Workload Monitoring Gap Compensation via Absence Length Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing workload monitoring systems face challenges in accurately accounting for large gaps in time series data, leading to incorrect ignoring of actual downtimes when creating client workload models.
Innovation Solution
The system transforms raw performance data into interval performance data by partitioning the monitoring period into equal intervals and assigning null and non-null values, then converts this data into an absence length sequence to compensate for gaps, using algorithms to derive the absence length sequence from the presence sequence, and treats time intervals with excessive absence length as downtime.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If raw performance data is interpolated to fill gaps in time series data, then the workload model can be created continuously, but actual downtimes are incorrectly ignored and reliability deteriorates
Solution Approach 1:
The patent segments the time series data into presence intervals (where data is available) and absence intervals (where data is missing due to downtime). By creating an absence length sequence that explicitly marks these segments, the system prevents interpolation from incorrectly filling downtime gaps while still maintaining continuous workload modeling for periods when data is available.
Solution Approach 2:
The absence length sequence acts as an intermediary data structure between the raw performance data and the workload model. It provides metadata about data availability that guides the modeling process, preventing inappropriate interpolation during downtime while allowing normal interpolation during periods when the monitored system is operational.
2Measurement precision
If gap compensation is implemented through complex data transformation processes, then data accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent divides the data processing into distinct sequential stages: (1) transforming raw performance data into presence/absence indicators, (2) calculating absence lengths for consecutive absent intervals, (3) applying threshold-based filtering to identify downtime, and (4) generating the corrected workload model. This segmentation makes the complex process more manageable and computationally efficient.
Solution Approach 2:
The patent transforms the raw performance data from its original numerical values into a different parameter space - specifically, presence/absence binary indicators and absence length measurements. This parameter transformation simplifies the subsequent analysis by converting a continuous data problem into a discrete classification problem that is computationally more efficient.
Data Source
AI summary
Gaps in performance data are corrected for through data transformations and conversion. A raw sequence is transformed by correction logic into an interval sequence by partitioning a performance monitoring period into equal intervals and assigning values based on the raw sequence. Locality sequence entries can indicate whether the interval sequence relies on estimation. The interval sequence is converted into an absence length sequence whose entries indicate null value periods in performance data. Conversion includes generating a presence sequence from the interval sequence, and deriving the absence length sequence from the presence sequence, by using a set-based algorithm or other mechanism. Excessive absence length values support treating intervals as downtime for the machine. Correction logic may include a stored procedure residing in a database, for example, which produces the absence length sequence without using a procedural language.