MPL Gaming Platform Goes from Zero to 40 Million Users on AWS

With the proliferation of smartphones and affordable mobile data plans in India, mobile gaming is quickly becoming a common pastime. According to Kantar IMRB, India currently ranks fifth among the world’s top mobile gaming markets. Bangalore-based Mobile Premier League (MPL) is one of the biggest and fastest-growing players, offering more than 40 games via its eSports platform. All games, including fantasy sports or the country’s favorite, Rummy, can be played for cash prizes. The MPL mobile app launched in September 2018 and acquired 10 million users within three months, which met the company’s one-year subscriber goal.

Today, MPL has more than 40 million subscribers, making up 14 percent of the total mobile gaming market in India, according to the 2019 report “The Power of Mobile Gaming in India.”

The startup launched on the Amazon Web Services (AWS) Cloud because many of its DevOps engineers had experience on the platform, which accelerated time-to-market. Scalability and automation were also a priority for MPL’s AWS Cloud infrastructure. The startup began with Amazon Aurora as its primary database, using Amazon Relational Database Service (Amazon RDS) to automate administrative tasks such as provisioning and backups. However, as its dataset grew—particularly its volume of unstructured data—MPL found that Amazon DynamoDB NoSQL database service worked better for gaming use cases because it offered low-latency data access and easy horizontal scaling.

Amazon DynamoDB can efficiently handle volume, velocity, and veracity for the data-heavy workloads typical of gaming companies. Additionally, the database automatically scales capacity to maintain performance during peak periods such as nationwide sporting events, when online traffic for MPL’s fantasy games can spike to 2.5 million hits per minute. Such automation is key in MPL’s architecture and resource planning.

Despite being a year and a half into operations, MPL’s DevOps engineering team have never had to expand since launch. “Our team of 12 engineers manages DevOps, reliability engineering, and 24/7 monitoring,” says Mukta Aphale, vice president of Reliability Engineering at MPL. “We don’t need a big team to run all our applications on AWS.” Its DevOps engineers currently use a microservices architecture for development, automating the deployment of more than 50 separate microservices using AWS CodeDeploy and AWS Lambda.

With such rapid growth, MPL has relied on support from AWS for debugging and periodic re-architecting to scale. “Nailing down the problem is often the toughest part of debugging, and AWS was always there to help us,” Aphale says. The startup recently subscribed to AWS Enterprise Support to prepare for the launch of a big campaign. Teams appreciate having an AWS technical account manager (TAM) and AWS Infrastructure Event Management (IEM) as part of the support package, with focused planning and ready assistance leading up to and during the launch.

“Having a dedicated TAM and IEM has led to a more detailed review of our architecture and guidance on how to optimize both the user experience and operating costs,” Aphale says. Early on, AWS hosted an AWS Dev Day event, where MPL teams used Amazon Inspector to uncover potential security vulnerabilities in their infrastructure. These vulnerabilities have since been resolved, and MPL has implemented Amazon GuardDuty for ongoing security monitoring.

AWS solutions architects held several discussions with MPL and shared the benefits of containerization. Following that, MPL conducted its own evaluation and moved ahead with containerizing its microservices using Kubernetes. The aim of this is twofold: to stabilize application performance at scale and improve operational efficiency. Its engineers use Amazon Elastic Kubernetes Service (Amazon EKS) to run the containers with high availability. “We appreciate being able to control our architecture and decide where and how to automate things. Amazon EKS is much more robust than open source Kubernetes,” Aphale explains.

Implementation of containers is just the first step in MPL’s cost optimization journey. During its first year of operations, the company was focused on scaling and right-sizing its architecture. Now, teams are working on optimizing infrastructure costs and looking to innovate further. Aphale says, “We’re pleasantly surprised with the proactive help from AWS to improve our architecture and save costs.”

Working with their assigned TAM on finding the right instance types for production versus non-production workloads has gone a long way toward reducing costs. In addition, MPL signed up for the AWS Enterprise Discount Program. Since beginning cost optimization exercises, MPL has cut infrastructure costs by 40 percent. “These savings have allowed us to scale our AWS infrastructure to roll out a lot more features and games, which creates a more interesting user experience with features such as audio chat,” Aphale says.

A large portion of MPL’s new customer base comes from referrals, so the business focuses heavily on customer satisfaction. For instance, because sub-millisecond latency is critical in gaming to ensure players don’t experience a lag between moves, MPL has been able to maintain an API response time of 30 milliseconds or less since its launch.

In working with AWS to fine-tune its architecture, engineers have also improved the app’s service availability. “Our app uptime has increased significantly to 99.9 percent,” Aphale says. Due to MPL’s confidence in the underlying AWS architecture, the DevOps team is more geared up to take on new challenges. Taking advantage of managed services from AWS and increased automation allow the team to take on development tasks they find interesting and to learn about technology applications such as containers firsthand.

In its road to expansion, MPL launched its gaming platform in Indonesia in July 2019 and is awaiting the launch of AWS data centers in the country by 2022. Aphale concludes, “AWS has been by our side throughout our entire growth journey, from debugging to stabilizing and optimizing to now expanding our product.”