Optimizing Cloud Costs & Application Performance for a Leading Sports Betting Platform

Brand: Not Disclosed

Industry: Sport Betting

Technologies: Erlang, Cowboy, DynamoDB, Kafka, AWS, RESTful APIs, Docker

Partnership: 2022 - Present

Overview

Our client is a leading online gaming and sports betting platform offering an extensive range of betting options, including sports, casino games, live casino, virtual sports, and lotteries. Known for its user-friendly interface and innovative products, the platform prioritizes secure, high-performance experiences for both casual and professional players.

The client faced rising cloud costs, long startup times, and high CPU usage due to inefficient scheduler processes and heavy in-memory storage of customer limit data.

The Challenge

Key challenges included:

Excessive In-Memory Load: 32 GB RAM dedicated to storing customer limit data, driving up cloud costs.
Inefficient Scheduling: Each customer limit had a dedicated scheduler, generating up to 200,000 processes monthly and causing high CPU usage.
Prolonged Startup Time: Application startup took over 20 minutes, delaying availability and responsiveness.

The business goal was to reduce cloud costs, improve system efficiency, and accelerate application startup without compromising reliability.

The Solution

Our team implemented targeted optimizations to resolve the issues:

1. Optimized Data Storage with DynamoDB and Caching:

Replaced in-memory storage of customer limit data with Amazon DynamoDB, integrated with a caching layer using a prebuilt Erlang library for faster access and fewer direct database calls.

2. Streamlined Scheduler Logic:

Refactored schedulers to activate only for limits expiring within the next hour, cutting CPU load. AWS EventBridge Scheduler was evaluated, but a custom solution proved 10x more cost-effective.

3. Enhanced Application Startup:

Eliminated the need to preload customer limit data into memory, reducing startup time from 20 minutes to under 1 minute.

Tech Stack

Business Impact

The implemented solutions delivered measurable improvements:

Memory Usage: Reduced in-memory load from 32 GB to 1.6 GB (20x reduction)
Cloud Cost Optimization: Achieved a 30% reduction in AWS costs
Scheduler Efficiency: Reduced active schedulers from 200,000 to 80,000 per month (60% CPU load reduction)
Startup Time: Improved application startup from 20 minutes to under 1 minute

These optimizations resulted in a faster, more efficient platform, significantly reducing operating costs while improving system responsiveness and scalability.

More projects

OctoPlay

coming soon...

Athena

coming soon...

Backlight

coming soon...

Frequently Asked Questions

How were cloud costs reduced so significantly?

By replacing in-memory storage with DynamoDB and optimizing scheduler logic, we reduced memory consumption and CPU usage, lowering AWS costs by 30%.

How was scheduler efficiency optimized?

Schedulers now only activate for expiring limits within the next hour, cutting the number of active processes from 200,000 to 80,000 per month.

Is the system scalable for future growth?

Yes. DynamoDB, caching, and streamlined schedulers ensure the platform can handle more users and data efficiently.

Can these optimizations be applied to other applications?

Absolutely. Similar approaches to memory management, scheduler logic, and caching can reduce costs and improve performance in other distributed systems.