introductions

Cloud computing provides flexibility and scalability to enterprises, but as we all know, even with the best intentions and predictions, it can lead to unexpected cost surprises. Our experience with cloud cost management has uncovered a specific “sequence of operations” to follow in order to optimize costs and maximize savings. This post describes a step-by-step approach.

Before we begin, it's important to remember that there are no quick fixes to long-term cloud cost management, and that it's important to implement established processes and tools that can be reviewed regularly to build an efficient cloud environment.

Also, although the Pinops Foundation has a 3-stage functional lifecycle, our ordering or operating methodologies are a bit more prescriptive. This is especially true when talking about cloud computing.

‍

‍

Step 1: Appropriate sizing

The first step in the process is to properly size the cloud instance. This includes choosing the right instance type and size based on actual resource requirements. Many companies simply estimate when provisioning an instance, and often choose an unnecessarily larger size “for safety.” This can lead to significant waste and unnecessary costs.

For example, a customer we worked with was running a large high-memory instance on a web server even though actual memory usage was very low (and, importantly, stable). By right-sizing to a smaller instance size, we reduced costs by 30% while meeting resource requirements. This is a simple example, but most environments we encounter have similar opportunities.

Also, as part of this step, consider “modernization” and identify instances running on older or more expensive instance types. The instance must be updated to the latest available type prior to proper sizing. This is because the cost-performance ratio is likely to be the best. After evaluating the required performance, you can adjust the appropriate size.

Simply put, proper sizing (and modernization) provides the right amount of compute, memory, and storage for workloads.

‍

Step 2: Scheduling and automatic scaling

Once the instance is properly sized, the next step is to take advantage of scheduling and automatic scaling features. With scheduling, instances are automatically turned on or off at specific times based on usage patterns. With automatic scaling, the number of instances dynamically increases or decreases according to actual demand. This keeps the application responsive while minimising waste during low usage times.

Many industries have very specific operating hours and associated requirements for computing resources. One example is the financial market. User access does not necessarily end, but most systems only process data when the stock market is open. In certain environments, these instances can be easily terminated or scaled up, but other environments that use less modern architectures are difficult to adjust. By utilizing automatic scheduling, automatic scaling, or burstable instance types, you can intelligently reduce expenses when there is a predictable load. Given the stock market's fixed hours and number of recognized holidays, it's easy to see where you can get significant savings by simply scheduling.

By combining proper sizing, scheduling, and automatic scaling, resources are closely aligned with workload requirements.

‍

Step 3: Book and take advantage of savings plans

After establishing an appropriately sized and expanded environment, businesses can use reservation and savings plans to further reduce costs. This provides discounted rates in exchange for a commitment to consistent use for 1 to 3 years.

Compared to pay-as-you-go or on-demand pricing, the savings can be over 70%. The important thing is not to pay for unused capacity by reserving only what you need after proper sizing. As discussed in another post in this blog post; you need to maintain a process to make the most of these opportunities, and if you have a dynamic environment like most companies, you should use the following platforms: OpsNow mitigates risk and ensures maximum efficiency.

Automated computing commitment is OpsNow's specialty, and we reserve instances after proper sizing and scheduling to reduce enterprise costs by 40% or more while ensuring high utilization and coverage. We do this through tools.

The greatest value is that OpsNow provides improved savings without the end customer having to commit to a typical 1- or 3-year term. By eliminating this risk, action can be taken quickly to ensure maximum savings.

‍

Step 4: Monitor usage and costs

The final step is continuous monitoring of cloud resource usage and costs. Cloud platforms provide detailed metrics that can be analyzed to identify waste and optimization opportunities. Leveraging notifications for more than one based on tag groups or other well-defined processes is an important and often overlooked step. Environments without a structured tagging process and no loose or inconsistent process for terminating unused resources are often considered cloud costs, but managed properly, can save 15% or more.

It's also essential to set usage thresholds and alerts to be notified of unusual activity. For example, a customer found that heavy batch jobs were unnecessarily using the CPU at maximum for several hours. By setting up alerts, we were able to take proactive steps to resolve the issue, and as a result, we were able to save more than 20% of our monthly costs.

Regular monitoring ensures that the environment remains the right size and scale and expands over time while also securing additional savings opportunities. Don't confuse performance management and customer experience with the financial aspects of cost optimization. Although they often use the same tools, their goals are different.

  
‍

When to consider Spot Instances

After an appropriate sizing, automatic scaling, and reservation process, businesses can further optimize costs by utilizing Spot Instances. This provides unused computing capacity at a discount of up to 70% compared to On-Demand.

The trade-off is that the cloud provider can reclaim Spot instances if capacity is needed. As a result, Spot instances are ideal for fault-tolerant workloads such as batch processing jobs, development and test environments, big data analytics, and any application with flexible startup and shutdown times. If you're using a Kubernetes environment, you probably already have an extensive Spot implementation, so it's a natural fit, but if you have too many failed requests or a state storage environment, balancing on-demand might make sense. ‍

‍
‍

Methods for optimizing code

In addition to resource allocation and purchasing choices, another way to improve cloud efficiency is code optimization. Well-written code reduces demand on infrastructure while enabling instances to take full advantage of functionality. Although not within the scope of this post, this can have a significant impact on performance and compute utilization, so it should be the first part of an optimization task sequence.
‍

Common code optimizations include:

• Caching frequently accessed data
• Use more efficient algorithms and data structures
• Parallelizing computations between threads/processes
• Optimizing and tuning database queries
• Compress larger data sets
• Prevent inefficient resource usage patterns

Even small code improvements can lead to huge cost savings on a large scale. Reducing overall resource requirements by as little as 10% provides significant instance reduction opportunities. For example, a company reduced CPU utilization for each web request from 800 ms to 200 ms through code optimization. This enabled the server to handle 5 times more traffic without modifying the underlying VM.

Also, most of the cloud runs on NGINX. If your environment has NGINX or a similar load balancer, adjusting these tools can provide dramatic improvements that are an overlooked bottleneck between services and customers.

‍

summation

To manage cloud costs, you must follow a best practice sequence of tasks. We first adjust the instance size, then enable scheduling and automatic scaling, then reserve capacity, and finally implement processes to monitor costs and identify anomalies. Effective implementation and periodic review can significantly reduce the hassle of unexpected cloud spending. Using the following tools, OpsNow addresses some of the issues by highlighting savings opportunities and providing reliable analysis. We think this is an important part. But just like anything else, you must play an active role in operations.

‍