SusyDB

Tagline: SusyDB is a tiny, high-performance, in-memory key-value store written in Go, optimized for simple cache, sessions, and counters.

---

Why SusyDB?

Teams often deploy Redis just to store a few session tokens or cache API responses.

SusyDB is for when you want:

• Zero Ops: Single binary, no dependencies.
• Simplicity: Just SET, GET, TTL, Hashes, and Counters.
• Pure Go: Easy to hack on, embed, or deploy alongside Go microservices.

What it's NOT:

• A durable database (No persistence yet).
• A distributed cluster (No sharding).

---

Architecture & Design

SusyDB follows a shared-nothing, single-process architecture. It handles concurrency using Go's native synchronization primitives.

graph LR
    Client -->|TCP| Server
    Server -->|Goroutine| Parser
    Parser -->|RWMutex| Memory

Loading

• Concurrency: Uses sync.RWMutex for thread-safe access to the global map.

  Note: While a global lock is simple, SusyDB minimizes contention by keeping critical sections purely in-memory and extremely short. For 99% of session/cache workloads, the network I/O is the bottleneck, not the lock.

• Eviction: Hybrid approach (Lazy check on Get + Active background sweeper).
• Communication: Simple text-based TCP protocol.

Read the Full Architecture Doc for deep dives into specific modules.

---

Features

• Protocol: 100% RESP-compatible (works with redis-cli).
• Data Structures:
  • Strings: SET, GET, DEL, SETEX (legacy TTL command).
  • Hashes: HSET, HGET, HDEL, HGETALL (Perfect for sessions).
  • Counters: INCR, INCRBY (Rate limiting ready).
• Pub/Sub: Lightweight Message Broker (PUBLISH, SUBSCRIBE).
• Embedded Mode: Use as a library import "github.com/Syed-Suhaan/SusyDB/pkg/core" in your Go apps.
• Hybrid Expiry: Lazy + Active TTL implementation.
• Architecture: Thread-safe design using sync.RWMutex.
• Observability: INFO command for stats.

---

Quick Start

Installation

go install github.com/Syed-Suhaan/SusyDB/cmd/susydb@latest

Running

susydb --addr :7379

Docker

# Build the image
docker build -t susydb .

# Run it
docker run -d -p 7379:7379 susydb

Usage

Connect with telnet or netcat:

$ telnet localhost 7379
Trying 127.0.0.1...
Connected to localhost.

SET user:1 Suhaan
+OK

GET user:1
Suhaan

INCR hits
:1

INCR hits
:2

INFO
$55
# Stats
keys:2
uptime_in_seconds:12

Embedded Library

SusyDB can be imported directly into your Go applications, bypassing the network layer entirely for ultra-low latency.

package main

import (
    "fmt"
    "github.com/Syed-Suhaan/SusyDB/pkg/core"
)

func main() {
    // Zero setup, pure Go
    db := core.NewKVStore()

    // Use it just like the server
    db.Set("session:123", "active", 3600)
    
    val, _, _ := db.Get("session:123")
    fmt.Println(val) // Output: active
}

---

Benchmarks

Fairness Disclosure

These tests compare Redis (v5.0.14) and SusyDB (v1.0.0) running natively on the same Windows 11 machine. Both used default configurations and the same Go-based benchmarking client.

Machine Specs: Native Windows 11, Localhost. Workload: 50 Concurrent Clients, 100,000 Requests, Small (16B) payloads.

Workload	Redis (RPS / P99)	SusyDB (RPS / P99)	Gain
A. Mixed (50/50 SET/GET)	99k / 5.2ms	154k / 1.8ms	1.5x
B. TTL (SETEX 60s)	17k / 24ms	170k / 0.9ms	9.5x
C. Counters (INCR Hot Key)	22k / 4.8ms	163k / 1.1ms	7.3x
D. Hash (Session HSET/HGET)	19k / 5.1ms	168k / 2.3ms	8.4x
E. Pub/Sub (Message Blast)	18k / 5.4ms	143k / 1.2ms	7.6x

Context: Redis is optimized for Linux; its Windows port is known to be slower due to OS-level threading differences. However, this demonstrates SusyDB's efficiency as a native Go solution for Windows-based microservices.

Scope Note: SusyDB is specialized for lightweight, in-memory session/cache workloads. It does not replace Redis features like replication, clustering, or disk persistence.

Reproduce this Bench:

# Mixed Workload
go run cmd/susy-bench/main.go -h localhost:7379 -c 50 -n 100000 -test mixed

# TTL Workload
go run cmd/susy-bench/main.go -h localhost:7379 -c 50 -n 100000 -test setex

---

License

MIT License. See LICENSE for details.