Rappi has optimized its benefit-cost ratio by 90% in the AWS cloud thanks to FinOps

Before the pandemic, Rappi had already seen significant growth in the number of users. However, when it set in, the demand for its services doubled, and even tripled at certain times. Rappy already had a team monitoring the services, but it wasn’t enough: the company required a cost-benefit analysis of the use of AWS technology and tools, to make it more efficient to support their businesses. This is how it created FinOps, a whole discipline and cultural practice for cloud financial management, and the benefits have been significant.

In 2019, the journey began towards greater certainty about the use and acquisition of infrastructure, towards knowing what worked and how much it cost the company, as well as the real benefits it provided. At first, Rappi operated with a reserve system, which really did not allow it to have resilience or elasticity at critical moments of day-to-day operations.

At times of high demand, as FinOps managers Lucas Zungri and Pablo Barrozo explain, they had very large peaks that they had to do away with, since the core principle of their business is the speed of deliveries, and the number of users grew during the pandemic.

Thus, the company looked for a system that would allow them to have scalability and high availability, and migrated their operation to Amazon EC2 Spot Instances. Today, almost 70% of Rappi's workload uses spot instances.

Spot instances allow it to take advantage of unused Amazon EC2 capacity in the AWS Cloud. Spot instances are available at a discount of up to 90% compared to on-demand pricing.

The company specifically used the Amazon EC2 Spot Instances to put together a cost optimization case, given the finance department’s need to understand the cost benefits of having or rejecting certain technological tools. Rappi knew it needed the technology, but it also needed to be certain that what was being used was what was really required, and that it was not spending unnecessarily or excessively. It was necessary to form a team for the purposes of optimizing the infrastructure but also giving it visibility, so that the various teams that make up the company could make the best business decisions.

So, how do you ensure that the team understands the investment made in infrastructure, and that it can be controlled and visible in order to perform said cost/benefit measurement? Two important premises emerged: decisions regarding the acquisition of infrastructure had to be made based on the efficiency, resilience and traceability of the technology. Having the ideal architecture to be efficient in the cloud, which did not necessarily mean the largest.

The second premise was to measure and monitor, but with the firm intent of educating all the teams about the tools Rappi use or would use. The monitoring allowed it to understand, for example, that there were technologies that were over provisioned, and others that were very old. By changing this, the efficiency of the platform increased, and the cost of the tool became more efficient.

Thus, the previous infrastructure was resized and the elasticity and recovery capacity in the face of demand increased, even in the most violent peaks during the day: in a peak hour more than 10,000 instances can be run. Rappi start by combining Amazon Elastic Kubernetes Service (Amazon EKS), Amazon Elastic Container Service (Amazon ECS), Amazon Athena, AWS Lambda and AWS Fargate for each situation that arises, and are thus able to be more resilient and scalable.

To manage costs, Rappi uses AWS Cost Explorer, AWS Cost & Usage Report and Amazon QuickSight. In this way, it has access to panels and analysis, adapted to the required use, and recommendations for each service minute by minute. In addition to detecting anomalies, the system is economically flexible.

“One of the most important steps in shaping FinOps was traceability and accountability: we had to let the teams know what they needed and how much it cost, as well as how effective it was. A cultural change had to be made, where everyone took responsibility for the use of their infrastructure and the costs. By which I mean, achieving a better cost-efficiency ratio for the cloud”, explains Lucas Zungri.

Rappi currently processes more than 25 million orders per month (across the nine countries where Rappi is present).

FinOps, the managers explain, serves as a guide for the entire company. It allows organizations to get the most value from their investments, collaborating with engineering, finance and business teams, using data on the cost impact of each decision that is made, regarding the use of a certain infrastructure or tool. Thanks to this development, today all the teams in the company can understand the use and cost of the cloud.

In addition, the performance of the technology is monitored in real time, as well as detecting any possible failures, to make the processes more efficient. Having up-to-date monitoring allows decisions to be made in the moment to optimize models (according to the price the company will pay) and the use made of the infrastructure, allowing the organization to be fully aligned. AWS Services guarantee that flexibility for Rappi. Additionally, using Amazon EC2 Spot Instances leads to significant cost optimization.

With Amazon EC2 Reserved Instances (RI) and Savings Plans for Amazon EC2, Rappi was already saving more than 40% compared to on-demand costs. Now, using spot instances, it has achieved extra savings of approximately 10%.

“Today with Units Economics, we can know the cost efficiency that each AWS tool offers us, and justify its use. All the decisions about what we do or do not use are made based on the speed, quality and efficiency that they represent, and that is something that the entire organization is already doing”, clarifies Pablo Barrozo.

As progress is made on the issue of good practices within the company, and thanks to the implementation of FinOps, there is a trend towards bringing all Rappi services on to AWS. The idea is to continue looking for opportunities for improvement in all areas, including and prioritizing optimal financial management of the infrastructure.

In the future, the use of artificial intelligence is expected to help detect anomalous behaviors or long-term trends, such as a possible increase in costs, or even tools that automate low usage. The use of Amazon EC2 Spot Instances is a cost optimization strategy, and it is planned to continue using it, and improving it, as part of a continuous process.