Score-Based Memory Access for Sensor Data Overflow
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The issue of memory overflow and data loss in data management systems due to large volumes of measurement data from sensors, exacerbated by network connectivity issues, leading to the discarding of valuable information.
Innovation Solution
Implementing a data management system with a data ring buffer and a scoring ring buffer that tracks time order and priority properties of data values, using scores to determine which data values to thin or retrieve, thereby optimizing memory usage and ensuring valuable data is sent to consumer devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If measurement data from sensors is stored in memory, then data availability for analysis is improved, but memory capacity is overwhelmed and data loss occurs
Solution Approach 1:
The patent changes the parameter of data selection from simple FIFO (first-in-first-out) to score-based priority selection. Each data value is assigned a score reflecting its importance, and memory eviction decisions are made based on these scores rather than arrival time. This parameter change allows the system to retain high-value data while evicting low-value data, resolving the contradiction between memory capacity and data availability.
Solution Approach 2:
The patent applies local quality by treating different data values differently based on their individual scores. Instead of uniformly evicting data based on arrival time, the system evaluates each data value's importance locally and makes eviction decisions based on that local assessment. This allows high-importance data to be retained even if it arrived earlier, while low-importance data is evicted regardless of arrival time.
2Quantity of substance
If data is evicted from full memory to make room for new data, then memory capacity is maintained, but valuable information is lost
Solution Approach 1:
The patent changes the eviction criterion from time-based (FIFO) to score-based. When memory is full, the system identifies the data value with the lowest score and evicts that data instead of arbitrarily evicting based on arrival time. This parameter change ensures that evicted data is the least valuable, preventing loss of important information while maintaining memory capacity.
Solution Approach 2:
The system performs self-service by automatically evaluating and scoring data values based on their inherent properties (time distances to neighbor data values). The scoring mechanism is self-contained and does not require external intervention to determine which data to evict. The system serves itself by making intelligent eviction decisions based on the scored importance of each data value.
3Reliability
If network connection is lost, then data management system continues receiving data, but memory fills up and data cannot be sent
Solution Approach 1:
The patent applies preliminary action by pre-calculating and assigning scores to data values based on their time distances to neighbor data values, even before network connectivity is restored. This scoring is done in advance so that when memory becomes full during network outages, the system can immediately identify which data to evict without delay, maintaining data continuity while managing memory volume effectively.
Solution Approach 2:
The system dynamically adjusts memory contents based on changing conditions. During network outages, the system continuously scores and re-evaluates data values as new data arrives, dynamically identifying which data to evict. This dynamic approach allows the system to adapt to network connectivity changes while maintaining optimal memory utilization and data continuity.
4Device complexity
If all data values are treated equally in memory management, then management simplicity is maintained, but data analysis insights are reduced
Solution Approach 1:
The patent changes the management parameter from uniform treatment to differentiated treatment based on scores. Each data value is assigned a score based on its time distances to neighbor data values, and management decisions (retention and eviction) are made based on these scores. This parameter change increases management complexity slightly but significantly improves analysis insights by ensuring high-value data is retained.
Solution Approach 2:
The system applies local quality by evaluating each data value's importance individually based on its local context (time distances to neighboring data values). Instead of treating all data uniformly, the system makes localized decisions about which data to retain or evict based on each data value's scored importance, thereby preserving analysis insights while managing memory effectively.
Data Source
AI summary
In some examples, a system includes a memory to store a first data structure containing data values received from one or more data sources, and a second data structure containing a plurality of scores associated with the data values, the data values in the first data structure arranged in a time order of the data values. The system includes a controller to: compute a score of the plurality of scores based on time distances of an associated data value to neighbor data values in the first data structure, and perform an access operation on the data values in the first data structure based on the plurality of scores.