GridGain Benchmarks Results

Benchmarks were performed to determine the performance of various cache operations and SQL based cache queries in GridGain^® Community Edition 8.7.3. The results obtained provide throughput, latency, and dStat probes measuring various metrics during benchmark execution.

Hardware Specification

Yardstick Configuration

GridGain was benchmarked using Yardstick framework with the configuration below.

If you want to reproduce the results of the benchmarks, please refer to this page which provides guidance on how to setup and run Yardstick against a GridGain cluster. Also, the configurations and scripts that were used on the GridGain side are available.

Results Summary

For all benchmarks, one local node and three remote nodes were configured with 128 client threads and a test duration of 300 sec.

• In-Memory Store Only
• With Ignite Persistence

Atomic Put Benchmark

Benchmarks atomic distributed cache put operation in GridGain. Randomly generated key-value pairs are stored in cache on multiple nodes.

Benchmarks Results Summary

Results show that atomic cache put operations in GridGain performed exceptionally well with approximately 330,000 operations per second.

• In-Memory Store Only
• With Ignite Persistence

Atomic Put and Get Benchmark

Benchmarks atomic distributed cache put and get operations in GridGain. Randomly generated key-value pairs are stored and retrieved from cache on multiple nodes.

Benchmarks Results Summary

Results show that more than 170,000 cache put and get operations per second can be performed in GridGain.

• In-Memory Store Only
• With Ignite Persistence

Transactional Put Benchmark

Benchmarks transactional distributed cache put operation. Randomly generated key-value pairs are stored in PARTITIONED cache with TRANSACTIONAL atomicity mode.

Results Summary

Following results show that more than 216,000 transactional put operations can be performed in GridGain.

• In-Memory Store Only
• With Ignite Persistence

Transactional Put and Get Benchmark

Benchmarks transactional distributed cache put and get operations together. Randomly generated key-value pairs are stored and retrieved in a single transaction from PARTITIONED cache with TRANSACTIONAL atomicity mode.

Results Summary

Following results show that approximately 61,000 transactional cache put and get operations per second can be performed together in GridGain.

• In-Memory Store Only
• With Ignite Persistence

SQL Query Benchmark

Benchmarks distributed SQL query operation on cached data. A simple SQL query was executed on the data stored on multiple nodes in a PARTITIONED cache.

Results Summary

Results show a throughput of approximately 98,000 SQL based cache query operations per second, which proves how quickly data can be retrieved from In-Memory Data Grid through cache queries in GridGain.

• In-Memory Store Only
• With Ignite Persistence