The origins of many of the concepts and assumptions embedded in PCP go back to earlier performance work that began almost 50 years ago. In the interim, the speed and complexity of systems has grown at least as fast as Moore's Law would have predicted, roughly 2**25 or more than 30 million times bigger and faster.
This presentation will provide some historical context for the genesis of PCP, then review the major design milestones in the evolution of PCP, with commentary on what worked, what was a mistake and what might be done differently if starting again. The true meaning of the original "co-pilot" analogy will be explained and we'll review how well that objective has been met.
Concluding comments will consider some crystal ball projections for future performance analysis scenarios.
This talk covers the current state of the integration between Grafana and PCP, and the rationale behind our increased focus on Grafana and scaling PCP.
The challenges in matching PCP time series to Grafana data models will be explored, with an introduction to using Redis as a scalable cache for historical PCP data.
Live demonstrations and discussion of future work for both PCP and Grafana projects will be included.
Findings from my first foray into PCP - updating the lmsensors metrics (PMDA) code.
The motivation for making these changes will be discussed. I will cover the various implementation stages, documentation, aids for new developers and ideas for further improvement.
Lukas Berk, Christian Horn
Open discussion soliciting ideas and RFEs in general from all attendees regarding initially the power management metrics, but moving into discussion of improving the experience for new PCP hackers and finally any more general needs or ideas others have for PCP development.
Removing process barriers before people contributing to PCP and PCP maintainers alike has been a goal of mine for some time. There are places where we could streamline processes in terms of time and accuracy through automation, and also remove single-points-of-failure making PCP a more resilient project overall.
This talk will also cover my recent findings while using Vagrant-based VMs to address:
An overview of use of PCP tools in the Red Hat customer support context, by worked example.
This talk will also cover the way in which the the PCP pidstat utility was extended to add in missing process argument informations, and the first-time experience of adding code to PCP.
Aconex is a cloud platform for connecting people involved in projects in construction and engineering. Aconex has seen significant value from its PCP deployment, and developed the Parfait project to efficiently interface Java web applications to PCP.
This talk will explore the mechanisms through which Aconex engineers added and utilize per-request tracing with their production web applications.
Features and improvements are being added to PCP all the time. This session will go through many of the features recently added in the past year, and then go onto covering the short term roadmap of active projects.
Demonstrations of features from some newer tools, REST APIs, and kernel metrics will be presented.
To close the day we will hold an open ended discussion. This session will be available for everyone to bring up any topics or directions you would like to see PCP take.
We will also hold a quick retrospective on the conference itself, where attendees and speakers can raise anything they would like to see done differently, or kept the same in future PCP conferences.
Locations for next years conference will also be open for discussion!