Hardware & System Requirements
"Enterprise Performance on Consumer-Grade Hardware"
Built entirely in Rust, PON-DB eliminates the need for heavy runtimes (like the JVM) or slow interpreters. This allows the engine to operate with a negligible memory footprint and native CPU efficiency.
1. Minimum Requirements (Embedded / IoT Edge)
Engineered for Edge Computing, IoT Gateways, and micro-servers.
-
CPU: 1 Core (x86_64 Architecture).
-
RAM: 128 MB (The engine can operate with less than 50MB of free system memory).
-
Storage: Capacity for the
.ponfile . -
OS: Linux (Kernel 4.x or higher), Alpine Linux (recommended for ultra-light containers).
2. Recommended Requirements (Production / VPS)
Optimized for high concurrency or datasets exceeding 100M+ records.
-
CPU: Intel Xeon / AMD EPYC (2+ Cores to leverage multi-threaded data ingestion).
-
RAM: 512 MB to 1 GB.
-
Storage: SSD / NVMe (While O(1) complexity mitigates seek latency, high-speed storage improves bulk write performance).
-
OS: AlmaLinux, Rocky Linux, Ubuntu Server, Debian.
Runtime Memory Footprint Comparison
| Feature | Traditional SQL Engines | PON-DB Engine / Reader |
| Dependencies | Java JRE / Python Libs / Shared Objects | Static Binary (Zero Dependencies) |
| Idle RAM Usage | 256MB - 1GB+ | < 1.8MB |
| Startup Time | Seconds (Boot-up/Init sequence) | Instantaneous (Micro-seconds) |
| Security | Software-level encryption (Latent) | Native AES-256 (Hardware Accelerated) |
Why PON-DB is Different
-
Zero Garbage Collection: Thanks to Rust's memory management, there are no "Stop-the-World" pauses. The latency you measure is the latency you get, every single time.
-
Portable by Design: The
pon-db-readeris a single, self-contained binary. No need to installmysql-client,psql, or Python libraries. Copy, execute, and scale. -
Hardware-Lock Ready: Built-in support for licensing tied to hardware UUIDs, ensuring security without penalizing query performance.
Developer Note: "If your hardware can run Linux, it can run PON-DB. We have successfully benchmarked the reader on instances with 1 vCPU and 512MB RAM, maintaining sub-25ms latencies on multi-million row datasets."