On-premise infrastructures have a new architectural paradigm—dHCI. It’s a natural progression from Converged Infrastructure (CI) and Hyper Converged Infrastructure (HCI), which have been around for a while. Now, vendors are starting to come out with Disaggregated Hyper Converged Infrastructure (dHCI). For reference, let’s start from the beginning.
Converged Infrastructure (CI) refers to a set of hardware components for storage, network, and compute, which is custom built to work together in a streamlined way. The intended benefits are to lower costs, centralize, and simplify management, decrease complexity, and leverage automation.
Hyper Converged infrastructure (HCI) takes the integration a step further and collapses both storage and compute into a single uniform appliance and often uses commercially available servers and introduces a virtualization abstraction layer to allowing for a software defined architecture. The systems are then made stackable, to expand capacity, yet maintain centralized management and control. The downside to this, is a loss of expansion flexibility.
In comes dHCI. Disaggregated Hyperconverged Infrastructure (dHCI) attempts to leverage the best of both CI and HCI. Much like CI, dHCI decouples the storage components from compute allowing them to be sized individually, and like HCI, it uses a software defined architecture on specialized appliances. These two features allow the greatest flexibility for growth, and accommodations for changing user needs.
There are a few key things that differentiate dHCI from HCI and even CI:
First, the Hardware. dHCI solutions lean much heavier on high-speed network switches to connect the storage and compute devices together, as opposed to having shared bus access of HCI appliances. Both Dell EMC and HPE Nimble have appliances for both storage and compute and they can both be managed by the same cluster orchestration application.
Second, we have the Hypervisors. The bigger players all leverage VMWare as the hypervisor, at least as an option and for many it’s the standard option, although KVM, Hyper-V, and even proprietary hypervisors like AHV and FusionSphere are available through some vendors. They each have their own vCenter plugins which provide different levels of features for managing both the appliances, and the resources on them. Vendors often tout some level of machine learning/artificial intelligence that monitors system parameters, flagging potential configuration errors, and warning of system health concerns, and simplified resource deployments.
The Storage management has become quite sophisticated. Although not limited to dHCI, some non-converged systems also include features like policy based storage protections, and applicability to individual data elements like vVols and data stores.
Lastly, dHCI is quite deliberately a cloud-oriented technology. All of the major vendors have mechanisms to integrate with all of the cloud platforms and leverage into hybrid cloud architectures.
dHCI is some ways represents the pendulum swinging back to flexibility to add storage and compute independently. With the maturity in the field, it brings with it the integration simplicity of modularized hardware components.
Written By William Vest, Sr. Systems Administrator