Using Linux tracing tools to analyse a LAMP stack

Tracing is a particular form of logging which collect information about frequent, low level events with a low impact on the mesured system. Each events contains various parameters and a precise timestamp.

This talk will present the LTTng project and demonstrate its use to analyse latency and other problems in a typical Linux webserver context. LTTng provides a Linux Kernel tracer and a userspace tracer (UST) with the corresponding control and visualisation tools. The kernel tracer can log information about a wide range of kernel events (Drivers, virtual memory, disk block i/o, scheduling, etc). The userspace tracer allow the instrumentation of any application with a simple API and the easy execution of the program to collect those events in a trace file.

You will learn how to instrument your program with examples in a web and database server, how to collect trace in several applications and in the kernel and what kind of analysis you can do with the collected information.