Asynchronous Replication Cycle Overlap for Storage Workload Sharing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In data storage networks, peer storage systems face inefficiencies in asynchronously replicating a shared logical storage unit to a target system, leading to performance degradation due to strict synchronization of asynchronous replication cycles, which results in inefficient use of system resources and increased response times.
Innovation Solution
Implementing a method where peer storage systems share the workload by dividing ownership of replication cycles in an alternating manner without strict synchronization, allowing each system to own alternating cycle numbers and synchronize only at the beginning of each cycle, thereby reducing performance degradation and optimizing resource utilization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If strict synchronization of asynchronous replication cycles is implemented, then write consistency is maintained, but system resource utilization decreases and response time increases
Solution Approach 1:
The patent divides the replication cycle ownership into segments assigned to different storage systems. Each peer storage system is assigned specific cycle numbers (e.g., even cycles to first system, odd cycles to second system), allowing independent operation during assigned cycles while maintaining overall consistency through the segmented approach.
Solution Approach 2:
The patent implements periodic action by having peer storage systems alternate their participation in replication cycles at regular intervals. The systems switch between active and passive roles in a periodic manner, with each system handling specific cycles based on a predetermined pattern, reducing the need for continuous synchronization.
2Reliability
If strict synchronization of asynchronous replication cycles is implemented, then write consistency is maintained, but response time increases
Solution Approach 1:
By segmenting cycle ownership among peer systems, the patent eliminates the need for one system to wait for another during its assigned cycles, significantly reducing response time while maintaining consistency through the segmented ownership model.
Solution Approach 2:
The patent establishes preliminary agreements on cycle ownership patterns before replication begins. Peer systems know in advance which cycles they will own and can prepare accordingly, eliminating delays caused by real-time negotiation or synchronization decisions during critical write operations.
3Reliability
If both peer storage systems actively participate in every replication cycle, then data redundancy is increased, but system complexity and coordination overhead increase
Solution Approach 1:
The patent segments replication cycle participation so that each peer system is actively responsible for specific cycles rather than all cycles. This segmentation maintains data redundancy through alternating participation while dramatically reducing coordination overhead by eliminating the need for both systems to coordinate every cycle.
Solution Approach 2:
The patent extracts the coordination requirement from every replication cycle by assigning specific cycles to specific systems. This removes the complexity of inter-system coordination from the majority of cycles, keeping only essential synchronization points while maintaining redundancy benefits.
Data Source
AI summary
Peer storage systems share the workload of asynchronously replicating a shared logical storage unit (LSU) to a target system. Peer storage systems (S1 and S2) that synchronously replicate a shared LSU share the workload of asynchronously replicating the shared LSU to a target system (S3) by dividing ownership of the replication cycles in an alternating manner without a strict synchronization of cycles between the peer storage systems. Rather, a given cycle number (e.g., Cycle 1) on S1 may not start and end with a same write operation as a same cycle number (Cycle 1) on S2, such that cycles on S1 (e.g., Cycle 1 and Cycle 2) overlap with cycles having the same number (Cycle 1 and Cycle 2) on S2. S1 may asynchronously replicate to S3 only the cycles it owns, and S2 may asynchronously replicate to S3 only the cycles it owns to S3.