On-Demand Snapshot Storage Service Instances
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing archival storage systems face challenges in efficiently indexing and managing large numbers of snapshots across multiple long-term storage services, leading to data corruption and high costs due to the need for cloud-based solutions and synchronization between instances.
Innovation Solution
The technique allows for the instantiation and management of long-term snapshot storage services on demand across geographical locations, using a multi-instance architecture that synchronizes operations through a shared object store without explicit communication, ensuring contention-free access and ownership of metadata for efficient retrieval and garbage collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple long-term storage service instances access the same snapshot data on a shared object store, then data availability and accessibility are improved, but data corruption occurs due to each instance assuming exclusive control over snapshot metadata
Solution Approach 1:
The patent introduces a shared object store as an intermediary between multiple long-term storage service instances and snapshot data. The object store provides a standardized interface that allows multiple instances to access snapshot data without directly competing for control of metadata, thus preventing data corruption while maintaining accessibility.
Solution Approach 2:
The shared object store serves multiple functions simultaneously: it acts as a common data repository for multiple storage service instances, provides geographic replication capabilities, and enables both read and write operations without requiring exclusive control. This universal interface resolves the contradiction between multi-instance accessibility and data integrity.
2Adaptability or versatility
If a shared storage service is hosted in a cloud service provider to allow global access to snapshots, then data accessibility from anywhere in the world is improved, but substantial costs are incurred due to usage of cloud compute resources
Solution Approach 1:
The patent allows storage service instances to be deployed in different geographic locations with local access to the shared object store. Each instance provides data accessibility to nearby clients without requiring all data to be replicated across all locations, reducing redundant cloud resource usage and costs while maintaining global accessibility.
Solution Approach 2:
The system uses selective copying through geographic replication where only necessary data copies are maintained in different locations. The shared object store enables efficient data distribution where copies are created only when and where needed, rather than maintaining full copies everywhere, thus reducing cloud resource costs while preserving global accessibility.
3Adaptability or versatility
If a shared storage service is used when the object store is on-premises, then accessibility is improved, but performance degrades because the service is not co-located with the object store
Solution Approach 1:
The patent introduces a new architectural dimension by allowing storage service instances to be deployed locally at different geographic locations with direct access to the shared object store. This eliminates the single remote service architecture, enabling clients to access data from the nearest instance while the object store maintains data centrally, thus improving both flexibility and performance simultaneously.
Data Source
AI summary
A technique allows instantiation and running on demand of long-term snapshot storage services of an archival storage system at various geographical locations. Storage service instances are configured to provide storage and retrieval of large amounts of point-in-time images or snapshots (e.g., recovery points) of application workloads stored as objects on one or more buckets of a shared object store. The storage service instances may contemporaneously serve snapshots of a same set of buckets on the shared object store without interfering with each other in a contention-free manner. That is, the technique enables storage service instances that are associated with snapshot workload data and/or metadata stored, e.g., as objects of a recovery point, on the same set of buckets to coexist without knowledge of each other. The storage service instances can be created and destroyed on-demand by splitting and merging the existing instances.