X
101660

HCI automation is widespread despite architectural challenges

April 7 2021
by Christian Perry


Introduction


The rise of infrastructure automation continues, placing IT teams on a level of efficiency that can mirror the invisible nature of cloud environments. Hyperconverged infrastructure (HCI) is increasingly viewed as an on-ramp to infrastructure automation that avoids the complexity involved with deploying automation tools and procedures on traditional three-tier infrastructure. Current HCI customers are automating a wide range of tasks, including resource provisioning and application deployment, that are crucial to creating efficient, highly responsive IT ecosystems.

The 451 Take

Most HCI customers already engage in IT automation to some extent – the practice is vital to ongoing digital transformation efforts in place within a majority of today's organizations. Innate, smooth automation capabilities in HCI platforms can prompt even the staunchest opponents of automation (for example, due to fear of skills displacement) to take the plunge into the automation of time-consuming tasks, including provisioning, backups and disaster recovery, among others. Although the fully converged nature of HCI generally lends itself well to automation, it will not necessarily carve a foolproof path into a broad software-defined IT ecosystem, since limitations on automation of some tasks – such as networking – still need to be overcome.

Infrastructure automation finds a home on HCI


A measure of success for any new infrastructure deployment by today's IT teams is the extent to which it can automate manual processes. According to 451 Research's Hyperconverged Infrastructure, Technology & Platform Innovation survey, HCI adopters are automating a wide range of tasks, including VM backup (38% of respondents), storage provisioning (38%) and VM sizing/provisioning (35%). The automation of these tasks is paramount to efficient day-to-day IT practices that cannot afford constant human attention in the face of relentless business requirements.

Figure 1
Tasks Automated on HCI Platform
451 Research, Hyperconverged Infrastructure, Technology & Platform Innovation

Of those, storage provisioning automation represents perhaps the most immediate inroad into broader datacenter automation, as the intrinsic complexity of storage automation on traditional platforms can prevent some organizations (particularly small customers) from doing it. Larger HCI customers are most likely to use storage automation on HCI, including 50% of those with at least 10,000 employees and 49% of those with at least 200 HCI nodes. Cloud-native use cases on HCI also appear to lead to more automated storage provisioning – it is used by 54% of organizations that deploy HCI on cloud services and 48% that use Kubernetes on HCI.

Other automated tasks on the rise on HCI signal a more comprehensive move into the truly transformative nature of software-defined IT. Anomaly detection and problem remediation are increasingly automated tasks in public cloud environments due to their ability to keep infrastructure running seamlessly and securely (particularly when combined with other tools, such as application security) – all without human intervention. These tasks will see the largest increase in automation on HCI platforms (in terms of percentage of organizations), compared with other common tasks, in the next two years.

Interestingly, these are potentially inversely proportional with alerts, which fewer customers expect to automate in two years. Alerts continue to form the foundation for effective infrastructure monitoring, ensuring that changes related to performance, availability, and activity on servers and other IT infrastructure elements are quickly brought to the attention of administrators. Although alerts and associated monitoring programs are generally effective for most IT teams, they can generate a significant amount of work for teams that choose to respond to alerts via manual intervention – for example, to disable or otherwise modify a service that exceeds a defined threshold for memory usage. Even simply clearing alerts and taking no other action takes valuable time from an administrator's day.

While automating alerts is a necessity in an age when no administrator has time to manually monitor infrastructure around the clock, it is the automation of the steps beyond those alerts that can prove truly transformative to an IT environment. So while alerts will still be present, some organizations will transition to automated problem remediation, where only the systems see and acknowledge the alerts, and take steps to better distribute resources or solve other issues, such as compliance violations. This will become crucial in busy datacenters, where even a lull in performance can be just as disruptive to business as an infrastructure outage.

Networking presents automation roadblocks


HCI customers are currently automating a range of tasks on these platforms, but there remains work to be done by vendors to ensure that the automation trend continues in the right direction. This need for continued improvement materializes in a comparison of technology features that customers consider important when selecting an HCI platform against HCI features that customers say need improvement. Perhaps most indicative are perceptions around networking, which is the second-most-common feature customers value in HCI platform selection (at 43% of organizations, behind ease of scaling at 50%). Meanwhile, networking tops all others in terms of features that need improvement, at 30% of organizations.

Although networking automation is widespread across traditional three-tier infrastructure environments, it is a work in progress with HCI, which continues to require manual configuration and intervention to solve network deployment and bottleneck issues – at least for some customers. Although east-to-west (or server-to-server) traffic is common in virtualized environments, storage traffic generally continues to use north-south traffic, which creates complexity but can nonetheless support automation across the different networking traffic routes. HCI, on the other hand, uses east-west traffic for both compute and storage. This can present efficiency and latency problems in a traditional datacenter architected to support the predominantly north-south network traffic of three-tier infrastructure.

While changes in network topology (such as increased buffers or a transition to spine and leaf networks) can help address problems introduced by HCI and potentially support greater automation, the fact remains that HCI deployments are typically managed by compute administrators with limited networking experience. As HCI deployments grow to support larger numbers of workloads, the networking traffic ultimately increases, and may require attention from dedicated network professionals. Considering that HCI has entered this strategic phase of workload deployment for most large customers, the inclusion of a software-defined networking element in HCI platforms will grow more important to help automate configuration of nodes and clusters and seamlessly solve traffic management issues.

Looking ahead


The ability to scale HCI across large datacenter environments is critical to the long-term success of the technology, and automation will play a major role in that course. Looking ahead two years, HCI could be in use for a who's who of automated tasks critical to the modern software-defined datacenter, including storage provisioning, application deployment, disaster recovery and anomaly detection. However, without continued innovation – especially on the networking front – HCI will remain only one component of the overall infrastructure picture, rather than dominating all of it.