For a growing number of organisations, the default option for new IT projects is the cloud and cloud storage, with even the more cautious businesses have adopted “cloud-first” or “cloud-preferred” strategies.
That’s because startups and fast-growing firms can build infrastructure in the cloud that would be hard to achieve quickly in any other way. The desire to promote the benefits of cloud can include even the vendors most often associated with on-premise technology – yet the cloud is not always the right choice.
CIOs have come to a more nuanced view of IT architecture, with many firms finding that the cloud is not as cheap – or even as flexible – as was once thought. There are situations where the cloud’s shortcomings means it is not always the answer.
Does cloud cost less, or more?
A big draw of cloud is its reputation for saving money, but the cloud does not always cut costs. In fact, for some workloads and types of IT operations, cloud can be more expensive than on-premise systems, a conventional datacentre, or co-location.
The problem, says Jon Collins, vice-president of research at analyst GigaOM, centres on a lack of forward planning and low discipline over how the enterprise uses cloud resources.
“Cloud does not save money per se, but it has been historically perceived and marketed as cheaper,” he says. “Less suitable workloads have been moved to cloud with associated migration costs, and no necessary gains. Meanwhile, cloud use has caused a proliferation of applications without any real cost assessment up-front, and that has resulted in cloud costs spiralling out of control.”
The growth of FinOps shows that CIOs trying to catch up with financial governance. Firms pay a premium for the cloud’s flexibility, but over time this can cost more than on-premise options. Then lift-and-shift of entire workloads or virtual machines to the cloud often means paying more without flexibility gains in return.
What makes cloud cost more?
Porting applications that weren’t designed for the cloud often causes performance issues and higher costs.
Some applications are just harder to run efficiently in a cloud environment. These include workloads that make extensive use of (non-cloud) databases, are very I/O intensive, or require connections to physical inputs such as internet of things sensors, or human decision-makers. If data and processing both happen in the cloud, these issues can be minimised, but problems arise when they are not.
Workloads that need very robust connectivity are usually better running over a local network or in a datacentre than via a public internet connection to the cloud. Examples include enterprise resource planning (ERP), supply chain management and manufacturing systems, as well as financial, healthcare, transport and critical infrastructure applications.
Performance can also suffer when an application has the potential to run in the cloud, but it is not optimised for it. Rewriting those applications to make effective use of the cloud – especially if they are already running well – can be both expensive and disruptive.
When is cloud difficult to manage?
From an IT management perspective, cloud infrastructure needs the right tools and skills to be able to run.
An IT department might, for example, have just the right skillsets to manage its in-house storage infrastructure to maximise uptime and utilisation. They will know in detail how their on-premise arrays and SANs work, and how to get the best from them.
With the cloud, users have less granular control over the IT assets they operate. Vendors have improved cloud management tools, but it is still the case that hardware is optimised for the operational requirements of the cloud provider and “average” workloads.
Even if IT teams can fine tune how they run cloud infrastructure, they face having to do so using multiple tools. They also need to understand how vendor performance tiers, availability and provisioning work. These may vary and be complex when taken together.
Analysts also point out that cloud-native applications can be more complex than conventional on-premise systems, and so consume more management resources.
CIOs also have to deal with a loss of control over IT assets, whether in the form of SaaS applications or more complex shadow IT set up by individual departments.
What are the shortcomings of cloud in data protection and compliance?
One area where cloud providers have closed the gap with on-premise technology is security and compliance.
Cloud providers have invested heavily in security over the past decade. Their size and scale allows them to attract strong talent and deploy the best defensive technologies. Cloud providers cannot afford to be victims of cyber attacks, and their security measures now are on par with or exceed those of all but the most security-focused users.
Nonetheless, there are still drawbacks to moving to the cloud from a data protection and compliance perspective.
Although cloud providers are built for resilience and offer high levels of availability, that is for the infrastructure as a whole. They provide less – or even no – protection for a customer’s data at file level, and so firms still need to invest in backup, recovery and local data protection.
Customers also need to consider where data is stored. Geopolitical events, as well as law and regulation, have made data sovereignty a big issue, and not just in industries such as healthcare or banking. Although the hyperscalers have responded, there are some applications where it is easier to store the data on-premise, at least for now.
How do I align business strategy and use of cloud?
The biggest potential drawback of the cloud is not technical at all. It is that organisations fail to align use of the cloud with their business goals. This can cause or exacerbate the financial, management, operational and data protection drawbacks of a cloud deployment.
It’s not that cloud isn’t a good solution, but it is often applied to the wrong business problems and in ways that don’t make the most of the cloud’s advantages.
“Organisations have seen cloud as a way of driving digital transformation,” says GigaOM’s Collins. “This could never be true in isolation from business strategy. For cloud to be the answer, it needs to be adopted strategically, which means business understanding and involvement.”