Storage Array Metadata Protocol for Copy Services
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing storage system copy services implementations require multiple implementations to cover all host types and switch types, consume host CPU resources, and double write bandwidth, and lack access to non-volatile memory, leading to performance and reliability issues, especially in data replication and recovery.
Innovation Solution
A new protocol that directs the replication engine to create metadata and transfer data between storage array components without passing through the replication engine, allowing for reduced bandwidth requirements and improved data integrity by mirroring writes within the storage array, enabling higher data rates and lower development costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data is routed through the replication engine for all operations, then data replication functionality is provided, but bandwidth requirements are doubled and data rates are reduced
Solution Approach 1:
The patent segments data operations into two paths: primary data operations bypass the replication engine directly between storage array components, while replication metadata operations flow through the replication engine. This segmentation allows bulk data transfers to utilize full bandwidth without replication overhead, while still maintaining replication functionality for critical operations.
Solution Approach 2:
The patent introduces metadata as an intermediary that carries replication control information separately from the primary data flow. This metadata mechanism allows the replication engine to manage copy services without requiring all data to pass through it, thereby maintaining replication reliability while preserving data transfer performance.
2Adaptability or versatility
If multiple implementations are used to cover all host types and switch types, then compatibility is improved, but device complexity increases
Solution Approach 1:
The patent implements a universal metadata protocol that can be used across different host types, switch types, and storage array configurations. This single metadata mechanism provides multi-functional support for various copy services (remote copy, point-in-time copy, continuous data protection) without requiring separate implementation code paths for each scenario.
Solution Approach 2:
The patent uses configurable metadata parameters that can be adjusted based on the specific host type, switch type, or copy service requirement. This allows the same basic infrastructure to adapt to different environments by changing metadata content and interpretation rather than changing the fundamental data flow path.
3Reliability
If host CPU resources are used for splitting writes, then copy services are provided, but host CPU MIPS are consumed
Solution Approach 1:
The patent extracts the write-splitting functionality from the host system and relocates it to the storage array metadata layer. The host only needs to issue standard write commands without any special processing, while the storage array's metadata mechanism automatically handles the replication splitting, thereby eliminating host CPU overhead for copy services.
Solution Approach 2:
The storage array's metadata system provides self-service write splitting by automatically intercepting and duplicating write operations to appropriate replication targets based on metadata configuration. This self-service mechanism eliminates the need for host-level software agents or CPU-intensive splitting logic, as the storage infrastructure handles replication autonomously.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus for controlling a storage system having a data replication function, comprises: a storage array component being operable to send notification to a replication engine that a write of data to a primary storage location by a host is subject to data replication; the replication engine being operable to receive the notification and in response to instruct the storage array to copy the data to a secondary storage location; wherein the data is copied to the secondary storage location unmediated by the replication engine.