Server virtualisation is far from dead

Container-virtual machine convergence

Hypervisor-based server virtualisation has repeatedly demonstrated the ability to expand functionality and capability to support emerging infrastructure requirements. In response to innovation and competition from containers, both traditional and emerging virtualisation providers are delivering cloud-inspired virtual infrastructure. Examples include vSphere 7 and Red Hat OpenShift Virtualisation.

Ultimately, converged virtual infrastructure holds the promise of more consistent compute capabilities across hybrid infrastructure software, multicloud, distributed cloud and edge computing. Gartner predicts that container-virtual machine (VM) convergence will be available on more than 75% of on-premise virtualisation hosts by 2026, up from less than 5% in 2021.

Container management

For early adopters of containers, reliable enterprise management at scale has depended on having the requisite container management functionality. However, as container-VM convergence increases, it may have an impact on existing operational tools and processes.

Investment in pure container management may introduce the risk of duplicate functionality and siloed processes. Instead, container-VM convergence holds promise to improve visibility and standardise on consistent, application programming interface (API)-driven automation and the benefits of managing modern infrastructure in a more consistent manner.

Cloud-hosted server virtualisation

Public cloud infrastructure-as-a-service (IaaS) is now taking hosted server virtualisation to a new scale. Cloud-hosted virtualisation is an approach that runs counter to the traditional deployment scenario operated in hyperscale IaaS, where the hypervisor is bundled as one part of an integrated set of cloud infrastructure services.

Cloud-hosted server virtualisation means that “bring/choose your own” hypervisor has become possible across hyperscale IaaS providers. Examples include VMware Cloud and Nutanix Cloud Platform.

Cloud-hosted will suit a specific set of potential rehost and refactor scenarios, as well as provide options for improving the elasticity and scalability of existing workloads, such as business continuity/high availability and datacentre consolidation. It will also serve as an on-ramp to cloud migration.

Edge computing

Edge computing is a distributed computing topology supporting information processing located closer to where the information is created and/or consumed. To do this, compute moves closer to distributed locations to meet the needs of applications that require a real-time response, including when the connection to the central datacentre or cloud service goes down or is unavailable for some time.

It is also done where the data volume is such that it is impractical to send to a central location without some form of pre-processing or filtering, as well as where the data cannot leave the edge location, because of networking issues, data residency/sovereignty, or a combination of both. A range of new solutions are targeting this requirement, including Sunlight and support for Arm processors.

Hyperconverged infrastructure

Hyperconverged infrastructure (HCI) software is a form of scale-out infrastructure that is designed to simplify and standardise infrastructure and IT operations. It provides a “building block” approach that includes a hypervisor and incrementally adds software-based storage virtualisation and can include network virtualisation. Given that it is software-based, HCI software allows a consistent hybrid solution spanning on-premise and off-premise, if and when desired.

Cloud-hosted HCI software overlaps with the trend toward modernised, cloud-inspired server virtualisation. This is happening alongside investment in container management by HCI software suppliers. Both are creating another pathway to adoption of distributed cloud, with emerging solutions from Microsoft, IBM and others.

Hardware assist for virtualised workloads

Hardware assist for virtualisation of industry-standard, general-purpose server hardware exists in many forms today, but it has continued to evolve, most recently through custom hardware engineering for hyperscale IaaS providers. In its many guises, hardware assist offers improved performance with varying use cases and associated costs.

Hardware-assisted virtualisation can provide a pool of shared compute resources that supports a range of existing and emerging server workloads. Of these, improved performance for artificial intelligence and machine learning is a popular emerging need. Examples include AWS Nitro, Intel, Nvidia and Pensando.

Consumption-based hardware

Hardware suppliers have responded to the shift to cloud delivery models by introducing cloud-like offerings in the form of subscription licensing and consumption-based infrastructure. Both cloud-hosted and consumption-based improve the ability to shift future infrastructure spending from capital expenditure towards operational expenditure – including for server virtualisation. Examples include HPE GreenLake and Dell Apex.

Michael Warrilow is vice-president analyst at Gartner