Distributed Metadata Servers Using Shared Key-Value Store
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In High Performance Computing (HPC) applications, metadata servers in cluster file systems act as bottlenecks due to complicated locking protocols and multiple round-trip communications required for metadata consistency in distributed metadata access.
Innovation Solution
A cluster file system with distributed metadata servers that utilize a shared low latency persistent key-value metadata store, providing an abstract storage interface for independent metadata request processing and employing a failure detection mechanism for directory locking, allowing parallel metadata access without coordination between servers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If distributed metadata servers are used to parallelize metadata access, then metadata access performance is improved, but coordination overhead and communication complexity increase due to ACID requirements and two-phase commit protocols
Solution Approach 1:
The patent segments the metadata storage system into independent distributed metadata servers that each maintain local state, eliminating the need for centralized coordination. Each server can process metadata requests independently, parallelizing access without requiring two-phase commit protocols or complex locking mechanisms across servers.
Solution Approach 2:
The patent extracts the coordination and consistency management functions from the metadata servers themselves, placing them in the client layer through the use of version vectors and conflict resolution algorithms. This allows servers to operate independently without ACID transaction coordination, significantly reducing communication overhead.
2Reliability
If locking protocols are implemented to ensure metadata consistency, then data consistency is improved, but access speed and parallelization are reduced due to multiple round-trip communications
Solution Approach 1:
The patent implements self-service consistency mechanisms where clients independently track metadata versions using version vectors and automatically resolve conflicts without server intervention. This eliminates the need for server-mediated locking protocols and multiple round-trip communications, allowing parallel access while maintaining consistency.
Solution Approach 2:
The patent uses version vectors as feedback mechanisms that allow clients to independently determine whether their metadata reads are consistent with the latest state. This feedback enables clients to proceed with operations without server coordination, improving access speed while maintaining reliability through conflict detection and resolution.
Data Source
AI summary
A cluster file system is provided having a plurality of distributed metadata servers with shared access to one or more shared low latency persistent key-value metadata stores. A metadata server comprises an abstract storage interface comprising a software interface module that communicates with at least one shared persistent key-value metadata store providing a key-value interface for persistent storage of key-value metadata. The software interface module provides the key-value metadata to the at least one shared persistent key-value metadata store in a key-value format. The shared persistent key-value metadata store is accessed by a plurality of metadata servers. A metadata request can be processed by a given metadata server independently of other metadata servers in the cluster file system. A distributed metadata storage environment is also disclosed that comprises a plurality of metadata servers having an abstract storage interface to at least one shared persistent key-value metadata store.