Akash

Akash is a Layer 4 load balancer written in Go. It supports multiple routing algorithms, active health checks, Prometheus metrics, TLS termination, and path-based routing.

⚠️ Warning: Akash is not production-ready. It was built purely for fun and learning purposes. It lacks stability, security, and robustness of a production-grade load balancer.

Website: akash.psidharth.dev

---

Tech Stack

---

Performance vs HAProxy (L4)

We benchmarked Akash against HAProxy, the industry standard L4 load balancer. The results show Akash performs on par with HAProxy in throughput and latency.

Metric	Akash	HAProxy (L4)
Total Requests	1,000,000	1,000,000
Throughput	71,290 req/s	71,585 req/s
Median Latency	2.1 ms	1.7 ms
95th Percentile	6.3 ms	7.2 ms
99th Percentile	12.1 ms	14.7 ms
Max Latency	129 ms	95.9 ms
Data Transferred	40 MB	40 MB
Success Rate	99.995%	99.995%

---

📊 Benchmark Insights

• Akash matches HAProxy in throughput (within ~0.5%).
• Latency distribution is nearly identical, with Akash slightly better at higher percentiles.
• Max latency spikes are different but within acceptable range.

For a Go-based project vs HAProxy’s C engine, Akash proves competitive.

---

Features

• Routing Algorithms

  • Round Robin
  • Least Connections
  • IP Hash
  • Weighted Round Robin

• Health Checks

  • Regular TCP-based health checks
  • Automatic marking of backends as healthy/unhealthy

• Metrics

  • Exposes Prometheus metrics on :9100/metrics
  • Tracks active connections, per-backend requests, and per-backend failures

• TLS Support

  • Optional TLS termination with provided certificates

• Path-Based Routing

  • Route specific URL paths to specific backends

• Graceful Shutdown

  • Handles termination signals
  • Closes active connections before exit

---

Getting Started

Prerequisites

• Go 1.21 or higher
• A valid config.json file (see example below)

Build

git clone https://github.com/your-username/akash.git
cd akash
go build -o akash

Run

./akash -config config.json

---

Configuration

Akash is configured using a JSON file. Below is an example:

{
  "host": "0.0.0.0",
  "listen": "1902",
  "algorithm": "round_robin",
  "max_connections": 1000,
  "timeout_seconds": 5,
  "health_check_path": "/health",
  "health_check_port": "80",
  "health_check_freq": 10,
  "tls_cert_file": "server.crt",
  "tls_key_file": "server.key",
  "Backends": [
    {
      "address": "127.0.0.1:8081",
      "weight": 1,
      "paths": ["/api"]
    },
    {
      "address": "127.0.0.1:8082",
      "weight": 2,
      "paths": ["/static"]
    }
  ]
}

Config Fields

• host: Address to bind the load balancer (default: 0.0.0.0)
• listen: Port to listen on (default: 1902)
• algorithm: Routing algorithm (round_robin, least_conn, ip_hash, w_round_robin)
• max_connections: Maximum number of active connections
• timeout_seconds: Timeout for backend health checks
• health_check_path: Path for HTTP health checks
• health_check_port: Port for health checks
• health_check_freq: Frequency of health checks (in seconds)
• tls_cert_file / tls_key_file: Paths to TLS certificate and key
• Backends: List of backend servers with address, weight, and optional paths

---

Metrics

Akash provides Prometheus-compatible metrics on :9100/metrics.

Exposed metrics include:

• akash_active_connections — Number of active client connections
• akash_backend_served_total{backend="..."} — Requests successfully served per backend
• akash_backend_failures_total{backend="..."} — Failed connections per backend

You can use Grafana to scrape metrics endpoint from Prometheus to build interactive dashboards

---

Future Plans

Planned features:

• WebSocket support
• HTTP/2 and gRPC proxying
• Retry policies and circuit breakers
• Rate limiting per IP

---

License

This project is licensed under the Apache 2.0 License.