agsandrew - Fotolia
Top infrastructure and operations technology myths of 2019
Admins are consistently evaluating technology to improve I&O efficiency. Cost, integration and business goals are key components to this process and cutting down any hype.
When admins research new infrastructure and operations technology, they're met with a sea of vendor information and industry hype. To navigate these waters, admins turn to industry analysts and colleagues to dispel myths and figure out what investments can truly benefit their organization.
With AI and machine learning trends, vendors are pushing "intelligent" offerings for self-healing infrastructure, said Paul Delory, research director for data center and cloud operations at Gartner's Infrastructure, Operations & Cloud Strategies conference.
This push for self-healing infrastructure a positive one, but there are questions admins should ask before diving head-first into any purchases -- and ultimately decide what infrastructure is right for their organization, he said.
9. Everything moves to the cloud; nothing moves to the cloud.
Over the next five to 10 years, organizations will have workloads in the data center, cloud and edge deployments, but analytics and security are the main priority – regardless of the infrastructure and operations technology stack, Delory said.
Of course, challenges to cloud adoption include data gravity, regulations and data sovereignty. Plus, latency and data storage needs that leave admins with questions surrounding storage capacities and unified data management.
8. Infrastructure must be multi-cloud
There are still questions around the meaning of true multi-cloud, Delory said. Most organizations have multiple clouds that they silo by use case or application. This isn't true multi-cloud from an admin or developer perspective, because they primarily end up working with one cloud tool on a daily basis or are locked into using one cloud provider.
When evaluating a multi-cloud setups, admins should ask if their organization has a good use case for multi-cloud, how much of an issue vendor lock-in is, should they use each cloud provider's native tools or cross-cloud tools, what their workflow will be and if they are getting the full benefits of each cloud offering.
7. Organizations should build its own infrastructure
For more traditional or risk-averse companies in industries such as finance or healthcare, admins spend a lot of time building their own on-premises data centers. Though with the emergence of integrated systems, not all organization-supporting infrastructure must be built from scratch.
"The first real integrated system we saw was … about a better delivery model for traditional infrastructure," Delory said.
He said that for purchasing infrastructure and operations technology, admins must define a build vs. buy boundary, and keep in mind which option provides the most benefit to IT operations.
6. HCI replaces other architectures
Delory said though there are benefits to hyper-converged infrastructure (HCI) the idea doesn't have tech valence at this point and has become a blanket term to include storage or software-defined infrastructure products.
"The early hyper-converged marketing was focused on being a storage product for a VMware environment," he said.
When working with clients interested in HCI, he usually clarifies what exactly the organization is looking to buy because offerings can range from full-stack software-defined infrastructure to enterprise-level storage technology, Delory said.
During HCI evaluation, admins should look at the operational benefits compared to existing infrastructure, HCI's ability to scale, if the organization can meet HCI network requirements and how much infrastructure HCI will replace.
5. CI replaces other architectures
According to Delory, composable infrastructure consists of truly disaggregated components -- such as CPUs, RAM and hard drives -- that are connected to a software composer, which allows admins to create servers from that specific resource pool via an API; this created server would be indistinguishable from any physical server.
Though the idea of converged infrastructure is promising, there's still some time before offerings reach this level of infrastructure and operations technology programmability, he said.
"If you look at products, [some vendor's] compute model is two Xeon CPUs and DRAM, which isn't much different from a blade server," Delory said.
The converged evaluation process should include surrounding virtualization needs, if the organization must regularly destroy and recreate infrastructure as well as if automation could bring the same benefits.
4. Containers replace virtual machines
Delory said that much like cloud, containers can replace older virtual machine (VM) technology, but for most clients they aren't the right fit.
"If you look at the two technologies and draw them out, there's actually not a whole lot of overlap. They're solving different problems," he said.
VMs are an infrastructure-level abstraction that include software-defined components such as network and storage. Containers, however, are an OS-level abstraction that help solve application delivery and operational issues.
3. Infrastructure automation saves money
The biggest misconception around infrastructure and operations technology automation is that doing so saves money and reduces headcount, Delory said. This often isn't the case, as running infrastructure automation requires hefty foundational and restructuring work in the data center.
In building a business case for automation, look at the work in three phases: efficiency gains, productivity improvements and cost reduction, he said.
"It's not going to eliminate work; it's going to shift it. Automate things because it makes your infrastructure better, not cheaper," he said.
2. Organizations must do DevOps
DevOps evolved to support agile infrastructure, Delory said. At its core, DevOps is a culture shift, which means that main barriers include policies, processes, available skills and off-the-shelf software tools.
Top DevOps skill sets include continuous integration/continuous delivery, cloud management, as-a-service knowledge and service delivery. Additionally, agile infrastructure and development models focus on short sprints to consistently address issues, patches and improvements.
If organizations decide to have DevOps, its managers must identify how closely their operational models align with agile infrastructure.
"If you don't have agile [production], then you don't have the problems that DevOps solves," he said.
1. AIOps is not fully autonomous (yet)
Within the past few years, the term "artificial intelligence" has made its way into vendor product offerings as a term for self-aware or automated, but using the technology is still in early stages. Still, vendors have tools for artificial intelligence in IT operations (AIOps).
"Generally, AIOps is used for decision support. But it's a species of monitoring and there's a long road between that and intelligent infrastructure," Delory said.
AIOps is approaching the Gartner Hype Cycle's Peak of Inflated Expectations, but should see more progress over the next five years.
To become more realistic for infrastructure and operations technology automation, organizations must address integration concerns. Admins must also address appropriate operational thresholds to let the software take actions for anomaly detection, root cause analysis, optimization analysis and intelligent automation.
