How Zepto Achieved Sub-Second Search and Cost Savings with OpenSearch Service OR2
Zepto, an Indian quick-commerce platform, migrated its Amazon OpenSearch Service data nodes to OpenSearch Optimized (OR2) instances to address scaling challenges. This change enabled sub-second product search latency and significantly improved indexing throughput by over 100%, alongside achieving 30% cost savings. The OR2 instances, designed for high indexing throughput and cost efficiency using local storage with S3 durability, allowed Zepto to manage increasing traffic, especially during peak festive periods. This case study details their architectural decisions and lessons learned from the production migration.
- →Zepto Migrates to OpenSearch Service OR2 Instances for Performance
- →Significant Improvements in Throughput and Cost Efficiency
- →OpenSearch OR2 Instances Optimized for High Indexing Throughput
- →Zepto's Event-Driven Search Architecture
Notes (4) ›
- Zepto Migrates to OpenSearch Service OR2 Instances for Performance
Zepto moved its Amazon OpenSearch Service data nodes from memory-optimized instances to OpenSearch Optimized (OR2) instances to manage scaling product search for its quick-commerce platform. This migration addressed challenges in maintaining sub-second latency and controlling costs amid growing indexing volumes and traffic.
- Significant Improvements in Throughput and Cost Efficiency
The migration resulted in over 100% higher indexing throughput and 30% cost savings for Zepto. This allowed the company to serve the same workload with two-thirds of its previous data node count while maintaining critical search performance during peak demand.
- OpenSearch OR2 Instances Optimized for High Indexing Throughput
OpenSearch Optimized (OR2) instances are purpose-built for high indexing throughput and cost efficiency. They utilize local storage as the primary data tier, with Apache Lucene segments synchronously copied to Amazon S3 for durability and resilience.
- Zepto's Event-Driven Search Architecture
Zepto's search platform uses an event-driven architecture powered by Amazon MSK for streaming and Apache Flink on Amazon EKS for priority-based indexing pipelines. This setup, combined with OpenSearch Service, ensures real-time, hub-specific product availability for customers.
https://aws.amazon.com/blogs/big-data/how-zepto-powers-sub-second-search-using-opensearch-service-or2-instances/
Related releases
- Amazon EC2 Marks 20 Years of Cloud Compute Innovation AWS News Blog ·
- GPU-accelerated Apache Spark on Amazon EMR with EC2 G7 instances runs up to 3.7x faster AWS Big Data Blog ·
- AWS Lambda functions now support full IAM resource-based policies AWS What's New ·
- Amazon MSK 3.7 reaches end of life in 7 days endoflife.date ·
- Amazon ECS adds automated agent connectivity detection and repair AWS What's New ·
- AWS Neuron 2.32 Enhances NKI, MXFP8 Training, and Variable-Size Collectives AWS What's New ·