Remote File Virtualization in Switched File Systems
Solution Overview
Problem
Remote file access across wide area networks (WANs) is impractical due to high latency and low bandwidth, exacerbated by 'chatty' NAS protocols like CIFS, which require numerous messages and lead to poor performance, especially when files are not frequently updated, and caching solutions do not effectively manage directory operations or stale data.
Innovation Solution
A switched file system with a central network file manager and remote network file managers that maintain mirrored copies of data and metadata, using lazy mirroring techniques to synchronize metadata and data, allowing remote sites to service client requests independently and reducing WAN traffic by caching and compressing data, while ensuring data integrity through oplock management and SMB signing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If file caching is used at remote sites, then file access speed is improved for cached files, but directory operations and stale data management remain problematic and require additional complexity
Solution Approach 1:
The patent creates a virtual copy of the entire file system namespace (including directories and files) at the remote site. This virtual file system mirror allows all operations including directory lookups to be performed locally without complex caching logic for determining what to cache. The virtual copy is maintained through synchronization protocols that handle updates from the central file server.
2Loss of time
If traditional caching solutions are used, then frequently accessed file data can be served locally, but directory operations still require remote communication and performance is not improved
Solution Approach 1:
The system performs preliminary action by pre-synchronizing the entire file system namespace (including all directory structures) to the remote site before they are needed. This allows directory operations to be performed immediately on the local virtual file system mirror without waiting for remote communication, eliminating directory operation latency.
3Loss of energy
If data is cached at remote sites, then network traffic is reduced, but ensuring data integrity and handling updates requires complex synchronization protocols
Solution Approach 1:
The system implements feedback mechanisms where the remote virtual file system monitor detects changes in the central file server's file system namespace and automatically triggers synchronization operations. This feedback loop ensures data integrity without requiring complex manual synchronization protocols, as the system self-regulates based on detected changes.
4Adaptability or versatility
If remote file access is implemented over WAN, then geographical distribution is enabled, but high latency and low bandwidth make file access impractical
Solution Approach 1:
The patent introduces a new dimension by creating a virtual file system layer at the remote site that mirrors the central file server's namespace. This virtual layer allows file access operations to occur in the local dimension (fast) rather than requiring communication across the WAN dimension (slow), while still providing access to remotely distributed files.
Data Source
AI summary
A plurality of network file manager switches interoperate to provide remote file virtualization. Copies of file data and/or metadata are maintained at a central site and at one or more remote sites. The network file manager switch at the remote site may satisfy certain client requests locally without having to contact the network file manager switch at the central site. A global namespace is maintained and is communicated to all network file manager switches.