Distributed File System Token Revocation Parallel I/O
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distributed file systems experience performance deterioration when multiple nodes attempt to read and write the same file simultaneously due to the need for sequential disk I/O processes and inter-node communications, leading to increased access times and reduced throughput.
Innovation Solution
The solution involves adding information about the node requesting a token to the revoke-request, allowing files to be delivered in a dirty state without being written back to the storage device, thereby reducing disk I/O processing and enabling parallel processing with minimal inter-node communications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a node holds a write-level data token for a file and another node needs to access that file, then data consistency is maintained through sequential token revocation, but system performance deteriorates due to multiple sequential disk I/O processes and inter-node communications
Solution Approach 1:
The patent applies preliminary action by having the requesting node (client B) perform the disk I/O operation in advance before receiving the token revocation confirmation. When client B needs to access a file held by client A, client B directly reads or writes the file from its local cache without waiting for the token revocation sequence to complete. This preliminary action eliminates the sequential dependency, allowing the I/O operation to proceed in parallel with the token management process, thus resolving the contradiction between maintaining data consistency and preserving system throughput
Solution Approach 2:
The patent uses the node's local cache as an intermediary to decouple the token revocation process from the actual data access. Instead of requiring direct sequential communication between the server and storage device, the local cache acts as a buffer that allows nodes to access files locally while the token management proceeds independently. This intermediary mechanism enables parallel processing of token revocation and data access operations, improving system throughput while maintaining consistency through the cache coherence protocol
2Reliability
If the system requires four inter-node communications and two disk I/O operations for token revocation, then data integrity is ensured, but access time increases and throughput decreases
Solution Approach 1:
The patent applies preliminary action by having the requesting node (client B) perform the disk I/O operation in advance before receiving the token revocation confirmation. When client B needs to access a file held by client A, client B directly reads or writes the file from its local cache without waiting for the token revocation sequence to complete. This preliminary action eliminates the sequential dependency, allowing the I/O operation to proceed in parallel with the token management process, thus resolving the contradiction between maintaining data consistency and preserving system throughput
Solution Approach 2:
The patent applies dynamics by making the token revocation process asynchronous and non-blocking. Instead of a rigid sequential process where I/O operations must wait for token revocation, the system dynamically allows I/O operations to proceed independently while token management continues in the background. The use of events and callbacks enables the system to adapt the execution flow, allowing overlapping of I/O and token management operations, thereby reducing access time while maintaining data integrity
3Reliability
If the system performs sequential processing of token revocation and I/O operations, then correctness is maintained, but system performance and throughput are reduced
Solution Approach 1:
The patent applies preliminary action by having the requesting node (client B) perform the disk I/O operation in advance before receiving the token revocation confirmation. When client B needs to access a file held by client A, client B directly reads or writes the file from its local cache without waiting for the token revocation sequence to complete. This preliminary action eliminates the sequential dependency, allowing the I/O operation to proceed in parallel with the token management process, thus resolving the contradiction between maintaining data consistency and preserving system throughput
Solution Approach 2:
The patent applies continuity of useful action by ensuring that both token management and I/O operations continue simultaneously without interruption. The asynchronous event-driven architecture allows the token revocation process to continue in the background while the requesting node performs its I/O operations independently. This continuous parallel execution of multiple useful actions maximizes system throughput while maintaining correctness through the eventual consistency model and cache coherence protocols
Data Source
AI summary
Disclosed is a distributed file system using the token method to allow file sharing by multiple clients (nodes) for improved system throughput. A storage device for holding files, and a server for managing tokens, and multiple clients for reading and writing the files in the storage device are connected over a network. On the distributed file system by which the client requests files by tokens, the server contains a token request means for requesting the return of tokens from the client holding the file. This token return request means sends a token return request means containing information on the client requesting the files by tokens and information showing the contents of the token.