Dispelling the top five myths of modern infrastructure

Myth 1: The cloud is a real place

Cloud is a style of computing where scalable and elastic IT-related services are delivered to consumers via the Internet. Don’t think of the cloud as a location, but rather as a new computing paradigm where loosely coupled services come together to form applications. It is fundamentally different from traditional IT infrastructure.

There are many reasons why applications or data cannot be moved to public cloud infrastructure, including regulatory compliance, latency, the need for offline access, and more. But that doesn’t mean they can’t use the cloud model. Many IT shops will want cloud services, but may need them in their datacentres or at the edge. The question becomes whether you want to buy turnkey services from a provider or build your own.

Now that cloud-first is the expected approach, you should ask:  

• Will public cloud be cheaper? What benefits can I get from the public cloud? 
• Should I migrate because it’s better, even if it isn’t cheaper? 
• Are there regulatory, compliance or insurance reasons I can’t move to the cloud?
• Can distributed applications tolerate latency between sites and the public cloud? 
• Can I have unified management for all my infrastructure?

Myth 2: Infrastructure should be built independently

Historically, IT organisations built their own infrastructure: They bought servers, storage and network switches — often from three different vendors — and assembled everything themselves. But today that is usually unnecessary. You can simply buy a turnkey, ready-made infrastructure stack with little or no assembly required.

One way to do this is to move to the public cloud — but that's not the only way. Integrated systems — including converged, hyper-converged, dedicated and composable infrastructure — bring the benefits of prebuilt infrastructure to the data centre and edge.

Then there is “distributed cloud,” where hyperscale public cloud providers place a substation of their infrastructure at your site in order to make their native services available to applications and data running in your data centre. Thus, it is now possible to acquire on-premises infrastructure as a service (IaaS) and platform as a service (PaaS) directly from your public cloud provider.

You need to determine where the “build/buy boundary” falls for you. You should ask:

• What are the operational benefits of a turnkey system, and do these justify the price premium? 
• Will you be locked into certain vendors or products?
• Do you have domain experts on staff, who can build and maintain a bespoke infrastructure stack? If so, is that the best use of their valuable time?

Myth 3: Automation can save money

Infrastructure automation is no longer optional. It is a response to the increasing scale and complexity of modern, distributed infrastructures, which cannot be managed without it. 

However, many IT organisations pursue automation in the mistaken belief that it will save them money. Many organisations want to reduce infrastructure spending, often by reducing headcount; they expect automation to replace both manual effort and the people doing it. This is unlikely. Infrastructure automation initiatives tend to shift costs, not reduce them. This is still highly valuable, since the time and money are put to better use on more valuable projects. But infrastructure automation typically doesn’t lead to actual cash savings.

Automate your infrastructure to make it better, not cheaper. To quantify its value, ask these questions instead:

• Can you improve service quality (for example, fewer errors or better compliance)?
• Can you add customer value?
• What costs will you reduce or eliminate through automation?
• What new costs will you incur?

Myth 4: Infrastructure can be autonomous

Some vendors claim to offer “autonomous” infrastructure powered by “artificial intelligence” that replaces manual operations (and maybe human operators). But these claims never withstand scrutiny. True autonomy would require the system to make judgments; with current technology, this is not even theoretically possible.

In reality, artificial intelligence for operations (AIOps) serves as decision support for human operators. And in that role, it can be quite valuable. For example, machine learning algorithms can comb through huge volumes of data, finding correlations that a human never could. AIOps excels at predictive analytics, such as pinpointing when resources will be exhausted or helping to identify the root cause of failures.

AIOps tools can even execute predefined routines in response to certain conditions — presuming a human being devises a solution to a problem, writes code to execute that solution, and defines exactly when and how the AIOps tool is permitted to run that code. AIOps cannot devise a solution independently or fix the problems it finds without human intervention.

Thus, AIOps is potentially very useful, but it is not autonomous. Instead, you should ask: 

• Where would you benefit from automated decision support?
• How will AIOps tools integrate with your automation toolchain?
• How can you govern the behaviour of bots and scripts?

Myth 5: You are ready for change

The reality is that most enterprise IT organisations simply are not equipped to make the transition to modern infrastructure and operating models. Many veteran I&O staff still lack expertise with new technologies. In many IT organisations, the lack of skills has reached crisis levels. Some modernisation initiatives have already foundered for want of skilled engineers and architects. 

Gartner data consistently shows labour shortages at the top of the job market because there aren’t enough people with modern skills. Infrastructure-as-code, Kubernetes, DevOps, AI, and cloud infrastructure are the skills in highest demand for I&O technical professionals. But it may not be feasible to hire outside experts to fill these gaps; candidates with these skills are rare and expensive.

As it grows more difficult to hire skilled personnel — and as they command ever higher salaries — IT organisations will need to grow skills internally. These skills are best developed as part of a formal talent enablement program, which identifies and prioritises the highest-value skills to the IT organisation.

As you build your talent enablement programme, ask: 

• How will we change our culture?
• What new people, roles, teams, and organisational structures will we need?
• How will we incentivise staff and managers to invest in skills growth?

In summary, don’t buy into the hype around new technologies, but don’t miss out on their promise. The less familiar you are with a technology, the more critical your evaluation process becomes. Involve technical experts early and often, but make your decisions based on business outcomes. Take a bottom-up view of the bottom line.