Right choices, not just right-sizing: How to Really Optimize FinOps

Right choices, not just right-sizing: How to Really Optimize FinOps

May 12, 2024
Get tips and best practices from Develeap’s experts in your inbox

A kid walks to their parent, a computer engineer, “why does the sun rise in the east?” asks the kid. “Look outside. Is the sun up?” asks the parent. “Yes,” says the kid. The parent immediately replies, “So it works! Don’t try to fix it!” 

This joke is so old that it knows how to code in Assembly. If it works- don’t fix it.  But what if using the default, reliable, working network configuration can cost you 300% more than a bespoke configuration? 

Before I dive deep into how to utilize FinOps, I’d like to share with you a few data points we learned at Re:Invent 2023 that shocked me:

  • 94% of EC2 spend is over-provisioning.
  • 80% of S3 objects are only accessed once.
  • 80% of RDS instances are running on non-aurora DBs.

The cloud is projected to be a 1.3 trillion USD industry by 2027. Can we imagine the size of a 5% or 36% cost saving? Hardly- but we cannot afford not to try.

On the other hand, organizations that fully utilize AWS advantages, have the know-how and skill to implement best practices, and make the genuinely right choices can experience:

  • 42% cost reductions.
  • 330% increase of VMs per admin.
  •  69% reduction in time-to-resolution. 
  • 1410% increase in code deployment frequency. 

Applying FinOפs best practices will allow you to achieve such KPIs. Although it may seem complex, I will share cost management advice and examples across the three primary domains of AWS: computing, storage, and networking. After the three technical domains I will share what is the most important and urgent step you should take in your organization both to cut costs and to maintain your prevalence in the cutting edge of the industry.

Domain 1 Computing

The first rule of thumb should be to modernize.  When I claim each generation of EC2 type is better than the previous generation, my students ask if by better, I mean more efficient, more powerful, or better priced. The answer is simple: “Yes.” More often than not, we can see a price reduction and better performance for each generation. Remember, even if we ‘only’ see a performance improvement, we can use it to improve our workload and reduce our spending.

Right-typing. AWS offers more than 30 types of EC2s. it can be tempting to use general purpose types that can be “good enough” for most things or the so-called “free tier eligible” that will be charged at a certain point. It is true that “pound for pound,” most of them can cost more, but this is why we use the term “cost-effective” and not “cheap.” In order to maintain efficiency, you must examine each task and look into the option of a specific type performing it at the same level (or better) with fewer machines or in less time creating a “pay more to cost less” scenario. 

Right-sizing. Right-sizing can be interpreted in two ways: fleet and machine. Right-sizing your fleet is well-documented, well-practiced, and easily automated using Auto scaling groups.  We can look into the correct size, like finding the right workload type. More often than not- an XXL machine will be more cost-effective than running a number of Micro-size machines to produce the same power. However, we need to consider the reduced flexibility, the need for fault tolerance, or distributed workload. Out of all the decision-making, this one requires a great deal of interdisciplinary thinking.

ML assistance. AWS Compute Optimizer: AWS Compute Optimizer is a service that uses machine learning to analyze historical usage data and provide recommendations for optimizing AWS resources. It offers suggestions for EC2 instances, Auto Scaling groups, EBS volumes, and Lambda functions to improve performance and reduce costs. Analyzing metrics like CPU, memory, and I/O usage identifies over and under-provisioned resources and suggests optimal configurations. While it can lead to significant cost savings and performance enhancements, users should review its recommendations in the context of their specific applications and ensure they align with security and compliance needs.

Domain 2 Storage

Reevaluate your needs: We live in a time of big data, and we, sometimes justifiably, wish to retain every data point because it’s better to have it and not need it than need it and not have it. It’s hard to argue with this logic. I will need to prove that those data points are not helpful and will never be helpful. However, we do not operate in an internal system. The corporate budget is usually not as far-sighted as five years, primarily just to the end of the year. 

My advice is the C4:

  • Compliance: Do I have any legal or security mandate to hold it (such as tax files)?
  • Core requirement: Do I use it in our day-to-day operations?
  • Creating opportunities: Can I articulate a scenario where this data will be helpful in a reasonable amount of time to create opportunities for my operation?
  • Cost: can I store this data in a more cost-effective way?

As mentioned above, 80% of objects in S3 are only accessed once. By recognizing the pattern, we can eliminate idle resources, consolidate data points that are only relevant in context to a summary, and avoid inflation of our storage costs that are fueled only by our FOMO.

