OpLock Revocation in Striped Storage Volumes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Managing rights, particularly opportunistic locks (OpLocks), in a striped volume set (SVS) environment is challenging due to difficulties in granting and revoking exclusionary rights across multiple clients and storage volumes, especially during data migration or failover operations.
Innovation Solution
A process is implemented to determine if an OpLock needs to be revoked, ascertain the location of the storage volume that owns the OpLock, and either process the revocation locally or send a recall message to a remote node to manage the revocation, ensuring all cached copies are invalidated to maintain data coherence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If OpLock is granted to a client for I/O operations on a data container in a striped volume set, then the client can perform efficient read and write operations with exclusionary rights, but managing and revoking the OpLock becomes complex when the data container is migrated or during failover operations
Solution Approach 1:
The patent segments the OpLock management by separating the data container location (first storage volume) from the OpLock residence location (second storage volume). This allows the OpLock to be managed independently at its resident volume while the data container can be migrated freely, reducing management complexity during failover and migration operations.
Solution Approach 2:
The patent introduces a recall message mechanism as an intermediary for OpLock revocation. Instead of directly managing OpLocks across multiple volumes, the system uses structured recall messages that traverse the storage network to notify relevant nodes and clients, simplifying the revocation process during data migration or failover.
2Speed
If the OpLock is cached at a first storage volume and resident at a second storage volume, then data access performance is improved through caching, but revoking the OpLock requires coordination between multiple storage volumes and nodes
Solution Approach 1:
The patent prepares for OpLock revocation by maintaining knowledge of the OpLock's residence location (second storage volume) separately from its cache location (first storage volume). This preliminary organization allows the system to quickly identify and coordinate with the correct node for revocation without searching through multiple volumes during the actual revocation process.
Solution Approach 2:
The patent implements a feedback mechanism where recall messages are sent from the first node to the second node that manages the OpLock residence. The second node processes the revocation and provides feedback by invalidating the cached copy, ensuring consistency across the distributed system while maintaining operational simplicity.
3Productivity
If multiple clients access the same data container in an SVS environment, then resource utilization is improved, but granting and revoking exclusionary rights becomes challenging
Solution Approach 1:
The patent segments the OpLock management by separating the data container location (first storage volume) from the OpLock residence location (second storage volume). This allows the OpLock to be managed independently at its resident volume while the data container can be migrated freely, reducing management complexity during failover and migration operations.
Solution Approach 2:
The patent introduces a recall message mechanism as an intermediary for OpLock revocation. Instead of directly managing OpLocks across multiple volumes, the system uses structured recall messages that traverse the storage network to notify relevant nodes and clients, simplifying the revocation process during data migration or failover.
Data Source
AI summary
A machine implemented process and system is provided. The process determines if any right granted to an entity for performing an I/O operation associated with a data container stored at a first storage volume is to be revoked. The right itself may be cached at the first storage volume and resident at a second storage volume. A recall message is sent to the node that manages the second storage volume. Any pending I/O operation is completed, after the message is sent and before the right is revoked. Thereafter, the right is revoked and the cached copies of the right are invalidated.