Cloud waste occurs when cloud resources are unutilized or underutilized. Resource under-utilization occurs when more resources are procured than are actually needed by virtual machines (VMs) at runtime. Cloud providers continue to charge for these provisioned resources regardless of whether they are used or not, resulting in unchecked expenditure.
FinOps teams and other stakeholders can measure cloud waste by calculating the difference—in terms of resources available—between the maximum computing load and the actual computing load.
Causes of cloud waste
If cloud waste persists, costs and revenue will remain unoptimized. To put a stop to excessive cloud expenditures and increase ROI, organizations must first understand the reasons for and sources of cloud waste.
Idle runtime
Idle runtime refers to the amount of time that resources spend waiting to be used. Idle runtime can be:
● Cloud resources sitting idle while an app is being executed
● Provisioned resources waiting for potential use by new machines
● Resources procured for longer than needed
● Orphaned resources not yet deleted
Idle runtime can occur due to unattended VMs, containers, and databases that continue to rack up network and CPU/hr charges, or there may be abandoned volumes that remain after pods or servers have been shut down that nonetheless continue incurring storage costs.
Flawed design
Engineers may unknowingly build apps or updates with suboptimal code, design, features, or third-party software or machines. When this happens, the flawed resources may consume excessive amounts of resources.
Consider a design that involves excessive API calls to cloud services, or one that requires unnecessary data transfers between VMs with inordinate frequency; both of these scenarios result in unwarranted charges that add up.
If not discovered and resolved promptly, these flaws may end up draining money to the tune of thousands of dollars.
Oversized resources
When engineers over-provision instances or deploy oversized infrastructure to handle potential traffic, usage, or storage spikes, it leads to oversized resources.
To avoid future complications and latency issues that result from undersized resources, developers often overestimate the resources needed, preferring to simply provision for the maximum rather than for the actual computing load.
This usually results in enterprises continuously paying for idle resources that may never be needed (or used).
Test waste
DevOps teams use testing environments to check for and fix bugs before applications or updates are launched. However, resources used in testing environments are rarely deleted or suspended because engineering teams want to keep them available as reference points or for future reuse. This leaves enterprises paying for idle resources.
Excessive centralization
Centralization entails storing all data in one data center regardless of usage duration and geographical location. In the early days of cloud adoption, centralized storage was the holy grail. It was easier to use, it facilitated data consistency, and it streamlined governance. However, as cloud storage evolved, with data volumes soaring and costs escalating exponentially, companies were forced to reassess centralized storage as a practice.
An important con of centralization is cloud waste. The total spend on storage, bandwidth, and multi-region redundancy can spiral out of control. Particularly, bill shock has become a regular occurrence due to unpredictable cross-regional data egress costs.
This is where decentralized cloud storage has a comparative advantage, as data storage and egress provisioned at the edge are usually more cost-efficient.
Additionally, in centralized storage, overprovisioning is a common occurrence because of the need to keep distant databases functioning at top velocity and prevent traffic or data volumes from overwhelming systems.
Excessive data storage
Data stored in the cloud is accessed with varying levels of frequency, from never at all to very often. Holding on to never-accessed data or using high-performance storage for rarely accessed data is a waste of valuable money and storage space.
Challenges in tackling cloud waste
Now that the sources of cloud waste have been identified, let’s discuss why cloud waste is so difficult to control.
Rightsizing
A foundational approach to cutting cloud waste is to rightsize. Rightsizing should be done right from the production environment through every stage of the software development life cycle to balance between optimizing spend and boosting performance. However, resource requirements depend on changing end-user traffic, and teams rely on estimates that can suddenly become too big or too small; this requires enterprises to continuously autoscale their resources according to use.
Autoscaling seems to be the perfect solution, and horizontal autoscaling has become a favorite with engineers. However, once in use, scaling apps vertically becomes fraught with risks of hardware failures, data loss, downtime, scalability limitations, and installation issues.
Unchecked costs
Engineering teams want to focus on delivering the best-performing applications to drive customer patronage, conversion, and satisfaction. As such, they may not monitor resource usage and costs in minute detail nor keep an eye on metrics such as cost per cluster, resource, or environment. This means they may not detect cost escalations nor be aware of their root causes and solutions. Without collaboration between engineering and finance teams, enterprises may end up paying far more than needed.
Another reason for unchecked costs is enterprises being unsure of ongoing and future resource needs. For instance, because engineering teams are unsure of how often a block of data may be accessed, they place it in hot storage. But then, the data is rarely accessed, and the enterprise consistently pays for hot storage when cold storage would have been more appropriate and saved the company a tremendous amount.
