Skip to main content

Why tales of cloud repatriation are largely wishful thinking

(Image credit: Image source: Shutterstock/bluebay)

What’s next in the development of cloud? There are a growing number of success stories, while public cloud providers are significantly increasing capacity and adding new services. However, others have begun talking about ‘cloud repatriation’ i.e. organisations moving some services back to on-premise infrastructure. There is also the growth of the IoT to consider - will the need for devices to process information in real time make cloud’s latency an issue?

The answers will not be found by talking to vendors, each of whom will be keen to push their particular agenda as the solution to every organisation’s needs, however diverse those needs may be. I suspect that cloud repatriation is wishful thinking by vendors experiencing a lack of server and storage sales. It suits specific use cases such as Dropbox or Netflix, who built their own data centres after exponential increases in sales made their public cloud costs soar. It might also be appropriate for very consistent and stable workloads that do not need scaling, although those could use reserved instances to ensure cloud costs remain under control. And it might be the result of a failed attempt to move services to cloud, for example after getting the architecture wrong.

In my experience of working with customers across a range of sectors and running a wide variety of applications, cloud remains the ultimate goal, and moving applications to cloud makes complete business sense. However, this will not all be SaaS, and it is definitely not one size fits all. Each organisation needs to find its own solution, driven by the applications used, the volumes of data handled and organisation’s future business strategy.

For most, the future will be hybrid cloud, making multi-cloud management a key requirement. The continued development of containerisation and general agreement on Kubernetes for orchestration across the major cloud providers will play a key role by making applications truly portable, thus encouraging migration between cloud providers. Edge computing is necessary for IoT and intelligent devices, but data analysis from these devices benefits from the scale and flexibility of centralised cloud processing and storage.

Three principles to avoid repatriation

To get cloud right, and avoid any potential need for repatriation in the future, organisations need to understand three things. First, cloud is not a single concept, but means different things to different people. It is pretty good for most applications, but it is not the best answer to everything, and other options are still valid. Cloud applications are designed around a number of assumptions which may not apply to some organisations and how they work.

Second, a move to cloud is not just an IT transformation – it requires a refocusing of the business, plus new skills and new ways of working, in order to succeed.

Third, cloud almost certainly will not save an organisation money unless that organisation fundamentally reengineers how it works. While cloud takes away in-house IT operations, it creates the need for new skills, such as billing management and managing one or more cloud providers to ensure they deliver the agreed service.

Applications will drive cloud migration

When considering cloud, SaaS should be the first logical choice. A good SaaS service will provide everything you need, be easy to use and costs will be similar or lower than in-house provision. Of course each vendor wants you to take their SaaS – which is fine if you can use it seamlessly. You need to consider where your data is going to be held (will it be secure?) and how the SaaS solution will work with other critical line of business applications.

For example, Microsoft Office 365 is in effect SaaS, and it makes sense for the majority of organisations that are using Exchange, SharePoint and MS Office. However, there are ramifications for other systems. Office applications in the cloud are patched and upgraded automatically. This means that if an organisation has another application which only integrates with an older version of Office, perhaps because the vendor has not yet developed an upgrade or no longer provides support, there will come a time when it will no longer talk to Office applications. This could be a significant problem without advance preparation.

If SaaS is not available, the next best option is PaaS, whereby you install an application on top of a managed database service or development environment. This requires the application to be using an up-to-date, widely supported database such as Oracle, SQL Server or MySQL. PaaS services providing 15-year-old Informix or ProgressDB environments are rather harder to find.

A third option is hosting on IaaS, which means moving the application, as-is or with minor enhancements, to operate from a cloud provider’s infrastructure. The only responsibilities retained in-house are licences and support from the application provider. Alternatively it could be run by a managed service provider, who will handle all the complexity.

A final option is to configure IT as a private cloud and run it in-house, ready to move to SaaS when a suitable solution becomes available. Some applications cannot be moved in the short or medium term as particular dependencies will require too much work to eliminate. When the choice is between paying, say, several thousand pounds annually to host an application as it is, or ten times that cost to redevelop it using an open API for cloud PaaS, the balance of benefit versus reward is clear.

Containerisation will help to address vendor lock-in

Once an organisation has moved services to the cloud and retired its in-house infrastructure, it has to accept what the chosen provider offers unless it migrates services again. There is a higher degree of lock-in to a cloud provider than to an IT vendor, and moving services between providers is not yet straightforward, as their services are not directly comparable. Vendors will naturally focus on their short-term income, rather than how applications will work together in the long term.

Fortunately, we are seeing the growing use of containers, which make it easier to move applications between cloud providers – although this won’t work for legacy applications, which will need to be fundamentally redeveloped. Cobol will not go into containers easily, whereas Python will.

As all the major cloud providers now support Kubernetes container management, organisations who have (re)developed their applications to containers will be able to take advantage of cloud broking between providers. This could either be an in-house role, if an organisation has the capability, or be provided by a third party.

Edge computing will still benefit from cloud

Although I’m a strong believer in the benefits of cloud for many applications, I expect to see a move back to decentralisation for the growing number of intelligent devices such as robotics in manufacturing. These devices are in effect small scale datacentres in their own right and need to process information in real time, so for them the latency of cloud is becoming a major issue and the need to have intelligence at the edge will increase.

This is simply on-premise computing re-imagined – the next step in the regular waves of centralisation and decentralisation which have characterised IT over the last 40 years. The growth of edge computing certainly doesn’t mean that cloud is dying. Each organisation will need to consider its own use case and choose the most appropriate solution, depending on how much real time processing they need. All will benefit from the scale and flexibility of centralised cloud processing and storage, from construction companies putting together consortia to deliver specific projects such as Crossrail and HS2, who require capacity for a finite amount of time, to public sector organisations who can hand routine applications to a cloud provider in order to focus on their core activities.

Even organisations working at the cutting edge of robotics and AI will benefit from cloud’s scale and capacity. However, their smart devices will need to rely on inbuilt intelligence, supported by cloud services.

Richard Blanford, chief executive, Fordway