Mustard magnate Jeremiah Colman made his money, it is said, not from what his customers ate but rather from what they left on the side of their plate. It’s the same with cloud. A technological approach that was supposed to save huge amounts of money has now become an industry that makes much of its profits from waste, as businesses spend thousands and millions of pounds on systems, applications, and processing that they don’t need - and for which they can’t account.
Today, most businesses have adopted cloud in some way, but many cannot determine if they are spending the ‘right’ amount, or are unable to understand or allocate their expenditure with any degree of accuracy. In a business that spends seven figures a month on cloud, it’s common for a quarter or more of this money to be spent on resources that are underutilised or lying idle - the equivalent of mustard lying untouched by the side of the plate.
Often, huge amounts of resources are wasted simply through an inability to effectively monitor resource usage. One company ran up a $300,000 bill over a summer when their autoscaling diligently spun up extra resources to cope with demand, and then failed to turn them off when they were no longer needed. Another forgot to spin down their cloud instances after a load testing session, and wasted tens of thousands over a long weekend.
It wasn’t supposed to be like this. When cloud was in its infancy a decade ago, one of its chief selling points was the enormous cost savings over traditional models of IT procurement. By slashing CAPEX on infrastructure and software and providing transparent OPEX, cloud was supposed to drive down costs while providing clear accountability over what was spent. How did we get here?
One main reason is that companies cost savings are no longer the primary driver for cloud migration. Today, businesses are adopting a ‘cloud first’ strategy so that they can innovate faster, cope with ever-increasing data volumes, and harness new technologies such as the Internet of Things. Cloud is less about cost and more about squeezing out the last drop of competitive advantage. As one VP told me: “It’s all about acceleration ahead of optimisation - until the bill hits $1m a month.”
The budget problem
Another key factor is the increasingly complex decision-making process involved in cloud purchasing. In the old world of datacentres, a select handful of people were responsible for modelling the migration and allocating resources. Nowadays this process involves dozens, even hundreds of people from the CFO and CIO down to engineers, devops, and heads of department such as the CMO, who all get a say in whether or not to run Reserved Instances.
As a result of this complexity and devolved responsibility, businesses are incapable of effectively tracking their cloud budgets, can’t separate their CAPEX from their OPEX and, ultimately, are unable to evaluate the cost of goods sold (COGS) - which is critical to improving gross margins across the business. Furthermore, this lack of visibility prevents organisations from optimising the costs of their cloud estate. Running a single server comes with dozens of associated charges each hour (including compute, network, storage, IOPS) each with significant scope for waste and optimisation, as the earlier examples show.
Agility and innovation may have replaced costs as the driving force behind cloud migration, but that doesn’t mean that a business can countenance wasting resources. It’s not uncommon for businesses to spend 25 per cent more on cloud than they actually need; given that Gartner estimates the public cloud market to grow to around a quarter of a trillion dollars by the end of this year, there are savings of around $60 billion to be made. How, then, can organisations start to seize back control over their cloud spending?
First, and most importantly, businesses must establish a comprehensive roadmap for their cloud migration where they determine their forecast expenditure and then map this against monthly goals. This should include total spending mapped against Reserved Instance (RI) coverage; account for wastage, such as underutilised resources; and assign spending allocation to each team to ensure that every business unit is accountable for the resources it uses.
Controlling cloud spend is about giving people the tools they need to track their resource use, which is why it’s crucial to give teams the daily data they need to make better decisions. It’s important to remember that cloud optimisation is an ongoing process of improvement, rather than being a matter of ‘fix-and-forget’.
Organisations should also appoint a cloud optimisation “Tsar” (or team), exclusively focused on ensuring that it first develops a financial model for the cloud, and then tracks against pre-determined performance metrics such as RI coverage. By effectively managing Reserved Instances in this way, a business can save between 30 and 60 per cent of their compute spending in the cloud. For a company spending between half and one million pounds on their cloud estate, the ROI of appointing a dedicated optimisation team should be abundantly clear.
Automate as much as you can
Remember, though, that financial success in cloud isn’t a single person’s responsibility: it is about creating a culture of cost management throughout the organisation - one that extends beyond the engineers and devops team, but involves everybody responsible for cloud decision-making.
Not all of this work can be done through dashboards and push-button applications. The goal of any cloud optimisation initiative should be to automate as many of these processes as possible, removing the human element of decision-making and establish a system of continuous improvement.
It is through steps such as these that some of the world’s biggest companies such as Uber, GE and Cimpress have saved hundreds of thousands of dollars a month, while enabling CTOs to finally understand which business units are accounting for their actions and spending in the cloud.
You might think that the only loser from cloud optimisation would be the big vendors themselves. However, unlike mustard makers, it’s in their interests to help their customers combat the costs of cloud, and deliver computing that delivers on the promises of ROI that’s always been one of cloud’s strongest selling points.
J.R. Storment is Co-founder of Cloudability
Image Credit: TZIDO SUN / Shutterstock