Moving to the cloud is supposed to save money. That strategy is often successful, but sometimes related expenses can start climbing. Some areas to keep a weather eye on can help mitigate losses.
Many enterprises continue to move to the cloud as a way of managing their computing expenses. Added security, reduction in IT-related expenses of managing systems in-house, and user convenience often top the list of corporate motivations. Unfortunately, a shifting of C-suite focus to other areas of business after migration can result in a slow escalation of costs associated with cloud, a trend that can eventually make that migration look like a not-so-great strategy. In fact, a 2020 Gartner report estimated that as much as 70 percent of public cloud expenditures are wasted money, largely because of inefficient cloud-management practices. With most cloud users employing either public or hybrid cloud arrangements, suboptimal public cloud use is a problem that must be taken seriously. Savings don’t automatically follow like pennies falling from heaven; they have to be cultivated.
Every Time It Rains, It Rains
A number of factors contribute to the problem. A first among equals can be the conflicts between the difference in mindsets between capital expenditures (CapEx) and operational expenditures (OpEx).
A standard paradigm for decades has been to consider computing resources as CapEx. Enterprises bought hardware such as servers, storage, and network devices, as well as software licenses and the people to run them (via their salaries) and strove to squeeze as much processing power as possible out of all that within the infrastructure’s lifetime. Moving operations to the cloud, even though some of that infrastructure remains, means migrating computer resources into the realm of OpEx. Those resources become an operational expense because they are now primarily a service being paid for rather than an investment in something physical that the enterprise “owns.”
This subtle change has contributed to many enterprises miscalculating the total cost of ownership prior to implementation, losing focus on expenses as cloud charges move away from IT budgets and are charged to departments using the resources. Enterprises often also make poor selections on service options available from their CSP.
One example of the results are problems such as overprovisioning, which is assigning more resources than are needed for a particular application or service out of concern there will be overloads or bottlenecking if too many users try to access the app or service concurrently. Under most CSP service plans, that resource use is billed even if it’s not actually being used. Another challenge is the example of temporarily enabling a particular service or cloud instance for a specific function, such as changing some application code or allocating some temporary storage, but then forgetting to disable that setting and having the charges accrue for days or longer even though that aspect is no longer in use. As a final example, the on-demand nature of cloud use and the frequent use of multiple CSPs for different aspects of cloud services also spread responsibility for efficient resource use among different enterprise groups and encourage the rationalization that “someone else” is responsible for overall monitoring. When everyone is in charge, ultimately often no one is.
You’ll Find Your Fortune Falling All Over Town
What can be missing is the realization that managing cloud-based computing expenses is not a task that should be left solely to an enterprise’s CSP or to a single department such as IT or finance. Conversion is a cultural change as well as a technical one and requires attention be given to a new set of concerns. Once an enterprise is in the cloud, it needs a new focus on optimizing and controlling those costs. There are numerous ways to deal with this shift in emphasis.
A starting point is to address this shift head-on by changing the awareness of good corporate citizens to be more cost-conscious. All employees need to be made aware of, and made ready to report on, resource misuses and performance problems.
One good idea is a multidepartmental delivery team or other core cloud team that is specifically assigned to keep track of enterprisewide cloud usage and costs. There needs to be clarity in CSP charges and conditions under which all service expenses accrue, so the team should study CSP billing statements closely enough to know where each charge originates. The team should either be empowered, or report to someone who is, to carry out tasks such as making decisions about rechanneling resources, reducing excess storage, eliminating unnecessary users, scaling cloud resources to pair up with actual demand, and consolidating databases, as well as providing performance analysis and continuous cost monitoring in real time.
