RTO-Based Data Restore Using Prioritized Cloud Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud environments with geographically dispersed storage zones, disaster recovery systems located far from failed storage nodes can be overwhelmed, leading to prolonged recovery times that fail to meet recovery time objectives due to resource constraints and distance.
Innovation Solution
Implementing extended attributes based on recovery time objectives for stored objects, prioritizing restoration by appending attributes like near-zero, near-threshold, and non-critical to objects, and using cloud object storage to sequence restoration based on these attributes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is restored sequentially from cloud object storage to target disaster recovery system, then data integrity is maintained, but recovery time increases and may not meet recovery time objectives
Solution Approach 1:
The system performs preliminary actions by appending extended attributes indicating restoration priority to objects before storage during normal operations. This pre-classification allows the disaster recovery system to immediately identify and restore critical objects first without needing to analyze all objects during the recovery process, thus reducing recovery time while maintaining data integrity through sequential restoration of prioritized objects
Solution Approach 2:
The restoration process is segmented into priority levels based on extended attributes (e.g., near-zero, near-threshold, non-critical). Instead of restoring all objects uniformly, the system divides the restoration process into segments handled in priority order, allowing critical data to be restored first to meet recovery time objectives while less critical data follows
2Reliability
If target disaster recovery system is located in geographically dispersed zone, then system availability is improved, but restoration speed decreases due to distance and network latency
Solution Approach 1:
Priority information is prepared in advance by appending extended attributes to objects during normal storage operations. This preliminary classification enables the geographically dispersed disaster recovery system to immediately begin restoring high-priority objects without delay, compensating for the slower network speed caused by geographic distance
Solution Approach 2:
The system rushes through the restoration of critical objects by identifying and restoring them first based on extended attributes. Less critical objects are restored later, allowing the system to meet recovery time objectives for important data despite the reduced restoration speed inherent in geographically dispersed architectures
3Productivity
If target disaster recovery system resources are consumed by other operations, then system utilization is optimized, but restoration priority is reduced and recovery time objectives may not be met
Solution Approach 1:
The restoration workload is segmented into priority tiers based on extended attributes. The disaster recovery system can allocate resources to restore high-priority objects first, ensuring recovery time objectives are met for critical data, while lower-priority restorations proceed when resources become available, thus balancing system utilization with recovery requirements
Solution Approach 2:
Priority classification is performed in advance during normal operations by appending extended attributes to objects. This preliminary action allows the disaster recovery system to immediately identify which objects require urgent restoration, enabling it to secure necessary resources for critical restorations even when the system is otherwise occupied with lower-priority operations
Data Source
AI summary
Computer-implemented methods for prioritization of data restoration based on recovery time objectives are provided. Aspects include receiving a request to store an object in a primary data volume, the object being associated with an application operating on a virtual server instance, appending an extended attribute to the object, and storing the object with the extended attribute appended in a primary storage device. Aspects also include periodically backing up the primary storage device to a cloud object storage device and transmitting a restore request to the cloud object storage device based on a determination that the primary storage device has become unresponsive. The restore request causes sequentially restoring a plurality of objects corresponding to primary data volume in an order based on the extended attribute appended to each of the plurality of objects.