Case Study: Zedia Powers High-Performance AdTech with Dragonfly for Cost Efficiency

Introduction

In the fast-paced world of ad tech, performance and cost efficiency are critical. For Zedia, a Brazilian startup delivering targeted ads via TV signals, managing high traffic loads while keeping cloud costs low was a major challenge. After struggling with Redis’s performance limitations, they turned to Dragonfly, a modern, high-performance Redis-compatible in-memory data store, to handle their real-time ad delivery system.

This case study explores how Dragonfly helped them:

• Reduce infrastructure costs by efficiently handling high connection volumes.
• Maintain reliability despite running on spot instances.
• Simplify architecture with Redis compatibility.

The Challenge: Scaling Ad Delivery Without Breaking the Bank

Zedia’s platform serves ads to over 150,000+ concurrent users during peak events, with each user device querying their servers every 15 seconds. Their setup relies on application-level caching to reduce database load. Previously, they also used Redis for storing active ads, but it couldn’t handle the surge in connections and requests, leading to crashes. Spot instances were also used for the application and Redis to cut costs. The primary pain points of the previous Redis setup are:

• Redis Limitations: At scale, the application couldn’t maintain enough connections and requests to fetch ads in real time.
• Operational & Cost Overhead: Running multiple Redis instances created unnecessary complexity, wasting server resources and requiring engineering forces for maintenance rather than innovation.
• High Cloud Bills: With local currency devalued against the US dollar, cloud expenses were a major burden.

Why Dragonfly?

After evaluating alternatives like KeyDB (which had stalled in development), they chose Dragonfly for:

• High Performance: Handles millions of requests per second with ease.
• Redis Compatibility: Zero code changes were required, so existing Redis-based workflows worked immediately.
• Cost Efficiency: Eliminates the need for horizontal scaling (at least for Zedia’s current workload), saving on AWS bills.
• Active Community: As a community user, Zedia is able to seek help and information via the community channels (Discord, forum, docs, etc.) easily.

"Dragonfly’s active community and Kubernetes support made the setup effortless. We didn’t even need to modify our application."
— Nicolas Vieira Pires & Vicente Menezes Pereira, Backend Team @Zedia

Migration to Dragonfly & Overcoming Spot Instance Challenges

Transitioning to Dragonfly was seamless for Zedia, thanks to Dragonfly’s high Redis compatibility. Since Zedia’s ad data was temporary by nature, they could adopt Dragonfly without application changes with an easy switchover. They initially deployed Dragonfly on AWS using a 1-primary, 2-replica setup, managed by the Dragonfly Kubernetes Operator running on spot instances.

The Spot Instance Dilemma

While Dragonfly easily managed their traffic peaks, their reliance on AWS spot instances created operational uncertainties. The most significant challenge emerged when all three spot instances running Dragonfly were terminated simultaneously, which is an infrequent but impactful event. During these occurrences, the behavior of attached EBS volumes became unpredictable. While the temporary nature of their ad data meant no permanent information was lost, the engineering team still needed to prioritize quick recovery for uninterrupted operations. Each incident required manual intervention to reactivate campaigns, creating unwanted operational overhead and delaying their ad delivery pipeline.

Optimizing for Resilience and Cost

With guidance from the Dragonfly team, they implemented two key improvements:

• Cloud Storage Based Snapshots: Instead of relying solely on EBS volumes, they configured automated snapshots to cloud storage (S3), ensuring reliable recovery even after spot failures.
• Right-Sizing the Deployment: Monitoring revealed that a single Dragonfly instance (with proper snapshot paces) could handle their workload. Given their tolerance for potential but minimal data loss between snapshots, they eliminated the replicas, further reducing costs without sacrificing performance.

Although note that this is not a universal solution. While it works for Zedia’s requirements of data resilience and cost efficiency, other high availability and backup options should be considered for different use cases.

"Dragonfly’s fast snapshotting and the ability to utilize cloud storage turned a major pain point into a non-issue. We now have peace of mind, even with spot instances."
— Nicolas Vieira Pires & Vicente Menezes Pereira, Backend Team @Zedia

Key Learnings and the Path Forward

The team’s experience with Dragonfly on spot instances yielded valuable insights. While the setup worked reliably most of the time, they discovered that the occasional simultaneous termination of all spot instances can cause recovery problems requiring manual intervention due to unreliable disk storage. This issue revealed the importance of robust backup strategies, prompting them to implement cloud storage-based snapshots for reliable recovery of their temporary but operationally critical data.

Looking ahead, Zedia plans to optimize further by exploring other platforms such as GCP for more favorable startup pricing while strategically transitioning core workloads to on-demand instances as their business scales. Through this evolution, Dragonfly has proven to be the ideal foundation, delivering the perfect balance of Redis compatibility, performance, and cost efficiency for their high-volume ad platform. Dragonfly’s performance, flexibility, and cost efficiency continue to support their infrastructure as they grow, proving that even data systems storing and serving short-lived data deserve enterprise-grade reliability.

About Dragonfly

Dragonfly is a high-performance, multi-threaded, in-memory data store designed for modern demanding workloads. For businesses processing high-velocity data with strict latency requirements, Dragonfly proves to be a game-changing alternative to Redis.

To try Dragonfly for free and experience performance improvements and cost savings of your own, sign up for Dragonfly Cloud today!