Performance Benchmark Report: RedisJSON vs MongoDB and ElasticSearch

Overview

The recent official benchmark demonstrates that RedisJSON (RediSearch) dramatically outperforms MongoDB and ElasticSearch in isolated writes (5.4× faster than MongoDB, >200× faster than ElasticSearch) and isolated reads (12.7× faster than MongoDB, >500× faster than ElasticSearch). In mixed workloads, RedisJSON maintains stable latency while ElasticSearch degrades.

Query Engine

RedisJSON’s development emphasizes performance; version 2.2 improves load and query speed by 1.7× over 2.0, with better throughput and lower data‑load latency.

Loading Optimization

NYC taxi benchmark results illustrate consistent performance gains across all new releases.

Full‑Text Search Optimization

Indexing 5.9 million Wikipedia abstracts and running full‑text queries shows substantial latency reductions when moving from v2.0 to v2.2.

Comparison with Other Frameworks

RedisJSON, MongoDB 5.0.3, and ElasticSearch 7.15 were tested on identical AWS m5d.8xlarge VMs (one client + three database nodes) using the YCSB benchmark, including a custom search operation.

Benchmark Setup

MongoDB 5.0.3 : three‑member replica set with text index.

ElasticSearch 7.15 : 15 shards, query cache enabled, RAID‑0 NVMe SSD.

RedisJSON* : RediSearch 2.2 + RedisJSON 2.0 on OSS Redis Cluster v6.2.6 with 27 shards.

100% Write Benchmark

RedisJSON* ingests data 8.8× faster than ElasticSearch and 1.8× faster than MongoDB, keeping sub‑millisecond latency (99 % of requests < 1.5 ms). It uniquely updates indexes in real time, unlike ElasticSearch’s near‑real‑time refresh.

100% Read Benchmark

RedisJSON* reads 15.8× faster than ElasticSearch and 2.8× faster than MongoDB, maintaining sub‑millisecond latency across the board.

Mixed Read/Write/Search Benchmark

In a realistic 65 % search / 35 % read workload (with 10 % writes), RedisJSON* sustains throughput 50.8× higher than MongoDB and 7× higher than ElasticSearch, while reducing latency up to 91× versus MongoDB and 23.7× versus ElasticSearch.

Full Latency Analysis

Across sustained loads (250 ops/sec and 6000 ops/sec), RedisJSON* consistently shows the lowest p99 latency (≈0.23 ms for reads, ≈3 ms for writes), whereas ElasticSearch suffers high tail latency due to GC and cache misses.

MongoDB vs ElasticSearch vs RedisJSON* Latency

MongoDB leads in search latency, but RedisJSON* remains the only solution keeping sub‑millisecond latency across all percentiles.

ElasticSearch vs RedisJSON* Latency

At 6000 ops/sec, RedisJSON* p99 read latency is 3 ms versus ElasticSearch’s 162 ms; update latency is 3 ms versus 167 ms.