Containers for a post-pandemic IT architecture

Strategic stack

The Accenture Technology Vision 2021 report recommends that businesses aim to build strategic IT stacks that will enable them to innovate quickly and take advantage of new opportunities in the post-pandemic world. Enterprises need an adaptive technology foundation, and they can’t afford to be weighed down by legacy systems. In Accenture’s Technology Vision 2021, a cloud-native strategy and microservices are the key.

Yet while the CIO’s voice is more likely to be heard by the heads of business, there is a constant nagging – legacy IT. It simply is not feasible for most established businesses to start afresh with an entirely new, cloud-native software architecture. They appear to be shackled to architectural blueprints, some of which are decades old, which no longer accurately reflect business operations and are wholly unsuitable for modern agile working practices.

In a recent Computer Weekly article looking at the Technology Vision 2021 report, Shaheen Sayed, technology lead at Accenture UK & Ireland, recommended that businesses looking to become tech-driven need to ensure their tech stack is supported by a robust cloud infrastructure.

She wrote: “Not only does this emancipate them from the burden of legacy systems, it allows them to transform to a platform-driven business that can draw on numerous digital capabilities to quickly scale to the priorities of today and forecast for the needs of tomorrow.”

Sayed believes that for businesses to achieve the greatest return on investment from their cloud migrations, they must avoid settling for a limited “lift and shift” model. In other words, they need to make more of their applications cloud-native.

Michael Warrilow, research vice-president at Gartner, says: “Understanding of ‘cloud-native’ varies, but it has significant potential benefits over traditional, monolithic application design, such as scalability, elasticity and agility. It is also strongly associated with the use of containers.”

However, many organisations have spent years switching from bare-metal servers to virtual machines (VMs). VMs have enabled IT administrators to take advantage of new, more powerful hardware to run multiple virtual servers in a single server rack, and because the VM replaces a physical server, server applications do not normally require modification.

Growth in containerisation

Gartner predicts that by 2022, more than 75% of global organisations will be running containerised applications in production, up from less than 30% today. The analyst’s figures are reflected in the latest Red Hat Enterprise Open Source Report 2021, which shows container adoption is already widespread. Of the 1,250 IT leaders surveyed, just under 50% said they use containers in production to at least some degree. A further 37% use containers for development only, while just 16% are still evaluating or researching container adoption, according to Red Hat.

“Containers are not a nascent trend any longer,” says Red Hat technology evangelist Gordon Haff. “In this past year, we have seen a lot of companies go to containers.”

While it took the best part of a decade for server virtualisation to gain mass adoption, Red Hat’s survey shows that containerisation has been gaining momentum in the enterprise. Over just a few years, it has become an important component of enterprise infrastructure. 

The first step in modernising infrastructure is to look for opportunities to move VMs from on-premise datacentres into the public cloud, where they can continue to run using infrastructure as a service (IaaS). Public cloud providers such as IBM offer automatic provisioning of servers and a VMware virtualisation layer in the cloud, which allows IT administrators to retain full control of their software stack.

Suraj Kumar, general manager of API, integration and microservices at Software AG, believes that simply moving a VM to the cloud will not give organisations the elasticity and modularity that helps them cope with an ever-changing business environment. 

Monolithic applications

Traditionally, VMs have helped companies separate monolithic applications from physical hardware. “We are now seeing more organisations breaking down monolithic applications into smaller services,” says Kumar.

This deconstruction of a larger application into smaller building blocks run in their own containers yields benefits in terms of flexibility and scalability, he says. If there are bottlenecks in a certain area, these can be scaled at a more granular level, instead of having to scale the entire monolithic VM.

Any discussion about containers usually leads on to discussions about container orchestration using Kubernetes. Running a small number of containers may be manageable, but the true value of containerisation is that it provides an elegant approach to delivering compute, storage and networking for microservices.

A cloud-native application may be based on hundreds of microservices, each of which needs monitoring and orchestrating. This is where Kubernetes fits in.

Red Hat’s report found that the use of containers and Kubernetes is likely to continue growing, with 30% of IT leaders expecting to significantly increase their container usage over the next 12 months and a further 42% expecting to slightly increase their container use.

Kubernetes is overwhelmingly seen as important to cloud-native application strategies for its container orchestration – two-thirds view it as “very” or “extremely” important, and 19% consider it “important”.

However, as James Sturrock, director of systems engineering at Nutanix, warns, the main challenges are that Kubernetes is deep and complex, and evolves quickly with its growing ecosystem of technologies, which means many organisations do not have in-house expertise and struggle to keep pace with developers’ needs.

“Without the right tools, the cost and management burdens can skyrocket when it comes to managing Kubernetes and cloud-native applications across different clouds,” says Sturrock.

There are a number of tools that promise to accelerate the containerisation of applications. One of these is Migrate for Anthos, which Google claims can deliver containerisation in hours or days.

One of Google’s customers, UK newspaper The Telegraph, used Migrate for Anthos to accelerate its application modernisation strategy. In a blog post describing the initiative, Andrew Gregory, systems engineer manager and Amit Lalani, senior systems engineer, said: “The Telegraph was running a legacy content management system (CMS) in another public cloud on several instances. Upgrading the actual system or migrating the content to our main website CMS was problematic, but we wanted to migrate it from the public cloud it was on.

“With the help of our partners at Claranet and Google engineers, Migrate for Anthos delivered results quickly and efficiently. This legacy (but very important) system is now safely in GKE and joins its more modern counterparts, and is already seeing significant savings on infrastructure and reduced day-to-day operational costs.”

A mixed world of IT infrastructure

As containers become mainstream, it is highly likely that VMs will remain a core part of IT infrastructure. Ed Hoppitt, Europe, Middle East and Africa (EMEA) director for apps and cloud-native platforms at VMware, says: “The interest today isn’t about how I migrate, but what I should migrate. The majority of conversations are about what is the right thing to do: what should I run, where and why?”

The company’s Tanzu product enables organisations to manage containers and VMs using the same management console. Hoppitt says containers and VMs both offer a way to encapsulate compute, networking and storage to a workload. “Do I retain workloads, replace them, rehost, replatform or refactor them?” he says. “It may not always make sense to rewrite something as a containerised application.”

What is apparent from Computer Weekly’s conversation with Hoppitt is that organisations deploying containerisation sometimes come across similar problems to those faced by early adopters of virtualisation. There really should be no need to reinvent the wheel.