Project: Building a High-Performance In-Memory Key-Value Store (TopicKV) — Phase 2 Capstone

Project: Building a High-Performance In-Memory Key-Value Store (TopicKV) — Phase 2 Capstone

Table of Contents

• Architecture: Sharded Store Design
• Core Data Model: Entry with TTL
• Sharded Storage for Scalability
• Thread-Safe GET with std::optional
• Thread-Safe SET with RAII
• TTL Expiry and Lazy Deletion
• LRU Eviction Policy
• Command-Line REPL with std::format
• Statistics and Metrics
• Performance Benchmarks
• Extension Challenges
• Phase 2 Reflection

Architecture: Sharded Store Design

To avoid a single global lock bottlenecking all operations, TopicKV uses sharding — 16 independent sub-maps each with their own mutex:

Core Data Model: Entry with TTL

Sharded Storage for Scalability

Thread-Safe GET with std::optional

Thread-Safe SET with RAII

TTL Expiry and Lazy Deletion

LRU Eviction Policy

Command-Line REPL with std::format

Extension Challenges

1. Persistence: Serialize the entire store to a binary file on SIGTERM (use std::ofstream + std::bit_cast for POD types)
2. Pub/Sub: Add a SUBSCRIBE key / PUBLISH key msg command using std::condition_variable and a std::deque<std::string> channel
3. Transactions: Implement MULTI / EXEC by buffering commands in a std::vector<Command> and applying atomically
4. Network interface: Wrap the store in a TCP server using std::jthread (from Module 14) — one thread per connection using the REPL dispatch logic

Phase 2 Reflection

Proceed to Phase 3: Inheritance & Composition →

Part of the C++ Mastery Course — 30 modules from modern C++ basics to expert systems engineering.