Cloud cost optimization: the next generation of cloud management

Understanding cloud costs is a complicated business. A cloud application typically consumes a variety of cloud services, each measured and billed in several different ways. Originally, a number of third-party platform vendors sought to address this complexity by providing advanced reporting and cost attribution. But as we recently reported, cloud providers have improved their capabilities to such an extent that the market for cloud cost management has consolidated.

Now that enterprises have started to get their heads around cost management, the inevitable next step is cost optimization – automatically ensuring that consumers pay the lowest price for the same level of cloud service. A slew of companies have entered the market, and we think they are ripe for acquisition, with managed service providers the most likely buyers.

The 451 Take

The cloud has brought unparalleled access to technology and capabilities to anyone with a credit card. With this choice comes complexity, and unravelling costs has proven to be a challenge – optimizing them is the next opportunity. Enterprises manually looking to reduce costs face an uphill challenge – tools and expertise are needed, which can be provided by MSPs as part of their value propositions.

Cloud broker-dealers: Potential targets

Broadly, players in the cost-optimization market primarily act as either cloud broker-dealers or waste managers. Many act in both areas, but for clarity we place them in the most applicable category.

On-demand pricing is the stalwart of the cloud – consume what you want, and pay in arrears for what has been consumed. But most cloud providers also offer alternatives that can provide discounts on average of 29%, according to 451 Research's Cloud Price Index. AWS's – and more recently, Microsoft's – reserved instances allow users to pay upfront for longer-term access to a virtual machine, with discounts of over 50% offered in return. Microsoft, Google and AWS offer spot instances, where the cost of the instance varies depending on apparent supply and demand as the providers look to sell unused capacity.

What's to stop someone from purchasing an inexpensive reserved instance over a period of three years, chopping it into hourly chunks, and selling those chunks to other users? If the price of an hour of that VM is cheaper than the on-demand price but greater than the average hourly price paid for the reserved instance, then the seller can make a profit with the buyer also getting a cheaper resource.

Similar models can be adopted with spot instances. The broker-dealer can sell access to cheaper VMs with the price set upon the risk of the spot pricing changed. During times when the spot price is below the sell price, the broker-dealer can collect a profit from the margin between the sell price and the cost price –this is essentially a financial hedge. When the price goes above the sell price, the broker can deliver the resources to users purchased on demand at a loss. As long as the sell price is cheaper than the on-demand price and the broker makes enough profit to cope with spikes in spot price, this can give both the broker and the buyer a gain. Reserved instances and spots are just two examples of potential broker-dealer models – there are many other models that could be explored and exploited in this market.

In theory, these methods should benefit cloud providers by syphoning risk away from them. Without broker-dealers, providers are fully responsible for planning capacity and balancing too much wasted capacity against too little such that demand can't be met. With the broker-dealer buying in advance, the provider has a better view of the future, and the broker-dealer assumes some of the risk. Although, of course, the provider might view the broker-dealer as sucking away on-demand revenue. Alternatively, automatically balancing workloads across combinations of spot, reserved and on-demand instances can minimize cost, while meeting requirements around availability and performance.

The market started with ClusterK, which was acquired by AWS in April 2015. Our interpretation at the time was that AWS was not keen on having a third party profiting from reselling its resources without adding any value. But since then, several similar companies have come to market focusing on this balance and optimization. For instance, cmpute.io (formerly known as Batchly) offers a SaaS execution engine for AWS that dynamically distributes resources across reserved, spot and on-demand instances. BidElastic has a similar remit, using a machine-learning plug-in to monitor AWS workloads and recommend a cost-efficient balance of instance types. Spotinst aims to offer a provider-agnostic endpoint for cloud-based applications employing load balancing, monitoring, an API gateway and a serverless engine to incorporate pricing intelligence and dynamics into workloads. 

Cloud waste managers: Potential targets

Because of the aforementioned complexity of cloud applications, there is a substantial risk of waste. Forgotten VMs left running, large expensive VMs being used where a cheaper one would suffice, or resources being left on during quiet periods when unlikely to be needed can all generate waste.

The need to 'trim the fat' started with AWS's Trusted Advisor, which provided recommendations on what resources were underutilized. Now automatic waste management is the name of the game, and several vendors have come to market promising substantial savings simply by cleaning up resources that aren't being fully utilized based on policy.

ParkMyCloud offers a bird's eye view of provisioned resources on AWS, Microsoft Azure and Google Cloud Platform and has a slick interface for 'parking' idle capacity, either according to a schedule or ad hoc. FittedCloud provides resource-level optimization and recommendations based on monitoring and usage-data analytics. GorillaStack offers proactive scheduling and automation of AWS compute and storage resources. Skeddly has adopted a pay-as-you-go approach to automating a broad variety of AWS services, and starts and stops AWS instances as and when required. RightScale's Optima is a stand-alone module of the company's larger cloud management platform that provides cost-saving recommendations to resource owners, enabling them to take automated actions and track optimization progress across an organization. DivvyCloud's cloud management platform provides a number of instance-optimization capabilities. British startup Nuvola Analytics promotes itself as a Cloud Financial Governance platform provider. Densify is a predictive analytics service that focuses on making the most of infrastructure.

Potential acquirers

Cost optimization is a key opportunity for service providers to add value and differentiate in a competitive market. The messaging of, 'Move to us, and we'll make sure you pay the least' is hugely compelling, and the fact that AWS and Azure offer adviser algorithms already demonstrates that there is an appetite to help customers make the best use of their cloud – automation of cost/usage is the next natural step. AWS, Google and Microsoft are potential acquirers.

A further (and perhaps greater) opportunity lies with MSPs looking to build and manage applications built on third-party clouds. The likes of Cloudreach, Rackspace, Datapipe and 2nd Watch add value to bare-bones infrastructure services by resolving issues of patching, monitoring, security and application management – why shouldn't they also be responsible for ensuring that users are paying the lowest price? Companies that aggregate cloud services, including Accenture, Vodafone, and others, could seek to add further value via M&A.

Finally, cloud management platforms could acquire cost-optimization tools to bolster their existing capabilities. Could Cloudability, the leading cloud cost management specialist today, expand its portfolio to be more focused on cost optimization? It has already purchased several companies in the sector, most recently CloudMGR in March. CloudHealth, OpsRamp, IBM, CA Technologies, BMC, Microsoft, HPE and many others play in cloud management.

New Alert Set

"My Alert"

Failed to Set Alert

"My Alert"