Photo by PixaBay User: NeuPaddy
I was at a CIO roundtable in Portland a couple of weeks ago and during the introductions, one of the CIOs said, “We’ve been using hyperconverged for the past five years on a Vblock.”
My tongue still hurts from biting it. Could anything possibly be more unlike software-defined hyperconverged infrastructure (HCI) than a Vblock? As just one small example, upgrading to a new version of vSphere on a Vblock requires a complex and expensive upgrade to the Vblock itself – and the last I heard, only VCE is authorized to do this type of engagement. In contrast, Nutanix enables 1-click upgrades that not only upgrade the OS and firmware, but also the hypervisor itself.
I am surprised that some IT leaders are still confused about the difference between hyperconverged and converged infrastructure. But, then again, many industry experts still don’t understand how HCI has surpassed converged infrastructure (CI) in terms of almost every important metric. A recent ComputerWeekly article, Datacentre decisions: Converged vs hyper-converged infrastructure, for example states,“From a scalability perspective, converged systems are more suited to larger applications such as Oracle, Microsoft Exchange or SAP installations.”
Legacy Packaged Infrastructure (LPI)
Traditional 3-tier infrastructure, i.e. centralized storage, storage network and servers, is comprised of three physically and logically separate products. Each “silo” has a separate upgrade cycle and management domain. Interoperability concerns between the separate silos can prevent an upgrade of one or more of the components. The silos tend to make the products complex to install and manage and inflexible for changing workload requirements. It typically takes month of planning just for organizations to purchase the individual components, and then more time to integrate them effectively.
Over the past 5 years, legacy storage, server, and switch companies banded together to simplify the deployment process, and speed selling cycles. They worked together on “converged infrastructure” solutions that offered two key points of value: 1) life-cycle management software to validate components and centralize software updates, 2) common support entry points to eliminate finger-pointing in the field.
It was unfortunate that HP (at the time) coined the term “converged infrastructure” to describe this infrastructure packaging. It is, perhaps, the biggest misnomer in all of IT. There is not a molecule of converged infrastructure in “converged infrastructure”. The only convergence comes from the ability (sometimes) to cut a single purchase order.
So for sake of semantic sensibilities, I’ll refer to these “converged” infrastructure systems as Legacy Packaged Infrastructure, or LPI, because they don’t change the underlying legacy server, storage, or network architectures. While LPI does help reduce deployment risk, the fact that it relies on separate physical and logical systems means LPI offers no aggregate capital cost reduction, no reduction in software layers, and no opportunity for integrating support and consolidating management panes.
HCI: True Innovation
Hyper-converged Infrastructure is the fastest growing market in IT infrastructure – tagged by Gartner to be an $8.6B market by 2020. This is in stark contrast to the declines in both legacy SAN storage and in LPI that relies on traditional SAN or NAS for storage.
HCI takes all of the intelligence out of proprietary arrays and puts it into software that runs on standard server hardware. HCI trumps the LPI reduction in infrastructure deployment times by typically enabling provisioning of new “nodes” in hours if not minutes. But HCI also addresses all of the other challenges of legacy infrastructure.
By consolidating compute and storage, HCI eliminates the storage fabric while slashing rack space and power requirements. It dramatically reduces staff time in administering the environment and eliminates the requirement for storage specialists.
HCI can easily save 40% – 50% or more in both CapEx and OpEx vs. 3-tier infrastructure. And a scale out deployment model allowing customers to pay as they grow also benefits continued advances in technology, further reducing overall costs compared with SANs.
The most important attribute HCI brings to organizations may be the same type of increased agility as enabled by public clouds. Fast deployment, greatly simplified management and scalability unencumbered by the physical storage controllers of an array enable customers to quickly spin up new services. And as with public clouds, HCI manufacturers can quickly deliver new services into the software defined environment.
Unlike public clouds, HCI still allows customers to maintain on-premises control for security, governance and compliance. It also enables a much more customized environment enabling tailored SLAs for every application.
What About Performance?
Many customers, and even so-called industry experts, mistakenly believe that HCI is good for VDI, ROBO and smaller applications – but that bigger workloads require an array. This is just untrue. HCI has been widely deployed during the last few years for Splunk, Oracle, SAP, SQL Server, Exchange, and many other mission-critical applications
All Flash Array (AFA) manufacturers like to promote the idea that a centralized repository of flash is somehow superior to distributed flash in HCI. But putting flash in a proprietary array at the end of a network designed for the latency of magnetic media intuitively makes no sense. Flash should be directly connected to the compute as it is with HCI, not remotely attached where it requires multiple hops, protocols, and performance-constraining controllers to be accessed.
Not only can HCI handle the same resource-demanding workloads as LPI, but it brings added reliability and resiliency unmatched by legacy 3-tier environments. It eliminates downtime required for firmware upgrades along with a single point of failure challenge with storage access failover. The software-defined nature of HCI enables capabilities such as self-healing and ongoing data integrity checks.
While nearly every leading legacy storage and server manufacturer today offers an HCI solution, they nearly always market it incorrectly. They look at HCI as simply another option among their storage and compute offerings.
HCI is not an end in itself, but rather an essential underpinning of the ability for customers to bring the same type of agility, simplicity and pay-as-you-grow attributes of public cloud to on-prem data centers. Legacy infrastructure, whether “converged” or not, lacks the simplicity, fractional consumption, scalability, resiliency and cost structure necessary to enable an effective on-prem Enterprise Cloud.