Costing cloud and weighing up use cases

When considering a move to cloud, it is important to be realistic about the real total cost of ownership (TCO) of cloud versus the alternatives, and to determine the business case for doing so.

Moving to cloud is not necessarily the right solution for certain businesses, infrastructure, or applications, and it may not always prove as cost-effective as many people think.

In fact, in some scenarios, the cost of running in the cloud could prove up to 10 times higher than running in a hosted environment or on-premises over a period of five years.

Organisations should undertake a careful needs analysis and calculate the total cost of ownership over an extended period, building in the cost of hardware, software, management and the cost to the business of potential downtime.

It is not as simple as looking at operating expenditure (opex) versus capital expenditure (capex).

There are specific instances where the cloud makes good sense. For example, for businesses with elastic compute requirements, such as an online retailer that has Black Friday specials and needs the ability to scale up quickly.

Cloud is also useful for organisations that simply don't have the budget for new hardware capex.

Many companies expect to 'just put everything in the cloud and the cloud providers will look after it'.

But there are misconceptions that the cloud offers dramatic savings and will allow companies to eliminate the cost of buying hardware and maintaining it. Many companies expect to 'just put everything in the cloud and the cloud providers will look after it'.

When taking the app to the cloud, companies are essentially renting compute power − the servers they spin up there still need to be patched, maintained and monitored, so they still have that ongoing maintenance cost.

Let's take the example of a typical medium-sized company with 100 employees, which is considering a move to cloud due to ageing infrastructure.

Its monthly costs for cloud-based compute power would amount to between R32 000 and R56 000, and it would still have to pay the resources it currently has to run patching and maintenance.

So that cost amortised over five years, with escalations of 6% per year, gives a total cost of R2.5 million over five years − just to rent compute power.

In contrast, if you spec the same requirements and buy the hardware, and have it hosted in a data centre, the cost over five years would be R250 000.

Downtime alone can add significant costs to the IT environment. In an exercise to put a number on it, factor in a hypothetical server going down halfway through a working day, plus the following day added as a waste as the imaginary company with arbitrarily 50 employees tries to bring services live again. This simple scenario would cost approximately R202 500.

The money lost in half a day alone would virtually cover the cost of an additional server to replicate the company’s production environment into a disaster recovery setting. This situation would alleviate concerns around a single point of failure associated with a company using its proprietary infrastructure.

The software-as-a-service model offers a solution to many companies. This 'infrastructureless' cloud has many advantages − such as the fact that the software provider takes care of all the maintenance, patching, hardware and all the backups.

However, this is not ideal for some applications, as companies do lose a certain amount of control − for example, you might not be able to plug in backup software to a particular application to extract that data and pull it into agnostic servers. So, business owners are trusting this software provider with everything. This raises the question of how to manage that risk.

SaaS could also prove problematic where operations depend heavily on an application, and power cuts or Internet downtime could bring the business to a standstill.

In circumstances like this, the business might look to both a fibre and microwave link to ensure continuity; or it might find that on-premises hardware and applications are all that is needed.

The cheapest option is still to have a server on site, and there are situations where that makes sense: for example, a very traditional business running just one mission-critical system in one office. Not only would it be cost-effective, but if the power and Internet connection went down, it would still be able to work.

It is important for organisations not to allow themselves to be 'pushed into cloud'. Instead, they need to consider how to build their environment to ensure it supports the business's needs and growth plans, in the most cost-effective way.

They need to look at it strategically within the context and nature of the business and how it operates, considering the long-term total cost of ownership, and the cost of potential downtime.

Rather than getting caught up in all the fancy new bells and whistles, they need to design something that is technically sound and gives the uptime and reliability they expect. Finding cheaper ways to deliver the required compute power and uptime is just the cherry on top.