What should businesses consider for a cloud-first strategy?

9 Aug 2021

Paul Meehan. Image: HP Enterprise Ireland

HPE’s Paul Meehan says organisations are adopting a cloud-first strategy without understanding the impacts or risks.

Cloud technologist Paul Meehan has more than 25 years’ experience in IT and seven years’ experience specifically in cloud. He is both a certified Kubernetes administrator and part of an elite global group of VMware certified design experts.

Meehan is now the hybrid cloud advisory lead at Hewlett Packard Enterprise (HPE) Ireland. In this role, he advises customers on adopting cloud technologies.

We asked Meehan how businesses can build success in the cloud, and the advantages that Kubernetes and hybrid cloud can bring.

‘Cloud-first is a common strategy today. However, an application assessment could result in a discovery that applications are unsuitable for cloud’

What do you believe are the key structures people need to put in place to be successful with cloud?

Organisations should start by examining their operating models to align them with their cloud strategies, to see where the gaps are. And if you don’t have a cloud strategy, start there.

Organisations are adopting a cloud-first strategy without having understood the impact or risks. Transforming to cloud without doing these things will result in significant cost, risk and operational issues that will need to be fixed later. And that is painful and costly.

Creating a cloud business office (CBO) that includes relevant stakeholders across a business allows visibility of all key governance issues as cloud is being adopted. A CBO also delivers ongoing optimisation of operations and cost, once workloads are running in the cloud.

Whoever you partner with in cloud transformation needs to be able to help you build out fully automated cloud landing zones with the right security, compliance and operational characteristics, regardless of industry.

Where do you feel the biggest challenges are with cloud adoption?

A combination of not knowing where to start and working in highly regulated landscapes such as government can stop initiatives before they start. Cloud-first is a common strategy today. However, an application assessment could result in a discovery that applications are unsuitable for cloud for a multitude of reasons. We see this across many clients we talk to and sometimes having to unpick a deployment delivered via shadow IT can be time consuming.

Close alignment and business interlock between application, strategy and operations teams is critical to ensure requirements for individual applications are understood. This allows appropriate guardrails to be designed into the operational model before those applications move to cloud.

HPE developed an approach called minimum viable cloud (MVC), which delivers services in full production on cloud. This includes deliverables such as automation, security models, runbooks and all other required assets.

What is the potential impact of Kubernetes and container technology on businesses?

Where once we had a virtual machine, now we have a container, but there is important terminology to consider. A virtual machine is a representation of a physical server in software. A container is a small instance of an application. Examples would be Spark or Kafka for data analytics and machine learning, or Nginx which is typically used as a web front-end.

A developer can assemble multiple application elements running inside containers, and provision them together, in a higher order object called a Kubernetes deployment.

This is what allows applications consisting of many sub-components to be scaled automatically on demand, and updated with very little impact to the application. Kubernetes provides the intelligence to manage all of this off-the-shelf, including discovery, scaling and updating containers within applications.

Application architects can focus on how applications should behave and where they should run. They do need to know how Kubernetes works to understand how applications need to be designed to leverage its features.

Public cloud services such as AWS’s Elastic Kubernetes Service, Microsoft’s Azure Kubernetes Service, or HPE’s Ezmeral container platform take away a lot of the platform management considerations. Ezmeral is the most advanced on-premises container platform on the market, as it is a fully integrated solution with persistent data storage tier, that can be used for AI and machine learning without data gravity concerns.

What are your thoughts on application portability in a hybrid cloud consumption model?

An application is typically made up of multiple components and this of course must include dependencies. I think it is a fundamental misunderstanding or expectation gap of multi-cloud that applications will enter and exit public cloud at will.

With egress charges applied for applications exiting public cloud, this is a prohibitive practice and it is much more likely that placement decisions will stick applications to where it makes sense to run them. The same applies to applications which are suitable for public cloud – they will likely stay where they are which is the right thing to do.

Many analysts advocate having a cloud exit strategy formulated, in case there are compelling reasons to move applications back to on-premises systems. At least then you can assume the best but plan for the worst, but ideally you will have run an assessment to ensure you make good decisions at the beginning. Public cloud has led to a huge increase in innovation and there are many reasons to use it, once the right decisions have been taken care of.

What are some of the biggest technology trends you see becoming prevalent in the next three to five years?

Serverless technology is based on even smaller objects than containers. Software code that is part of an application is broken into functions that can execute as required, saving compute cycles and additional cost.

Many customers are already consuming this technology in the cloud, but this also possible to do on-premises using some initiatives from the Cloud Native Computing Federation such as OpenFaaS.

It is clear we will end up with a combination of virtual machines, containers and serverless applications, depending on the business need. And of course, we will still have a small proliferation of mainframes and physical servers.

SD-WAN technology is becoming prevalent as CIOs look to decouple wide-area networking from underlying hardware, just as it was done with hypervisors five years ago. This allows networks to be built on-demand with policy-based automation with the right guardrails. The outcome of this will be the ability to connect different sites together to build services and applications without manual tasks slowing down the process.

In the open-source world and in the world of containers, service mesh technology is becoming a fundamental construct to create low-latency networks and allow applications to communicate internally or externally, in a secure manner no matter where they live. Solutions such as Istio and Linkerd will become mainstream technology in the next five to 10 years as cloud-native adoption takes hold.