Meesho Migrates from Redis to Dragonfly: Cuts Latency by 50% and Costs by 60%
Learn how Meesho, a leading Indian e-commerce platform, migrated from Redis to Dragonfly, cutting latency by 50% and reducing infrastructure costs by 60%.
March 20, 2025

Meesho, a leading e-commerce platform in India's tier 2+ cities, operates with a strong focus on cost efficiency and performance optimization. For their business, low latency is critical, and maintaining an affordable price point is essential. As a cost-conscious platform, Meesho meticulously optimizes every bit of spend while continuously refining latency to deliver a tier-1 experience to tier-2+ users, even on flaky networks.
When you serve 175 million annual transacting users, every millisecond counts. To enhance performance and efficiency, Meesho recently migrated their high-throughput data workloads from Redis to Dragonfly. This transition resulted in a 50% reduction in latency and an impressive 60% cost savings.
In this case study, we’ll explore how Meesho addressed their caching challenges, why they chose Dragonfly, and the impact it has had on their infrastructure.
Meesho Infrastructure Overview
With ~175 million annual transacting users, Meesho operates on a massive scale. As a result, their infrastructure needs have evolved beyond their initial setup. Their technology foundation combines cloud-native solutions like Google Bigtable with self-hosted databases like MongoDB and ScyllaDB. Running primarily on Google Cloud Platform and managed through Terraform, Meesho’s application is mainly written in Java and Golang.
A critical component of this setup is a robust caching layer, which ensures low-latency data access across various high-traffic user workflows. Many services throughout the Meesho app must simultaneously query both a complex product catalog and detailed user information. This architectural setup makes caching not just beneficial but essential to scaling and maintaining a lightning-fast user experience, particularly for two specific service categories: e-commerce purchase flows and personalized product recommendations.
The Problem: Balancing Performance and Cost at Scale
As Meesho scaled operations to meet growing user demand, their caching needs evolved. The cost of running Redis grew unsustainably with scale, and higher-than-expected latency impacted their real-time user experience.
They required a solution that could handle millions of requests per second while maintaining data consistency and atomicity. Performance and cost efficiency were critical factors in ensuring a seamless real-time experience for users.
On the product side, optimizing caching infrastructure directly impacts business metrics. Faster load times would improve e-commerce conversion rates, driving revenue, while lower latency would enable more personalized product recommendations.
Meesho sought a caching solution that could provide strong performance at scale while optimizing costs for long-term sustainability.
Requirements for a New Solution
Meesho’s in-memory caching requirements were clear. They needed a solution with the ability to reliably handle millions of requests per second with data atomicity to ensure consistency and reliability for critical user workflows. They wanted a solution with strong community support for ongoing improvements and troubleshooting and one that scales efficiently without escalating costs.
Why Dragonfly?
After evaluating several caching systems, such as Valkey, Memcached, and Mcrouter Proxy, Meesho selected Dragonfly for its unique advantages. Dragonfly’s multi-threaded, shared-nothing architecture would allow Meesho to achieve significantly better throughput and latency than with Redis. Dragonfly would also deliver superior performance at a much lower cost, perfectly aligning with Meesho’s scaling requirements. Additionally, Dragonfly’s vibrant open-source community on GitHub and Discord would provide the support and collaboration Meesho was looking for.
Migration and Deployment
Meesho completed the migration from Redis to Dragonfly for some of the high-throughput use-case processes over the course of two months. This involved evaluating Dragonfly’s compatibility and performance against Valkey, Memcached, and Mcrouter Proxy. During the POC phase, Meesho encountered a few challenges, but close collaboration with the Dragonfly team proved invaluable.
Dragonfly’s team was highly responsive, providing early-access binaries with fixes that Meesho tested before they became generally available. This iterative approach helped resolve key issues quickly, ensuring a smooth transition. The team conducted performance testing and addressed challenges like memory spikes and data replication before the final migration of data over to Dragonfly.
Dragonfly’s Role in Meesho’s Ecosystem
Dragonfly is now one of the cornerstones of Meesho’s caching layer, powering several critical use cases. One of Dragonfly’s most vital uses is serving as the feature store for Meesho’s machine learning models, which power personalized user feeds. This use case involves hundreds of GBs of data and peaks at 8 million requests per second, making performance and reliability non-negotiable.
Dragonfly also excels at supporting high-traffic workflows such as initial app loading, product browsing, and transaction processing, ensuring low-latency responses for millions of users.
The Dragonfly Impact: Lower Latency and Lower Costs
The latency was reduced by roughly 40 to 50% and the cost reduction was around 60%.
— Shubham Sharma, Senior Software Architect at Meesho
Since migrating to Dragonfly, Meesho has realized significant benefits. By rigorously testing Dragonfly under real-world high-traffic conditions, Meesho helped identify and resolve critical issues related to large-scale caching.
The Dragonfly team was highly responsive, sharing early-access binaries for fixes, which Meesho validated before a broader release. This close collaboration improved Dragonfly’s stability and Meesho’s scalability, ensuring it could efficiently support multi-threaded workloads in its large-scale deployments.
“The latency was reduced by roughly 40 to 50% and the cost reduction was around 60%,” shared Meesho Senior Software Architect Shubham Sharma. Meesho also saw an improvement in throughput, enabling them to handle more concurrent requests while using fewer resources. These improvements underscore the importance of choosing the right caching solution for high-throughput workloads.
While maintaining compatibility with their existing infrastructure, Meesho’s critical workloads now handle 8 million requests per second with Dragonfly’s horizontally scalable solution. As Meesho continues to scale, Dragonfly has become essential to ensuring their platform remains fast, reliable, and cost-efficient.
Beyond the Migration: New Use Cases and Continued Optimization
With Dragonfly’s caching solution firmly in place, Meesho is well-positioned to continue scaling their platform while delivering a seamless, real-time experience to millions of users. The team is also exploring opportunities to collaborate with Dragonfly’s community and present their success story at future events.
Meesho will look to implement new use cases on Dragonfly as part of this process. “Two of the use cases are currently in production, and we have plans to roll it out wider over time,” said Meesho’s Sharma. The migration has reinforced Meesho’s commitment to adopting cutting-edge infrastructure solutions that balance performance with cost efficiency.
About Dragonfly
Dragonfly is a high-performance, multi-threaded caching solution designed for modern, high-throughput workloads. With its Redis-compatible API and shared-nothing architecture, Dragonfly delivers unparalleled performance and cost efficiency for enterprises like Meesho.
To try Dragonfly for free and experience performance improvements and cost savings of your own, sign up for Dragonfly Cloud.