Journal-Based Any Point in Time Backup for Virtual Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for protecting virtual machines, such as snapshots, disrupt applications and are limited in frequency, leading to challenges in maintaining a low recovery point objective (RPO) and efficient data protection.
Innovation Solution
A data protection system that replicates IO streams as a journal, allowing for continuous backup generation without native snapshots, enabling any point in time backups with reduced overhead and storage efficiency through deduplication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If snapshots are used to backup virtual machines, then data protection is achieved, but applications are disrupted and snapshot frequency is limited
Solution Approach 1:
The backup process is segmented into two independent components: (1) a production storage system that continues to serve applications without interruption, and (2) a separate backup storage system that receives replicated data through a data stream splitter. This segmentation allows backup operations to proceed independently without affecting application performance or availability.
Solution Approach 2:
A data stream splitter acts as an intermediary component that intercepts data streams between the production storage and the host system. This intermediary copies data for backup purposes while allowing the original data stream to continue flowing to the production storage, enabling simultaneous backup and application operation without disruption.
2Reliability
If snapshot frequency is increased to reduce RPO, then data loss between backups is reduced, but system limits prevent frequent snapshots and storage resources are consumed
Solution Approach 1:
Instead of creating traditional snapshots that require quiescing and stunning virtual machines, the system continuously copies data streams to backup storage in real-time. This copying mechanism eliminates the need for frequent snapshots while achieving near-zero RPO, as backup data is continuously synchronized with production data without being constrained by system snapshot limits.
3Productivity
If mirroring systems are used to compensate for snapshot frequency, then backup frequency is improved, but storage resources are substantially consumed and management overhead is significant
Solution Approach 1:
The system performs preliminary data preparation and filtering before storing backup data. By continuously monitoring and filtering data streams to identify only changed blocks, the system prepares data for efficient storage before it reaches the backup storage system. This preliminary action enables high-frequency backups while minimizing storage resource consumption through selective data transmission.
Solution Approach 2:
The system changes the parameter of data transmission from complete data copying to selective block-level data transmission. By monitoring data changes and transmitting only modified blocks, the system achieves frequent backup operations while significantly reducing storage resource requirements compared to traditional mirroring systems that copy entire data sets.
Data Source
AI summary
Generating any point in time backups without native snapshot generation. Production data is split such that a journal stream is sent to a data protection system, which may be local or remote. The journal stream includes a data stream and a metadata stream. Backups are synthesized at the data protection system by rolling at least a portion of the journal. A backup for any point in time represented in the journal can be synthesized.