Bottleneck detector for executing applications
View Patent ↗A bottleneck detector may analyze individual workloads processed by an application by logging times when the workload may be processed at different checkpoints in the application. For each checkpoint, a curve fitting algorithm may be applied, and the fitted curves may be compared between different checkpoints to identify bottlenecks or other poorly performing sections of the application. A real time implementation of a detection system may compare newly captured data points against historical curves to detect a shift in the curve, which may indicate a bottleneck. In some cases, the fitted curves from neighboring checkpoints may be compared to identify sections of the application that may be a bottleneck. An automated system may apply one set of checkpoints in an application, identify an area for further investigation, and apply a second set of checkpoints in the identified area. Such a system may recursively search for bottlenecks in an executing application.
1. A system comprising:
a processor;
a bottleneck analyzer that:
identifies a workload processed by an application, said workload having multiple instances;
captures timestamps each time a workload instance encounters each of a plurality of checkpoints in said application; and
for a first checkpoint, curve fits said timestamps across a plurality of workload instances to generate a first historical curve;
for a second checkpoint, curve fitting said timestamps across said plurality of workload instances to generate a second historical curve; and
compares said first historical curve to said second historical curve to identify a difference between said first historical curve and said second historical curve; and
determines that said difference indicates a bottleneck.
2. The system of claim 1 , said first historical curve having a first correlation coefficient, said second historical curve having a second correlation coefficient, and said difference comprising a difference between said first correlation coefficient and said second correlation coefficient.
3. The system of claim 2 , said first historical curve and said second historical curve both being linear curve fitting.
4. The system of claim 1 , said first historical curve having a first slope and said second historical curve having a second slope, said difference comprising a difference between said first slope and said second slope.
5. The system of claim 1 , said first historical curve and said second historical curve being a function of load.
6. The system of claim 1 , said workload being a process.
7. The system of claim 1 , said workload being a thread.
8. The system of claim 1 , said timestamps being relative time difference from a baseline timestamp.
9. The system of claim 8 , said baseline timestamp being a starting timestamp for said workload.
10. The system of claim 8 , said baseline timestamp being a previous timestamp of a previous checkpoint.