ZKIM CAS: A Novel Content Addressable Storage System

Abstract

This paper presents ZKIM CAS (Content Addressable Storage), a decentralized storage system that uses BLAKE3 hashing for content addressing, FastCDC (Fast Content-Defined Chunking) for efficient data management, and a multi-layer storage architecture for high availability. ZKIM CAS addresses the challenges of content deduplication, distributed storage, and efficient content retrieval in decentralized messaging platforms. The system generates deterministic Object IDs (ZOIDs) from content hashes, enabling automatic deduplication and efficient content distribution. We describe the architecture, implementation details, and performance characteristics of ZKIM CAS, demonstrating its effectiveness in production deployments. The system achieves content deduplication rates of up to 95% for similar content and supports multi-peer retrieval for large objects, reducing storage overhead and improving retrieval performance.