If you do not know, use your intelligence. 

AWS offers intelligent tiering solutions for two of its storage services: Amazon S3 and Amazon Elastic File System (EFS). These are designed to optimize costs by automatically moving data to the most cost-effective access tier based on how frequently the data is accessed. 

Amazon S3 Intelligent Tiering is a dynamic storage solution that optimizes costs by automatically moving data between two tiers based on access patterns. The Frequent Access Tier is used for regularly accessed data. In contrast, the Infrequent Access Tier holds data not accessed for 30 consecutive days, with no retrieval fees and automatic monitoring and movement. This system ensures cost-effectiveness without manual intervention or analysis of access patterns, making it ideal for data with unpredictable access frequencies.

Amazon EFS offers a lifecycle management feature that automatically moves files not accessed within a specified period (typically 30 days) to the cost-optimized Infrequent Access (IA) storage class. This tiering approach reduces costs while maintaining millisecond access latencies and transparent file accessibility. It’s designed for users who want to save on storage costs for infrequently accessed data without compromising performance or access speed. It provides an effortless solution to balance cost and access needs.

Domain 3: Networking

Since February 2023, AWS has charged for all public IPs, not only unattached EIPs. This is a very rare occasion of AWS increasing the price and not reducing it, from nothing to $3.5$ per month per IP. It may not seem much, but across thousands of machines in an enterprise environment, that is fertile ground for finding cost savings.

Implement AWS-managed services in front of your workload. Use serverless and manganese services such as API Gateway, Elastic Load Balancers, and NAT Gateway/instance to consolidate your access to public access points across several (or even hundreds of) services. This is safer, more manageable, and will now be even more cost-effective.

Monitor for idle recourses and remove them. Make sure to periodically scan, using AWS Flow Logs, for any underutilized ELBs, NAT/Transit/internet Gateways, and EIPs. Once you have findings, look for counter-indications to termination, such as a critical system backup, legal requirements, or false positives.

The shortest way can be the fastest AND cheapest way. Us VPC gateway endpoint for S3 and DymnoDB all the time. It has no cost and is easy to set up. Interface endpoints can save you some data transfer costs and offer better speed and security. Use Cloudfront and S3 as a backend to any static or almost static (such as simple Java scripts) websites. Fast, durable, and cost-effective.

Domain 0-training your organization.

As I mentioned at the beginning of the article, AWS can offer an extensive array of solutions tailored to meet businesses’ evolving needs. The key to unlocking its full potential lies not in the power of AWS  tools but within your team’s proficiency. Only with the right know-how on AWS services and best practices can they make smart decisions, make the most of resources, and run operations smoother than ever.

Equipped with comprehensive knowledge of AWS services and best practices, your workforce gains the insight needed to make well-informed decisions, optimize resource utilization, and enhance operational efficiency. Moreover, studies consistently showcase an ROI exceeding 200% for organizations that prioritize training, underscoring the profound impact trained teams can have on driving cost reductions and maximizing the value derived from AWS investments.

Beyond mere cost savings, AWS training fosters a culture of innovation and adaptability within your organization. Empowered with newfound skills, your team embarks on a journey of exploration, embracing advanced services and emerging technologies with confidence. This spirit of innovation not only accelerates your organization’s digital transformation journey but also cultivates a competitive edge in an ever-evolving market landscape. In essence, AWS training transcends the realm of education—it serves as a strategic enabler, propelling your organization toward sustained success and resilience in an increasingly digital world.

Optimizing AWS for Cost and Performance

Making the right choices is the key to optimizing recuses utilization and providing the most cost-effective solution. it is evident that navigating the complexities of AWS requires more than just right-sizing. Through the journey of this article, we have unveiled the crucial significance of making informed decisions and optimizing resource utilization across computing, storage, and networking domains. By embracing these principles, organizations can unlock substantial cost reductions and performance enhancements. Furthermore, let us not underestimate the transformative power of ongoing training and upskilling. By equipping your teams with the necessary knowledge and skills, you pave the way for innovation and resilience in the face of digital evolution. As we bid farewell, may these insights serve as guiding lights on your path to maximizing the value of AWS investments and driving sustained success in the ever-changing landscape of technology.

Become an AWS expert with Develeap

We’re Hiring!
Develeap is looking for talented DevOps engineers who want to make a difference in the world.