Atomic Write Management for Memory Subsystems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional memory subsystems face performance inconsistencies due to the aggregation of entire data pieces before writing, leading to underutilization of memory components while waiting for data aggregation, and inefficient handling of atomic write operations.
Innovation Solution
Decoupling the writing of data to memory components from updates to translation tables, where chunks of data are written as received with atomicity indicators, and translation table updates are accumulated and applied only after successful writing, ensuring atomicity and improved performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire piece of data is aggregated in working memory before writing to memory component, then atomicity of write operation is maintained, but memory components are under-utilized and performance is inconsistent
Solution Approach 1:
The patent segments the write operation into two independent parts: (1) data writing to memory components, and (2) translation table updates. This segmentation allows data to be written in chunks as it arrives without waiting for complete aggregation, while translation table updates are deferred until all data is successfully written, thus maintaining atomicity without blocking memory component utilization.
Solution Approach 2:
The patent performs preliminary action by writing data chunks to memory components as they arrive, before the entire data set is aggregated. The translation table updates are prepared in advance but held pending until completion confirmation, allowing memory components to work continuously while atomicity is preserved through the deferred update mechanism.
2Reliability
If data aggregation is completed before writing to memory component, then atomicity is ensured, but memory components wait idle resulting in under-utilization
Solution Approach 1:
The patent extracts the translation table update operation from the data writing process. By separating these two operations, the system can write data chunks to memory components immediately upon arrival without waiting for complete aggregation or coordinate translation table updates, thereby eliminating idle waiting time while ensuring atomicity through deferred update application.
Solution Approach 2:
The patent introduces an intermediary mechanism (the controller's coordination logic) that decouples data writing from translation table updates. This intermediary manages the timing and synchronization, allowing memory components to operate continuously while the controller ensures atomicity by applying translation table updates only after confirming complete data writing.
3Stability of the object's composition
If translation table updates are performed immediately with each data write, then data consistency is maintained, but atomicity of the entire write operation is compromised
Solution Approach 1:
The patent merges multiple translation table updates into a single batch update operation that is applied only after all data chunks are successfully written. This merging ensures that either all updates are applied together (maintaining atomicity) or none are applied (preserving data consistency through rollback), eliminating the partial update scenario that compromises atomicity.
4Reliability
If the system waits for complete data aggregation before sending to memory component, then atomicity is maintained, but memory component utilization is inefficient
Solution Approach 1:
The patent introduces dynamic control where the timing of translation table updates is flexible and adaptive. Updates are deferred dynamically until data writing is confirmed complete, allowing memory components to operate at full utilization throughout the process. This dynamic approach replaces static aggregation-waiting behavior with adaptive coordination that maintains atomicity while optimizing resource utilization.
Data Source
AI summary
Exemplary methods, apparatuses, and systems include receiving an instruction to atomically write data to a memory component. A plurality of write commands for the first data are generated, including an end of atom indicator. The first plurality of write commands are sent to the memory component while accumulating a plurality of translation table updates corresponding to the write commands One or more translation tables are updated with the plurality of translation table updates in response to determining that the final write command has been successfully sent to the memory component.