Tips for cloud performance optimization

At Capital One, we are constantly running at full speed when it comes to cloud initiatives, in fact, we were the first U.S. bank to move completely to the cloud in 2020. Our teams don’t see this momentum as merely a series of initiatives. They see their work as helping in the always-changing space of cloud technologies. With this space constantly evolving, there is so much to do, and yet such little time to do it; not to mention, resource challenges and competing priorities of working on the cloud at enterprise scale. With that in mind, there is one priority that cannot be ignored: cloud performance optimization.

The journey isn’t just about being in the cloud. It’s about optimizing in the cloud.

As the Director of Software Engineering at Capital One, I lead a team tasked with identiyfing rightsizing and cloud optimization opportunities within our organization. We know the goal of any cloud journey is not simply about being in the cloud. That's not good enough. It's about using the cloud to its optimal capacity while also galbanizing support amongst our internal users with attention paid to exemplary cloud native best practices.

Our team develops tools and strategies that help other Capital One teams improve performance in the cloud. We then reach out to teams and users to share our resources and tools in meetings and one-on-one talks. We have established some wonderful partnerships with various lines of 

businesses and helped train them on how to run their optimization tracks independently. In addition, we have worked with numerous application teams to deep dive into their specific applications; identifying opportunities and helping them streamline their deployments for improved performance and reduced costs.

How did my team and I get tasked to chase after these rightsizing opportunities by Capital One? First, we knew there were risks for not optimizing in the cloud. To ensure we’re all on the same page with optimizing in the cloud, we established best practices and shared standards to the table to ensure we’re all working on the same page. Second, and most importantly, we mapped out all the potential gains. Being in the cloud alone isn’t enough; that is why we turned to optimization.

Like all other companies that have lean approaches, we have resource challenges. These obstacles have not set us back, but instead made us more nimble, which was particularly helpful when working at full capacity during a global pandemic. In fact, thanks to our preemptive work navigating into and through the cloud during the pandemic, even with all of the health risks, social challenges and drastic changes in how almost all of us worked in tech, it posed almost no problems with our work moving forward. Despite resource challenges, we were well-prepared for something few could foresee.

Our first focus was on cloud cost optimization and efficiencies, which is a natural step once teams are dialed into resiliency, availability and improved performance. My team promoted automation through releasing several tools to help implement features such as predictive scaling and optimized close to 300 instances.

During this process, we learned that the biggest challenge was getting leadership buy-in! We noticed that when you influence the right leaders, the actions come automatically.

Another aspect is data. Using data to derive usage, costs and forecasting has been key in presenting the case for optimizing in the cloud. Our data has played a huge role in getting teams to buy in. In an ideal world, you would have such metrics available to each leader on a weekly basis so they could see the progress and opportunities within their scope of responsibility.

One team member, Data Analyst Igor Trubin, explains the mechanics behind data collection. “We found a way to collect adequate cloud usage performance data to automatically recognize the workload patterns of different cloud objects and their subsystems (clusters of servers, databases, containers, disk volumes and networks), including the cost of using them.” Some details of this method are shared in our blog post on optimizing your public cloud for maximum efficiency.

The work to improve cloud performance management and optimization never ends, and these other facets mentioned limited resources, costs & efficiencies and data add to the complexity of our journey. The iterative spirirt at Capital One compels me and my team to constantly recreate the spaces in which we find outselves virtually. While we know we’ve accomplished so much with cloud technologies, I don’t believe in resting on one’s laurels. So, we press on every day, discovering new ways to lead, innovate and inspire in the cloud space.