NUMA Storage Path Selection Using Access Performance Metrics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data access methods in persistent storage systems do not efficiently optimize data paths based on performance, latency, and reliability, leading to suboptimal data access times and potential data loss risks.
Innovation Solution
A storage controller that calculates and selects the preferred data path for accessing persistent storage devices using access performance information, including latency, bandwidth, load, and reliability indicators, and periodically updates this information to adapt to changing system conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If data access systems use static data paths without optimization, then system complexity is reduced, but data access efficiency and performance deteriorate
Solution Approach 1:
The patent implements dynamic data path selection by continuously monitoring access performance metrics (latency, bandwidth, load, reliability) and recalculating preferred paths in real-time. The system transitions from static to dynamic path selection, allowing data access routes to adapt automatically to changing system conditions, thereby improving data access efficiency without requiring permanent structural complexity
Solution Approach 2:
The system establishes a feedback loop where access performance information is collected from actual data operations, processed to identify optimal paths, and used to guide subsequent data access decisions. This closed-loop control mechanism enables continuous improvement of data access efficiency while keeping the control logic centralized and manageable
Solution Approach 3:
The system pre-calculates and stores access performance information for multiple data paths before actual data operations occur. By maintaining a pre-computed preferred data path based on current performance metrics, the system avoids real-time calculation overhead during critical data access operations, thus improving efficiency without adding significant runtime complexity
2Reliability
If data paths are selected without considering access performance metrics, then calculation overhead is reduced, but data access performance and reliability deteriorate
Solution Approach 1:
The system monitors and responds to changes in key performance parameters (latency, bandwidth, load, reliability) of data paths. By detecting when these parameters cross certain thresholds or change significantly, the system triggers recalculation of preferred paths, thereby maintaining high reliability without continuously monitoring all parameters at full resolution
Solution Approach 2:
The patent introduces an intermediary data structure (preferred data path information stored in memory) that mediates between the complex performance metrics and the actual data access operations. This intermediary layer processes and simplifies the decision-making, hiding the complexity of multi-parameter optimization from the core data access logic while ensuring reliable path selection
3Productivity
If access performance information is not updated regularly, then system operations are simpler, but data access efficiency and performance optimization deteriorate
Solution Approach 1:
The system implements periodic updates of access performance information at strategically chosen intervals rather than continuously. By updating preferred path information periodically based on changing system conditions, the system maintains data access efficiency while avoiding the overhead of continuous monitoring and recalculation, thus reducing time loss without sacrificing performance optimization
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method comprising: obtaining a list of data paths to at least one persistent storage device through a plurality of NUMA nodes; associating with each data path, access performance information; receiving a request to access one of the at least one persistent storage device; calculating a preferred data path to the one of the at least one persistent storage device using the access performance information; and accessing the one of the at least one persistent storage device using the preferred data path.