Concurrent Key-Value Store

Project Overview

This is an implementation of a concurrent key-value (KV) store using a client-server architecture. Both client and server are multi-threaded processes that communicate through a shared memory region. The system uses a ring buffer for inter-process communication and a thread-safe hash table to store key-value pairs.

System Architecture

Components

Ring Buffer

• Thread-safe circular buffer for request submission and retrieval
• Supports concurrent access from multiple client and server threads
• Implements blocking behavior when full (client) or empty (server)

Key-Value Store

• Hash table implementation with fine-grained locking for concurrent access
• Supports put(key, value) to insert or update values
• Supports get(key) to retrieve values

Request-Status Board

• Shared memory region for tracking request status
• Allows clients to check completion status and retrieve responses

Building and Running

To build the project:

make

To run the server:

./server -n <num_threads> -s <init_table_size>

Where:

• -n specifies the number of server threads
• -s specifies the initial hash table size

To run the client:

./client -n <num_threads> -w <window_size> [-f] [-i workload_file]

Where:

• -n specifies the number of client threads
• -w specifies the window size per client thread
• -f forks the server process automatically
• -i specifies the input workload file

Design Decisions

• Used fine-grained locking in the hash table to allow multiple threads to access different buckets concurrently
• Implemented efficient synchronization in the ring buffer to minimize contention
• Organized shared memory to support multiple client threads with dedicated status windows

Performance Considerations

• Multiple server threads can process requests concurrently to improve throughput
• Window size controls the maximum number of in-flight requests per client thread
• Thread-safe data structures are optimized to minimize lock contention