The use of hyper-converged infrastructure (HCI) is nearly universal today, both in enterprise on-premises datacenters as well as forming the basis for services available from public cloud providers. We all know that tightly coupled hardware and software in HCI is key, but what has become apparent is that the foundation for the numerous benefits of deploying HCI is, ironically, the oft-forgotten ‘I’ of HCI – the actual infrastructure. Some call the infrastructure ‘compute platforms’, others ‘infrastructure solutions’, but let’s face it, we all know these machines as servers, so let’s call them that.
The original hyper-converged infrastructure, we’ll call this HCI 1.0, is built on servers that contain a small number (at most 24; typically 12 or fewer) of storage devices. The small number of devices are typically not all flash and rarely utilizing the latest protocol of NVMe SSD. These servers are popular, and commonly available. However, they are primarily designed to provide compute resources, as opposed to I/O and storage – which has created an imbalance in the world of HCI. So, many enterprise customers that have deployed hyper-converged infrastructure for some workloads, still rely on a traditional 3-2-1 SAN infrastructure for other more storage intensive workloads– thus defeating the very purpose of implementing a simple, single vendor HCI.
- HCI 1.0 meets a need for SMB companies with small databases or single workloads like VDI
- HCI 1.0 has been built on commonly available servers – compute first, storage second
- HCI 1.0 doesn’t meet the need of enterprise environments with storage intensive workloads
- HCI 1.0 isn’t meeting simplicity, single vendor, or ease of management for the enterprise
So, we can agree the intentions and execution in small environments have been on point for HCI– but there is room for improvement for the enterprise environment.
Hyper-converged infrastructure should be able to:
- Replace your SAN Infrastructure
- Have a load carrying capacity to handle a multi-tenant of workloads big and small
- Enable the latest storage device protocol in NVMe at the lowest cost point
So, is there a server out there that can meet what hyper-converged infrastructure should be, and not just what it has been in HCI 1.0? The simple answer is yes, there is. A new approach to server infrastructure where storage AND compute needs have been considered is here. Axellio’s FabricXpress™ platform is that server. We want to put infrastructure back in hyper-converged infrastructure – this is HCI 2.0. HCI 2.0 brings storage capacities that rival the traditional SAN scalability within a small footprint, combined with the consolidated networking & compute and features of HCI 1.0.
In our next blog we will visit the next generation infrastructure of HCI 2.0. This new approach involves this notion of more devices (72 NVMe SSDs per server), supported by a PCIe fabric that enables those devices to be exploited, thus enabling a lowered cost point and greater performance. We will break down the technical aspects of how a larger device count contributes to a lower cost – although it may seem counter-intuitive.