Lack of awareness
Organizations may unwittingly subscribe to more expensive plans simply because they are unaware of discounts, or they may be ignorant of the comparative cost benefits of spot or on-demand instances versus reserved instances.
For example, for a relatively predictable workload, companies can cut storage costs by up to 70% by simply committing to long-term storage using reserved instances.
Reducing cloud waste: 6 best practice solutions
In most cases, cloud waste comes in the form of a few cents or dollars here and there, which when added up across resources and over time become thousands of dollars. Below, we explore six ways to eliminate cloud waste.
1. Adopt an efficient cloud cost management solution
Stakeholders must monitor and audit cloud resources regularly to prevent cloud waste. However, proper monitoring, governance, and auditing require full visibility into your cloud environment.
To achieve this, you will need cloud cost management tools that allow you to splice data for granular visibility, empowering you with insights on oversized or underutilized resources, engineering flaws, and other issues that may be contributing to cloud waste. All of this will drive decisions on where to invest less or more, as well as enable engineering teams to address design flaws and other engineering issues promptly.
Additionally, your cloud cost management tool should have cost reporting capabilities, automatically fix cost anomalies, and provide detailed graphs of resources incurring the highest cost for easy discovery of potential or ongoing cloud waste.
Consider a solution that lets you perform ad hoc spend analyses across preferred dates, regions, accounts, resources, or other parameters. This should include various grouping, filtering, and visualization mechanisms.
2. Implement FinOps
Foster a culture of cost optimization and accountability. This entails DevOps teams making data-driven decisions and provisioning resources with both cost and performance implications in mind.
Encourage cross-team collaboration as well, with engineering and finance teams spearheading the anti-waste campaign.
3. Park idle resources
Provisioned resources are charged but rarely used 24/7. Because of this, companies should park, or shut down, resources in testing, production, staging, and other environments when not in use; these can then be restarted when needed. While EC2 and RDS instances are the most commonly parked resources, ECS, EKS, and many other cloud resources can also be parked to cut waste.
However, engineering teams may have problems automating cloud resource parking because techniques to do so vary across services and instances. The exception is for EC2 and RDS instances, which can be easily automated using Lambda functions.
Still, leaving resources unparked can quickly accrue waste to the tune of tens of thousands of dollars. Engineers can thus optimize for cost, time, and effort by focusing mostly on parking non-production or infrequently used resources.
4. Classify and optimize data
Consider your organization’s data usage patterns and store data accordingly. Aside from deleting unutilized data, this would entail selectively storing data in:
● High-performance storage: For frequently accessed data, higher cost
● Warm storage: For data you access less frequently, less expensive
● Cold storage: For rarely accessed data, low cost
To optimize data storage spend, make sure to automate your data storage policies so that data is moved as needed to the appropriate tier, according to usage frequency. Doing this will reduce storage data costs to the minimum and save your engineering teams the hassle of manually moving data around.
5. Automate rightsizing and cloud cost management
Rather than over-provisioning, which results in waste, engineering teams can take advantage of autoscaling. However, it’s critical to control autoscaling limits on the console provided by your cloud service provider (CSP), as if not properly managed, it can quickly spiral into cloud waste.
Additionally, autoscaled instances are built to automatically balance across your availability zones, and some instances may not auto-terminate after completing the required task. An Amazon EC2 instance, which is billed per second, is an example of this. Depending on the traffic and number or capacity of instances autoscaled, you may be dealing with thousands of wasted spend.
Enterprises must thus invest in an efficient cost management tool to proactively detect anomalies, notify them when they’re approaching forecasted spend, and stop autoscaling when their budget is surpassed.
6. Leverage CSP cost savings plans
To stay on top of hidden cost-saving options and the availability of spot instances, organizations need to study and query their cloud provider’s plans and pricing tiers.
Bid for spot instances during low-demand seasons; use them for tasks such as batching and analytics at the cheapest prices possible, and release them once the task is completed.
Also make sure to minimize your use of on-demand instances, as they can be quite expensive.
Lastly, leverage savings plans and reserved instances where the commitment benefits align with your usage volumes and patterns.
Stop cloud waste with ManageEngine CloudSpend
Cloud waste diminishes profitability, stalls growth, and may end up impairing application performance when engineers can’t provision needed resources due to previous or ongoing waste. The keys to eliminating cloud waste are to understand its sources and implement the solutions discussed above. Still, without the right cloud cost management solution, this can prove impossible.
ManageEngine CloudSpend allows you to guardrail cloud costs, detect and fix cost anomalies in real time, and intelligently pinpoint the resources driving cloud waste. Additionally, it will point out cloud spend optimization features to improve application performance. Sign up for a free trial today or request a demo, and see how you can stop cloud wastage in its tracks.
