Application Analyzer Partitioning for Isolated Software Profiling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing performance analysis tools in enterprise computing environments fail to isolate the effects of individual applications, leading to inadequate identification of misbehaving software that degrades system performance, causes user experience issues, and increases energy consumption, particularly in large-scale heterogeneous systems.
Innovation Solution
An application analyzer circuitry that includes application identifier, partitioning, experimentation, and analysis components to isolate and analyze applications under controlled conditions, measuring low-level processor and memory behaviors, and providing actionable results in a scalable and automated manner.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing performance analysis tools are used to monitor system performance, then general visibility into system load is provided, but the ability to isolate and identify individual misbehving applications is insufficient
Solution Approach 1:
The system segments the analysis function by dividing it into multiple specialized circuitry components: application identifier circuitry for selecting target applications, partitioning circuitry for isolating applications to specific processor partitions, experimentation circuitry for modifying execution conditions, and analyzer circuitry for measuring performance metrics. This segmentation enables precise application-level analysis without requiring complete system redesign.
Solution Approach 2:
The patent introduces an intermediary analysis system that sits between the application layer and hardware resources. This intermediary circuitry intercepts and measures performance metrics from applications without directly modifying application code or requiring changes to the underlying hardware, thereby achieving precise measurement while maintaining system compatibility.
2Measurement precision
If conventional monitoring techniques are used to track application behavior, then general system metrics are captured, but detailed low-level processor and memory behaviors remain undetected
Solution Approach 1:
The system extracts performance measurement functionality from the general monitoring infrastructure and implements dedicated hardware circuitry specifically for capturing low-level processor and memory behaviors. This extraction enables direct measurement of cache misses, branch predictions, and memory access patterns that are invisible to conventional software-based monitoring tools.
Solution Approach 2:
The patent replaces software-based monitoring mechanisms with hardware-level measurement circuitry. By using dedicated processor circuits and memory interface circuitry to directly measure low-level behaviors, the system achieves higher precision and lower overhead compared to software instrumentation approaches.
3Measurement precision
If aggregate performance measurements are used to identify problematic applications, then overall system load is visible, but differentiation between normal background activity and detrimental behavior is not achieved
Solution Approach 1:
The system applies local quality by measuring and analyzing performance metrics specifically for each application partition rather than relying on aggregate system-wide measurements. Each application's performance is evaluated against its own baseline and against defined thresholds, enabling precise differentiation between normal and detrimental behavior patterns for individual applications.
Solution Approach 2:
The analysis system performs self-service by automatically establishing performance baselines for each application, continuously monitoring against these baselines, and autonomously identifying misbehving applications without requiring manual configuration or intervention. The system self-manages the entire analysis workflow from baseline creation to anomaly detection.
4Productivity
If reactive troubleshooting methods are used to identify misbehving applications, then problem resolution is achieved after user complaints, but preventive identification and fleet-wide scalability are limited
Solution Approach 1:
The system performs preliminary action by proactively and continuously monitoring application performance across the enterprise fleet before problems manifest to users. Performance baselines are established in advance, and the system continuously compares current behavior against these baselines to detect and alert on misbehving applications before they cause user complaints or system failures.
Solution Approach 2:
The patent implements a universal analysis system that can be deployed across heterogeneous enterprise fleets with different hardware platforms and application workloads. The same circuitry and methodology are used to analyze performance across diverse devices, enabling fleet-wide scalability and consistent problem identification regardless of the specific application or hardware configuration.
Data Source
AI summary
This disclosure is related to software application analysis and, more particularly to, methods and apparatus to analyze software applications. An example non-transitory machine readable storage medium includes instructions to cause programmable circuitry to at least determine a first partition of a first processing resource, reallocate applications executing on the first partition, allocate an application under test to execute on the first partition, monitor execution of the application under test, and output a report associated with the execution.