We stand close to a decade after the introduction of hyper-converged infrastructure and have a more definitive answer about where it fits into the enterprise IT landscape. Like many other new methodologies and technologies, HCI was hyped to replace everything that had come before and to solve all IT problems. As is usually the case, the reality was less than the hype, and we eventually worked out where hyper-converged infrastructure fits: among other platforms.
There are a few properties of HCI that bring value to the data center:
- Simple deployment
- VM-centered management
- Policy-based management
- Scale-out architecture
- Single-vendor platform support
These benefits make HCI an ideal platform for a variety of common virtualization use cases from server consolidation, virtual desktops, edge and remote office/branch office (ROBO) deployment to testing and cloud migration, and even data protection and analytics workloads.
Dig deeper on the following top hyper-converged infrastructure use cases.
There is a lot of HCI sold to provide a general-purpose VM platform to consolidate and contain servers with a common system. One part of the value is in VM-centered management with HCI; infrastructure administrators can focus on the VM service rather than managing a storage network or RAID groups.
Most HCI is ideal for server consolidation with tier 2 or 3 workloads, without the requirement for a lot of fine-tuning of VM resources or configurations. Tier 1 applications will demand more detailed management of the platform and VMs, possibly compromising the value of the simple, policy based VM-centered management on HCI.
VDI and desktop as a service
Virtual desktops are a classic scale-out workload, with hundreds to thousands of identical VMs that often get busy all at the same time. The related benefit of HCI is scale-out capacity; each HCI host is an island of compute and network resources and adds to the pool of storage resources. If three HCI hosts can deliver 250 desktops with good performance, then 12 hosts can handle a thousand of the same desktops, and 60 hosts -- in multiple HCI clusters -- can probably handle five thousand desktops.
HCI has no scaling inflection points, where you would need to buy a different resource such as a new storage array. HCI allows easy scaling as you grow your deployment or scale as you sell a desktop as a service. This coupling of resource purchase to demand and value is a crucial result of using HCI.
One of the challenges of a conventional virtualization platform is that dedicated storage arrays don't scale down very well; a 1U storage appliance seldom has all of the features of an enterprise array. When your edge computing environment needs a couple of x86 servers for virtualization, you want fully featured shared storage.
Most HCI platforms offer almost the same features in their smallest deployment as in their largest. A three-node HCI deployment will have all the features of a larger cluster and will integrate into the same VM management by policy. The same HCI platform can scale out for larger edge requirements with the same features and management systems.
ROBO deployments are another resource- and cost-constrained environment. HCI offers the ability for a highly available virtualization cluster with just two to four physical servers.
For a truly cost- and space-constrained deployment, some HCI platforms support a single-server deployment with high availability through replication to another location. A crucial factor is having HCI management that spans dozens or hundreds of ROBO locations into a single view for the central IT team.
Test and development
Test and development environments are usually a reduced-size copy of the production environment and often have tighter budgets than production. A handy feature for a nonproduction platform is automation to allow a new and clean copy of production to be deployed regularly.
Most HCI platforms have great automation tools, and most allow replication from a production cluster to another cluster for nonproduction tasks such as test and development.
One of the attractions of cloud computing is that you don't need to worry about infrastructure, but the challenge is that on-premises applications can't be simply moved to the cloud and become cloud-native.
Managed HCI platforms on public cloud provide a migration path from on premises to the public cloud without needing immediate refactoring of applications.
Backup and DR
The VM-centered management of HCI often includes VM-centered replication within an HCI cluster or from one HCI cluster to another. Replication within a cluster is commonly used for backup and data protection, often with storage efficiency features such as deduplication and compression.
Replication to another cluster, preferably in another physical location, is excellent for DR. Some HCI platforms even include automation for the DR failover, while others use hypervisor failover products.
Logging and analytics
One of the early criticisms of HCI was that CPU, RAM and storage all scaled together, so a workload that required a disproportionate amount of one resource would demand the unnecessary purchase of the other resources. To accommodate storage-centered applications such as log aggregation and analytics, most HCI vendors have storage-dense nodes either to simply expand storage capacity without compute or to provide clusters with a higher ratio of storage to compute.
The distributed storage architecture in HCI is particularly well suited to read-heavy workloads such as analytics or the sequential write behavior of logging applications. For applications that will progressively ingest data, requiring more capacity over time, the scale-out-as-you-grow nature of HCI will allow you to pay as you grow and, thus, a better TCO than buying five years' worth of storage capacity upfront.
There might not be one HCI to rule all enterprise virtualization, but there are many hyper-converged infrastructure use cases. If you have a scale-out workload or don't want to think about managing anything but your VMs, then HCI might be for you.