Decoupled Replication Sessions for Asynchronous Fan-Out Storage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data replication methods, particularly in concurrent fan-out asynchronous replication, are inefficient as they require maintaining different sets of source data snapshots for each target site storage system and often necessitate waiting for all target systems to complete a replication cycle before starting the next, leading to performance issues.
Innovation Solution
Implementing decoupled replication sessions in an information processing system, where a source storage system captures snapshots and generates differential data for multiple target systems in parallel, allowing each replication session to proceed independently and commence the next cycle without waiting for others to finish, thereby avoiding the need for separate snapshots and enhancing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If concurrent fan-out asynchronous replication is implemented using conventional approaches, then data can be replicated to multiple target site storage systems, but the source site storage system must maintain different sets of source data snapshots for each target site storage system, increasing device complexity
Solution Approach 1:
The patent merges the snapshot management for multiple target sites into a single shared snapshot structure. Instead of maintaining separate snapshot hierarchies for each target site, the system creates one snapshot per replication cycle that is shared across all target sites, thereby reducing device complexity while maintaining concurrent fan-out replication capability
Solution Approach 2:
The patent implements a universal snapshot structure that serves all target site storage systems simultaneously. A single snapshot can be used as the basis for replicating data to multiple different target sites, making the snapshot mechanism multi-functional and eliminating the need for separate snapshot sets for each target
2Reliability
If conventional asynchronous replication is used, then data replication can be performed in cycles, but the source site storage system must wait for all target site storage systems to complete a given replication cycle before commencing the next replication cycle, reducing productivity
Solution Approach 1:
The patent takes preliminary action by creating all necessary snapshots for multiple target sites at the beginning of a replication cycle before any data transmission begins. This allows subsequent replication operations to proceed independently and in parallel, as the snapshot foundation is already in place for all targets, eliminating the need to wait for sequential completion
Solution Approach 2:
The patent segments the replication process into independent parallel operations. After the initial snapshot creation, each target site's replication can proceed as an independent segment without blocking others, allowing the system to initiate multiple replication cycles concurrently rather than waiting for sequential completion
3Adaptability or versatility
If separate snapshot hierarchies are maintained for each target site storage system, then data can be replicated to multiple targets, but storage system performance is adversely impacted due to the overhead of managing multiple snapshot sets
Solution Approach 1:
The patent combines the snapshot management overhead into a single unified structure that serves all target sites. By merging the snapshot hierarchies into one shared structure, the system eliminates the redundant overhead of managing multiple separate snapshot sets, thereby improving storage system performance while maintaining multi-target replication capability
Data Source
AI summary
A first storage system in one embodiment participates in a cycle-based asynchronous replication process with multiple additional storage systems. The first storage system establishes replication sessions with respective ones of the additional storage systems for replication of source data. For each of multiple cycles, the first storage system captures a snapshot of the source data for a current cycle, generates differential data between the snapshot and a corresponding snapshot for a previous cycle, and transmits the differential data to the additional storage systems in their respective replication sessions. In conjunction with completion of transmission of the differential data to a given one of the additional storage systems, the corresponding replication session determines whether or not a snapshot of the source data for a subsequent cycle is available. If there is no such snapshot available, the replication session triggers capture of the snapshot of the source data for the subsequent cycle.