Valkey Hashmap Benchmark

Benchmarks comparing concurrent query methods for multiple Valkey/Redis hashmaps:

1. Concurrent HGETALL: Using MULTI to batch multiple HGETALL commands
2. Concurrent EVALSHA: Using Lua script execution server-side

Prerequisites

• nvm (Node Version Manager)
• Valkey 8 server

Setup & Running

nvm use

npm install

npm run start-valkey

npm run setup-valkey

npm run benchmark

npm run stop-valkey

Multi-threaded Valkey Server Behavior

With io-threads 6 configuration, Valkey distributes I/O operations across multiple threads

Concurrent HGETALL (MULTI/EXEC)

• Each client connection sends MULTI followed by multiple HGETALL commands

• Commands are queued and executed atomically per connection

• Multiple concurrent connections can be processed in parallel by different I/O threads

• Network I/O for command parsing and response serialization is parallelized

• Actual command execution remains single-threaded on the main thread

Concurrent EVALSHA

• Lua script execution is atomic and blocks the main thread

• While one script runs, other connections must wait

• I/O threads can still handle network operations for queued requests

• Less parallelism compared to MULTI/EXEC approach despite multi-threading

Expected Performance Characteristics

1. Small hashmaps (2000 fields): Minimal difference; network overhead dominates

2. Medium hashmaps (4000 fields): MULTI/EXEC begins showing advantages through better parallelism

3. Large hashmaps (10000 fields): MULTI/EXEC significantly outperforms due to:

• Better utilization of I/O threads

• Non-blocking concurrent execution

• Reduced contention on the main thread

The benchmark simulates 4 concurrent clients to stress-test parallelism capabilities.