Thread Dump Profiling for Cloud Service Bottleneck Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated test systems for cloud-hosted business transaction management and analytics services do not effectively test for performance or efficiency, especially with increasing numbers of tenants, transactions, and dataset sizes, leading to a need for optimized resource utilization and execution configurations.
Innovation Solution
A system that profiles performance by analyzing thread dumps to derive a relational model, using thread intensity as a proxy for latency and resource utilization, and provides insights through a dashboard for optimizing code execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If automated test systems are used for testing code execution, then testing coverage is improved, but performance testing capability deteriorates
Solution Approach 1:
The patent introduces thread dumps as an intermediary mechanism that captures execution state information without interfering with code performance. These thread dumps serve as a mediator between the testing system and the actual code execution, enabling performance analysis through collected data rather than direct measurement, thus resolving the contradiction between comprehensive testing and accurate performance measurement
Solution Approach 2:
The system creates copies of execution state information through thread dumps instead of directly measuring performance during execution. By copying and analyzing thread dump data, the system can assess performance characteristics without the measurement process itself impacting the actual code performance, thereby maintaining both testing coverage and measurement accuracy
2Measurement precision
If detailed performance profiling is implemented, then bottleneck identification is improved, but system overhead increases
Solution Approach 1:
The patent applies partial action by collecting only the specific thread dump information needed for performance analysis rather than comprehensive system state monitoring. By selectively capturing thread execution states and analyzing only relevant portions, the system achieves sufficient bottleneck identification precision while minimizing the overhead associated with data collection and processing
Solution Approach 2:
The system uses lightweight thread dump objects that are created temporarily for analysis and then discarded. These disposable thread dump structures require minimal memory and processing resources compared to persistent profiling mechanisms, enabling detailed performance analysis with reduced system overhead. The temporary nature of these objects allows comprehensive profiling without long-term resource consumption
Data Source
AI summary
A method, system, and device for profiling execution of a set of code is disclosed. The method includes obtaining information associated with a plurality of thread dumps from execution of a set of code; determining, based at least in part on one or more dimensions, one or more relationships between tasks processed during execution of the set of the code; and providing the one or more relationships.