Take advantage of L3 Cache to prioritize Linux process. / Linux Plumbers Conference: Developing the Kernel, Libraries and Utilities

Take advantage of L3 Cache to prioritize Linux process.

This proposal has been rejected.

One Line Summary

Use the latest Intel® Xeon® feature to view and control over how shared resources such as last-level cache (LLC) is used by applications, virtual machines (VMs) and containers.

Abstract

The use of applications, virtual machines (VMs) and containers, in data centers is a demanding task nowadays. Linux helps to prioritize those task by using “cgroups”, which limits and isolates the resources to be used such as CPU, memory disk I/O, network, etc. A new feature introduced by Intel® Xeon® brings the capabilities to monitor and control the last-level cache (LLC) and memory bandwidth by CPU package. This feature is called Resource Director Technology (Intel® RDT).

A user space software was developed to manipulate the Intel® RDT features. In principle, this software programs the technologies via Model Specific Registers (MSR) on a per core or hardware thread basis. Clear Linux team has developed an orchestrator which monitor the system process and by using different policies choose and assign a percentage of LLC to a best candidate process.

This talk will show how to use the user space library and the orchestrator to prioritize Linux process and use efficiently the LLC.

Speaker

Miguel Bernal Marin
Intel
- Website: http://clearlinux.org/
Biography

Miguel is an engineer at Intel’s open source technology center. He is responsible for the kernel maintenance in the Clear Linux Project for Intel Architecture. Also, he teaches computer science in the Zapopan Institute of Technology. He has a Ph.D. in robotics and computer vision. He is a Linux developer enthusiast and has presented in some technical and scientific conferences.