The team must be aware of special discounts offered by their CSP(s) and be prepared to use them when appropriate. Three prominent examples are spot instances, reserved instances, and tagging. The first two are alternatives to on-demand use of cloud resources, which is when the enterprise temporarily needs more computing power, storage, or other services. Spot instances are spare computing time offered on an auction basis by major CSPs. Users bid for these instances, but they can be lost on a few minutes’ notice when a particular user is outbid, so effective use relies on keeping track of statistics such as recent bid prices, which the CSPs offering that service provide. Reserved instances are usually offered in conjunction with committing to a multiyear service plan or a large amount of potential reserved time and are essentially on-demand time at a lower price (as much as 80 percent) and still available only as needed. Their downside is that using them is a further step toward vendor lock-in. Tagging is attaching an electronic tag to a resource that can add data about how and what that resource is being used for and to associate the resource with some entity. Principally used by Oracle, it can identify, for example, what department is using a cloud resource for what reason and can greatly simplify cloud cost accounting. Again, though, use commits at least some cloud functions to that single vendor.
Every CSP also offers additional services, such as cloud spending management apps, automation of various functions, containers, and other management tools. Whatever entity is placed in charge of cloud oversight needs to be able to deploy these additional services to best advantage, even though they increase dependency on the vendor offering them, because they’re currently available and part of the CSP’s fees already. For example, automation controls from some CSPs can be set to automatically turn off resources after peak hours if they’re not being used, rather than rely on someone to remember to do it manually.
It's also important to go through the exercise of mapping out cloud processes and roles and updating computing resource architectural diagrams, as well as updating them when conditions change. Useful for troubleshooting and accounting, particularly in situations of frequent staff turnover, they can help spot spikes in usage and other potential areas for attention. The mapping will also help identify resources that are provisioned incorrectly and help determine appropriate levels to reset them to.
Shadow IT is a greater pitfall than usual in a cloud environment. Unauthorized software and equipment modifications can have unpredictable effects on cloud app access and performance. A good suggestion for combatting this is to create a service catalog with an auditing process that lists all the cloud services available to users and customers, accompanied by a rule requiring new service requests to go through a catalog-related ordering process.
Trade Them for a Package of Sunshine and Flowers
The three largest CSPs (Amazon Web Services, Microsoft Azure, and Google Cloud) offer built-in cloud expense management and cloud cost optimization solutions that can analyze their statements to help users compare usages, costs, and charges across vendors as well as break charges down by internal user groups. Some of the data appearing in those tools can appear to be cryptic, so in addition, numerous independent software vendors offer applications that help users understand CSP service statements and drill down for usage information. Exact features vary, but using one could be a justifiable additional expense to help break down cloud use and financial accountability. Vendors offering solutions of this kind include Asignet, Bindadox, Brightfin, Etma, Nutanex, Profit Enhancement Systems, Quantconnects, Tangoe, Uptrends, Valicom, Virtana, and VMware.
Today’s tools can not only give enterprises insights into the effectiveness of efforts to avoid poor use of cloud resources, but also help discover architectural optimizations that can make cloud use even more efficient. One example of this kind of benefit occurs when an application that’s been migrated from an in-house server to the cloud is rearchitected to take advantage of virtualization. Rather than tying up a single physical server, the restructured app can coexist on a single virtual server with other apps, making more efficient use of resources.
Although migrating to the cloud provides a faster rollout of apps, simply moving applications structured for server-based architectures to the cloud isn’t the best strategy. The apps need to be replaced over time by new versions that are “cloud-native,” which is to say modularized for cloud delivery models such as Software as a Service, Network as a Service, Platform as a Service, etc. While it’s beyond the scope of this article to delve into that process, apps versions in the cloud that were originally built for older architectures often display poorer performance if they’re not optimized for cloud environments, either before migration or relatively soon thereafter. Postponing this optimization risks newer priorities crowding out such modifications once a cloud migration is initially accomplished.
Use DevOps processes to better deploy production and testing environments within available cloud spaces. Replacing larger apps with microservice structures promotes more-efficient use of cloud computing cycles. Use of containers to modify existing apps is useful because the microservices the containers use are not retained in virtual memory when they’re not active.
Pennies Become Dollars
While this is not an exhaustive list of the ways in which small economies that add up can be found to reduce cloud environment expenses, it does provide an overview of many avenues for exploration. It’s important to be aware that operating in the cloud puts an even higher premium on following management practices that, while they’ve always made good business sense, are even more important to maintain to keep from undermining a cloud migration strategy. Once an enterprise is in the cloud, the devil is in the details to a greater extent than ever before.