Merging Sample-Based Profiling Data via Normalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing performance profiling modules face challenges in merging sample-based profiling data from multiple sessions, as they often require normalization to provide semantically meaningful presentations, especially when using sampling techniques that involve time-multiplexed hardware resource sharing.
Innovation Solution
The method involves calculating total sample counts, converting them to extrapolated event counts using specific formulas, and then merging these counts to determine the interval and active fraction for the merged database, allowing for the creation of a comprehensive performance overview of a computer system over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sampling techniques are used to monitor events, then hardware resource sharing is enabled and monitoring capacity is increased, but the data requires normalization and complex merging procedures to provide semantically meaningful presentations
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing normalization factors (active fraction and interval values) during data collection. This allows the merging process to simply retrieve and apply these pre-computed values rather than performing complex normalization calculations during merging, thus reducing the complexity of the data merging operation while maintaining the high monitoring capacity enabled by sampling techniques.
2Loss of information
If multiple monitoring sessions with varying time durations are conducted, then comprehensive performance coverage is achieved, but the merging of data sets becomes complex to ensure semantically meaningful presentations
Solution Approach 1:
The patent applies parameter changes by transforming the raw sampling data into normalized parameters (active fraction and interval) that are session-independent. This transformation allows data from monitoring sessions of varying time durations to be merged by simple addition of the normalized parameters, eliminating the complexity that would otherwise arise from attempting to merge raw sampling data directly while preserving comprehensive performance coverage.
Data Source
AI summary
In one embodiment, a computer system comprises one or more processors, and a memory module communicatively connected to the one or more processors. The memory module comprises logic instructions stored on a computer readable medium which, when executed on the one or more processors configure the one or more processors to normalize data from a first data set comprising sample based profiling data to form a first normalized data set, normalize data from a second data set comprising sample based profiling data to form a second normalized data set, combine the first normalized data set and the second normalized data set to form a merged data set, and present a portion of the merged data set on a user interface.