Object Storage
Object storage in the context of Panzura CloudFS takes on a specific role in the overall architecture. While data is ultimately stored as objects in the underlying cloud object storage, Panzura introduces an abstraction layer that transforms object-based storage into a familiar file system interface.
- Data Decomposition: Files are broken down into data blocks. These blocks are then stored as objects in the underlying object storage and metadata pointers are created that record the data blocks that comprise files.
- Metadata Management: Panzura maintains a global metadata store, which includes the metadata pointers. In conjunction with immutable snapshots, this metadata store is crucial for reconstructing files on demand — including restoring previous versions — and ensuring data consistency.
- Global File System: By managing metadata and data block locations, Panzura creates a virtualized, global file system that appears as a single, unified namespace accessible from any location.
Key differences from traditional object storage:
- Metadata-centric: Panzura's approach uses lightweight metadata for file management and reconstruction.
- Data deduplication: Panzura often employs deduplication techniques to optimize storage usage by storing identical data blocks only once.
- Global consistency: Panzura ensures immediate global data consistency by managing metadata centrally.