1 architecture, that often means having to swap out with a bigger, faster, and more costly storage appliance. The same idea applies when better or faster hardware is needed in the physical server (RAM, CPU, etc): costly upgrades, expensive downtime, and a hefty administrative workload to complete the project. The final, and the potentially fatal, flaw for the IT department utilizing 3-2-1 architecture is the storage. Being the “1” in the 3-2-1, storage represents a single point of failure for the entire architecture (leading to the “Inverted Pyramid of Doom” moniker). While many storage devices can be implemented redundantly, redundancy usually means more than double the cost of the storage, already an expensive component. Instead, many organizations rely on backups and prayers to protect against the catastrophic failure of the storage layer. Despite all of its flaws, the 3-2-1 architecture did get the job done in terms of delivering features like high availability, VM live migration (aka vMotion), and cluster-wide shared storage. It was also the only viable way to effectively implement virtualization for many years. Luckily, there are now other alternatives to this architecture profile.

Converged Infrastructure

Before HCI, there was converged infrastructure. To tackle the complexity of the 3-2-1 architecture, the idea of converged infrastructure was to combine some of the different component layers into a single “system” and SKU, most often combining servers and storage. Sold as a single system, the hardware and software components were pre-tested together to avoid incompatibility issues and speed up deployment time. However, these “converged” solutions were generally the same separate components, just pre-installed, pre-wired and delivered in a rack. The next stage in converged infrastructure was combining and integrating the different components into a single appliance. It wasn’t difficult to add more compute resources to a storage appliance in order to run virtual machines, and that’s what some vendor solutions offered. Clustering would make the storage appliances highly available. 

Generally, converged infrastructure solutions were meant to be hardware platforms onto which 3rd party hypervisors like VMware or Hyper-V could be installed with relative ease. These converged infrastructure appliance solutions are primarily what exists in the market today; most have adopted the term hyperconverged infrastructure due to the buzz factor. The problem with converged infrastructure solutions is that they generally mimic the same storage architectures as the 3-2-1. These clustered, converged storage and compute solutions relied on virtual storage appliances (VSAs) running as VMs to manage storage in a similar way that shared SAN and NAS controllers functioned. VSAs are the best example of this problem as they effectively virtualize all the inefficiencies of the SAN architecture from the 3-2-1. VSAs consume large amounts of CPU and RAM from the appliance, keeping it from being used by other virtual machines. Because the hypervisor and storage are still from two different vendors in these converged solutions, the VMs must consume the storage through a number of protocols and files system layers (and VSAs) that reduce storage efficiency. Each of these layers, including VSA, adds hops to the data I/O path. Only the emergence of flash storage has enabled these converged infrastructure solutions to provide efficient storage for virtualization. 

The Real Meaning of ‘Hyperconverged’ 

When the term ‘hyperconverged’ was coined, it meant a converged infrastructure solution that natively included the hypervisor for virtualization. The “hyper” wasn’t just hype as it is today. This is an important distinction because it has specific implications for how the architecture can be designed for greater storage efficiency and simplicity. Who can provide a native hypervisor? Anyone can, really. Hypervisors have become a market commodity with very little feature difference between them. With free, open source hypervisors like KVM, anyone can build on KVM to create a hypervisor unique and specialized to the hardware they provide in their hyperconverged appliances. Many vendors still choose to stay with converged infrastructure models, perhaps banking on the market dominance of VMware even with many consumers fleeing the high prices of VMware licensing. Saving money is only one of the benefits of HCI. By utilizing a native hypervisor, the storage can be architected and embedded directly with the hypervisor, eliminating inefficient storage protocols, files systems, and VSAs. The most efficient data paths allow direct access between the VM and the storage; this has only been achieved when the hypervisor vendor is the same as the storage vendor. When the vendor owns the components, it can design the hypervisor and storage to directly interact, resulting in a huge increase in efficiency and performance.

In addition to storage efficiency, having the hypervisor included natively in the solution eliminates another vendor which increases management efficiency. A single vendor that provides the servers, storage, and hypervisor makes the overall solution much easier to support, update, patch, and manage without the traditional compatibility issues and vendor finger-pointing. Ease of management represents a significant savings in both time and training from the IT budget. 

What about the Cloud?

Cloud computing has been around even longer than HCI and many have already begun implementing the cloud into their IT infrastructure in various ways. Most market indicators are pointed toward organizations using a combination of on-prem infrastructure with cloud-based infrastructure or services in what may be called hybrid cloud architectures. As a fully functional virtualization platform, HCI can nearly always be implemented alongside other infrastructure solutions as well as integrated with cloud computing. For example, with nested virtualization in cloud platforms, an HCI solution can be extended into the cloud for a unified management experience. Not only does HCI work alongside and integrated with cloud computing but it offers many of the benefits of cloud computing in terms of simplicity and ease-of-management on premises. In fact, for most organizations, HCI may be the private cloud solution that is best suited to their environment. Like cloud computing, HCI is so simple to manage that it lets IT administrators focus on apps and workloads rather than managing infrastructure all day as is common in 3-2-1. HCI is not only fast and easy to implement, but it can be scaled out quickly when needed. HCI should definitely be considered along with cloud computing for data center modernization.