
System performance analysis
Showing 5 of 19 projects. Click any project card for scope, mentors, and proposal studio.
Mentors: Student: Shiyao
This project is to extend some timeseries functionalities based on the existing query languages. The tasks to be worked on are: 1. grammar extensions supporting scalar operands in expressions, either as literal constants or from functions returning a scalar result. 2. additional functions complementing existing functions to operate across both the time domain (with one scalar result per metric instance) and across instance domains (with a single vector result). 3. new statistical functions such as top_k(expr) and nth_percentile(expr, n) and rolling_avg(expr, n). 4. new functions for subsampling and interpolation of vector operands to match time series samples with other vector operands in an expression.
Mentors: Student: Sohaib Mohamed
Port BPF tools to Performance Co-Pilot PMDA and implement PCP/Htop DynamicTabs. The Htop utility recently acquired the ability to display multiple Tabs in its user interface. Currently, these display only process information. The aim of this project is to support generalized Tabs, where information beyond processes can be displayed in columns. The classic example of this would be being able to display cgroups on Linux.
Mentors: Student: Ashwin Nayak U
<p>This project involves advancing the performance analysis capabilities provided by the PCP integration within the pbench benchmarking and analysis framework. The problems to be worked on are:</p> <ol> <li>Modifying pbench to leverage pmlogger's remote collection capabilities and leverage PCP's archive compression feature for efficient storage.</li> <li>Enhancing pbench with the ability to enable live display of collected metrics via Grafana.</li> <li>Modify PCP to provide new pmcd agent metrics, pmlogger and pmlogconf templates tailored to recommended performance analysis data collection for target workloads (database, web server, computation, low-latency networking, etc.)</li> </ol>
Mentors: Student: Yushan Zhang
<p>The Performance Co-Pilot (PCP) timeseries are series of time-stamped values gathered from hosts making performance data available. Clients obtain target timeseries values by submitting timeseries query language to pmseries tool. This proposal is to expand five new types of timeseries operation functions based on the existing query language, includes extending the grammar of the query language, implementing the parse module corresponding to the extended grammar, and implementing the mathematical operation functions accordingly. The extended languages and the timeseries functions will still accessible via REST API.</p>
Mentors: Student: Miroslav Foltýn
<p>Goal of this project is to improve and build upon existing Grafana intergration for Performance Co-Pilot.</p> <ul> <li>Developing a <strong>reactive dashboard</strong> and using PCP metrics</li> <li>Implementing support for PCP <strong>derived metrics</strong> in the Vector data source, as described in the PCP REST API documentation</li> <li>Improving the <strong>metric search</strong> process by creating a new page for live, full-text search on PCP metric names, instances, instance domains and their descriptions, if any</li> </ul> <p>Based on performance metrics from PCP and a predefined rule set, the <strong>reactive dashboard</strong> should display only relevant metrics to the user. For example, if the CPU utilization is high, it should show an overview of CPU statistics and processes contributing load. The user should then be able to drill down into more specific areas of interest - e.g. present graphs about lock contention. For each recognized performance issue a link should be presented to the user on how to resolve the issue. In this way the dashboards guide users toward possible root cause of performance problems.</p>