VoidStore is a high-performance, event-driven in-memory key-value database built from scratch in C++ using Linux epoll. It implements core systems found in real-world data stores, including non-blocking I/O, RESP protocol parsing, TTL expiration, and append-only persistence.

To understand how systems like Redis handle networking, concurrency, and data persistence internally—beyond just using them as black boxes.

g++ src/*.cpp -Iinclude -o server
./server
redis-cli -p 8080

Handles ~60K requests/sec on local machine (single-threaded, epoll-based event loop).

Measured over multiple runs using redis-cli with piped input (100K SET operations) under minimal system load. Results include client-side overhead.

Observed range: ~60K–63K ops/sec.

Single-threaded design; performance can be further improved using pipelining and multi-threaded request handling.

• Event-driven server using epoll (non-blocking I/O)
• Redis-compatible RESP protocol parsing
• Multiple commands parsing using pipelining
• Multi-client TCP handling using sockets
• In-memory key-value storage with TTL expiration
• Append-Only File (AOF) persistence for durability
• Command support via redis-cli

Single Command

SET key value
SET key value ttl
GET key
DEL key

Multiple Commands

 printf '*3\r\n$3\r\nSET\r\n$4\r\nkey1\r\n$6\r\nvalue1\r\n*2\r\n$3\r\nGET\r\n$4\r\nkey1\r\n' | nc localhost 8080

• Designed an event-driven architecture using epoll for efficient multi-client handling
• Implemented lazy TTL expiration during read operations
• Built a custom RESP parser to handle Redis protocol communication
• Implemented AOF logging and replay to recover state on restart
• Used unordered_map for O(1) average-time operations
• This project focuses on understanding system internals rather than replicating full Redis functionality.

Systems like VoidStore (and Redis) are commonly used for:

• Caching frequently accessed data
• Managing user sessions and authentication
• Real-time features like chat presence
• Background job queues and rate limiting

• C++
• POSIX sockets
• epoll (Linux I/O multiplexing)
• Redis RESP protocol
• STL (unordered_map)

• LIST / SET data structures
• Background TTL cleaner (active expiration)
• RDB snapshot persistence
• Replication support
• Non-blocking socket improvements