20 Questions
What type of data does PerfSpect collect?
Performance data
True or false: PerfSpect is a programming language.
False
What is the similarity analyzer used for?
To compare multiple workloads based on telemetry data
True or false: The online phase of PerfSpect collects application performance metrics.
False
How many default files are generated during the offline post-processing phase?
3
What can PerfSpect be used for?
To solve a problem that was causing a system to have a lower latency
What can PerfSpect be used for?
Generating insights into applications
True or false: The Metric_out_average file is used to analyze results and deduce insights.
True
What does PerfSpect do before being released to the public?
Validated internally
What type of hardware does PerfSpect use to collect PMU data?
Intel hardware
What are workloads that are more frontend or backend bound clustered in?
Lower left and lower right quadrants
True or false: The 3rd Gen Intel® Xeon® Scalable processor has four general programmable counters.
False
What is the Umask filter used for?
Reducing the number of groups needed
True or false: The Umask filter was extended to 64-bit in the 4th Gen Intel® Xeon® Scalable processor.
False
What is PerfSpect?
A tool to collect performance metrics from Intel hardware
How many fixed counters does the 3rd Gen Intel® Xeon® Scalable processor have?
4
What is the most used file to analyze results and deduce insights?
Metric_out_average
What is the purpose of PerfSpect?
To collect performance data
What does PerfSpect require for use?
Nothing
What type of processor does the Umask filter extend to?
64-bit
Study Notes
- PerfSpect is a telemetry tool that collects performance data from your workload or application on your production servers.
- PerfSpect is lightweight and can be used without any need for programming.
- PerfSpect uses perf drivers to collect PMU data from Intel® hardware.
- PerfSpect can be used to characterize your workload or application and identify potential bottlenecks.
- PerfSpect can be used to post-process raw telemetry data to generate insights into your application's performance.
- PerfSpect is a performance analysis tool that collects raw telemetry data from the processor microarchitecture.
- The online phase collects system stats, and the offline post-processing phase converts the raw telemetry data to more readable performance metrics.
- There are three default files generated during the offline post-processing phase: Metric_out is a time series dump of all the metrics calculated, Metric_out_average is the most used file to analyze results and deduce insights, and Metric_out_p95 is a 95th percentile file to analyze results.
- The 3rd Gen Intel® Xeon® Scalable processor has four fixed counters, which frees up the scarce general programmable counters to reduce the number of groups needed to collect telemetry for a certain metric.
- The Umask filter was extended to 64-bit in the 3rd Gen Intel® Xeon® Scalable processor.
Learn about PerfSpect, a lightweight telemetry tool used for performance analysis of workloads and applications on production servers. Discover how it collects raw telemetry data and post-processes it to generate insights into application performance.
Make Your Own Quizzes and Flashcards
Convert your notes into interactive study material.
Get started for free