Elasticsearch Deployment and Use Cases in Major Chinese Companies

1. JD.com to Home Order Center Elasticsearch Evolution

JD.com’s order center faces massive read‑heavy traffic, so it offloads query pressure to Elasticsearch while keeping the source data in MySQL. The cluster now stores over 1 billion documents and handles 500 million queries per day, using a real‑time active‑standby architecture with VIP load balancing, gateway client nodes, and data nodes with one primary and two replicas.

The team tuned shard count to balance single‑ID lookup throughput against aggregation pagination performance, and archived older orders to a historical database.

2. Ctrip Elasticsearch Application Cases

2.1 Ctrip Hotel Order Elasticsearch Practice

Ctrip built a lightweight, easy‑to‑deploy Elasticsearch index for hotel orders, using real‑time indexing of sharded databases and a dedicated web service to improve query convenience.

2.2 Ctrip Flight Ticket Elasticsearch Cluster Operations

Data flows from Kafka through ETL pipelines into various storage tiers (HDFS for cold data, databases/cache for hot data). Traditional BI reports coexist with real‑time machine‑driven decision loops, forming a closed‑loop data usage model.

2.3 Ctrip Large‑Scale Elasticsearch Cluster Management Insights

The largest Ctrip log cluster runs 120 data nodes on 70 physical servers, indexing 60 billion documents daily (≈25 TB new index files, 50 TB with replicas), sustaining peak indexing rates of millions of records per second, retaining data for 10‑90 days, and managing 3 441 indices, 17 000 shards, and about 1 PB of disk usage.

3. Qunar Order Center Elasticsearch Solution

Qunar’s hotel order volume grew to 1 million daily orders, exceeding the capacity of a hot‑table sharding approach. They introduced Elasticsearch to store searchable fields while the relational database kept detailed order data, achieving a split‑model where simple ID queries hit MySQL and complex searches use Elasticsearch.

The index is configured with eight primary shards, holding 140 million documents (≈64 GB) on a cluster with 240 GB total disk capacity.

4. 58.com Information Security Department Elastic Stack

The department deployed the Elastic Stack for log storage, near‑real‑time analysis, and security monitoring, covering ingestion, storage selection, performance challenges, master and data node optimizations, high‑throughput low‑latency search tuning, and Kibana visualizations for operations.

5. Didi Elasticsearch Multi‑Cluster Architecture Practice

Didi built a platform with over 3 500 Elasticsearch instances and >5 PB of data, handling >20 million writes per second. All writes pass through a Kafka‑based Sink service that streams data to multiple clusters for disaster recovery, while all queries go through a Gateway service exposing HTTP, TCP, and SQL interfaces, providing access control, rate limiting, DSL‑level throttling, and multi‑cluster failover.

6. Practical Order Search Solution Based on Elasticsearch

The solution leverages Elasticsearch’s structured query capabilities and real‑time update features to replace traditional order reporting bottlenecks. The architecture wraps Elasticsearch and sharded databases behind a unified order service, serving both front‑end/back‑end applications and reporting systems.