We built another object storage
11 hours ago
- #object-storage
- #high-performance
- #cloud-computing
- High-performance object storage is expensive at scale, making it unusable for many.
- FractalBits addresses the high-performance trap by offering affordable high IOPS.
- Modern workloads like AI and analytics demand low latency and efficient metadata handling.
- Traditional object storage struggles with small objects, high request costs, and lacks directory semantics.
- FractalBits introduces a metadata engine using an on-disk radix tree for efficient path-like keys.
- The system is built with Zig for performance and Rust for the S3-compatible API server.
- FractalBits operates on a BYOC (Bring Your Own Cloud) model, ensuring cost transparency and data sovereignty.
- Benchmarks show significant cost reductions compared to S3 Express One Zone.
- The solution supports native directory semantics, atomic rename, and strong consistency.
- FractalBits aims to bridge the gap between high performance and affordability in object storage.