Introduction

Performance Co-Pilot¹ (PCP) is a toolset for monitoring the status and activity of distributed systems. PCP can monitor, in near-realtime, a great variety of operational metrics from the target machines' operating systems and other running services. These tools help ensure effective operations by facilitating anomaly detection, fault remediation, and capacity planning. With PCP, you can: 1) see how your system is working, 2) identify when it is broken, and 3) optimize future operations.

Components

The Open Source distribution of Performance Co-Pilot includes:

• target-side plug-ins that relay real-time metrics from services such as Apache, Sendmail, and MySQL, as well as from the target machine's underlying operating system. Each plug-ins provides a viewport to the inner workings of a a particular system component.
• several client-side utilities, e.g. 1) a logger for archiving metrics, and 2) an inference engine that monitors metrics and triggers user-defined actions, such as sending alerts or initiating a failover. These client utilities enable operators to collect data, analyze it, and take action.
• APIs for 1) creating client-side utilities, and 2) creating target-side plug-ins to monitor new services. With these APIs, PCP can be easily extended to interoperate with your custom system.

And, with PCPMon² - an Open Source PCP data graphing application - you can graph PCP metric streams in real-time, as well as replay metric archives.

Architecture

PCP has a thoroughly distributed architecture that allows scattered clients to directly request metrics from target machines known to them. PCP requires no central aggregating server, which might become a bottleneck. Although, if desired, such an aggregating server might be constructed using PCP as a foundation.

Support Matrices