Storage Subsystem Asynchronous Write Acknowledgment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Storage systems face significant latency issues due to the need to copy old data out of its physical location before writing new data, especially when using slow stable storage like hard disk drives, which delays write operations and increases wait times for clients.
Innovation Solution
Implementing fast stable storage, such as battery-backed NVRAM, within the storage subsystem to queue and defer copy and write instructions, allowing these operations to be processed asynchronously and reducing latency by separating the acknowledgment of completion from the client and the actual processing time on slow storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If copy-out operation is performed before write operation in storage subsystem, then data integrity for snapshot images is maintained, but write latency increases significantly
Solution Approach 1:
The patent implements a write-in-place policy where write operations are performed immediately without requiring prior copy-out operations. The system maintains snapshot consistency through metadata updates and copy-on-write mechanisms triggered only when snapshots are created, rather than requiring copy-out before every write operation. This eliminates the latency penalty while preserving data integrity through alternative consistency mechanisms.
Solution Approach 2:
The patent separates the write operation from the copy-out operation, allowing writes to proceed independently. Snapshot copy-out operations are segmented to occur only when snapshots are created, rather than being coupled to every write operation. This segmentation allows the system to maintain reliability through targeted copy operations while minimizing latency through immediate write-in-place operations.
2Loss of time
If fast stable storage is added to storage subsystem to defer operations, then write latency is reduced, but device complexity increases
Solution Approach 1:
The patent extracts the fast stable storage requirement from the storage subsystem itself and places it in the storage server. This allows the storage subsystem to remain simple while the server handles deferred operations using its own fast storage resources. The subsystem processes writes immediately and defers copy-out operations asynchronously, eliminating latency without adding complexity to the subsystem architecture.
Solution Approach 2:
The patent introduces an intermediary deferred operation queue in the storage server that mediates between immediate write requests and asynchronous copy-out operations. This intermediary layer allows the storage subsystem to maintain simplicity while the server manages the complexity of deferred operations, separating concerns and reducing overall system complexity.
3Reliability
If storage server waits for storage subsystem to complete operations, then data consistency is ensured, but system productivity decreases
Solution Approach 1:
The patent enables continuous write operations in the storage subsystem without waiting for copy-out completion. The storage server continues processing new write requests while copy-out operations proceed asynchronously in the background. This continuity maintains data consistency through metadata tracking while maximizing productivity by eliminating wait states.
Solution Approach 2:
The patent implements feedback mechanisms where the storage subsystem notifies the storage server of operation completion status. This feedback allows the server to track consistency state without blocking productivity, enabling the system to maintain reliability through informed decision-making while keeping write operations flowing continuously.
Data Source
AI summary
A storage server receives a write request from a client system including new data and a location to store the new data. The storage server transmits a copy instruction to a storage subsystem to relocate old data at the location and transmits a write instruction to the storage subsystem to overwrite the old data with the new data. The storage subsystem includes fast stable storage in which the copy instruction and the write instruction are stored. After receiving each instruction, the storage subsystem sends an acknowledgement to the storage server. When both instructions have been acknowledged, the storage server sends an acknowledgement to the client system. The storage subsystem performs the instructions asynchronously from the client system's write request.