IBM Support

Optimizing a PowerKVM guest using PCI passthrough

Technical Blog Post


Abstract

Optimizing a PowerKVM guest using PCI passthrough

Body

This article describes a method to optimize a PowerKVM guest to minimize NUMA effects when dedicating PCI adapters using PCI passthrough to optimize performance and reduce CPU overhead to achieve near baremetal performance on a POWER8 system.

KVM guest domains configured with dedicated high speed I/O adapters through PCI passthrough can benefit from optimizing location of CPU and memory in the host. This tutorial shows how to add commands to the domain xml file to pin CPU and memory to a specific NUMA node and provides a comparsion of network throughput between untuned and tuned throughput measured by iperf. For the example workload and server configuration, a gain of 30% throughput was achieved by NUMA tuning.

[{"Business Unit":{"code":"BU054","label":"Systems w\/TPS"},"Product":{"code":"HW1W1","label":"Power ->PowerLinux"},"Component":"","Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"","Edition":"","Line of Business":{"code":"","label":""}}]

UID

ibm16170841