Why are the load averages high, while there is hardly CPU utilization? How can I discover processes with a memory leakage? How can I avoid that they harm the overall system performance? Can poor response times be caused by zombie processes? What is the current size of the dynamically scaling page cache? Can I avoid aggressive behavior of my page cache? What is the real impact when I modify the nice value of a process? How can I avoid that all bandwidth of my network connection is used for bulk transfer at the expense of the interactive transfers? Which process will be killed when my system runs out of memory and swap space? How can I protect my critical processes against out-of-memory killing? How can I avoid that batch processes consume too much disk bandwidth at the expense of my interactive processes? What are scheduling policies? How can I run batch processes that only consume the leftovers of the processor capacity? Why is the kernel swapping out pages, while there is sufficient memory free?
As a Linux system administrator, lots of performance-related questions may cross your mind. These questions (and many more) will be answered in the master class Linux performance analysis and tuning (description in Dutch). The master class is lectured by Gerlof Langeveld, the creator/maintainer of atop, atopsar and the kernel module netatop. Gerlof is trainer/consultant at AT Computing (website in Dutch), a company that provides professional Linux training and consultancy.
The master class provides an insight into the methods and techniques to measure the performance of a Linux system, to interpret the results, and to implement improvements. It is targeted at experienced Linux system administrators
and takes 4 days.
During the master class, the hardware components are discussed that are most likely involved in the loss of performance by overloading: CPUs, disks, memory and network connections. In the theoretical lectures, you will learn how the Linux kernel deals with the capacity of these hardware components (CPU scheduling, disk scheduling, caching, memory management, swapping, and many other kernel mechanisms) and how this background knowledge correlates to the counters shown by atop and other relevant measurement tools. Furthermore, the training covers mechanisms like control groups that can be used to reallocate the bandwidth of these hardware components (see detailed technical content of the course).
This master class is regularly scheduled in one of the class rooms of AT Computing in Utrecht, The Netherlands. However, it is also possible to organize the master class as an on-site training at your location (english spoken with english documentation).