mpolednik.github.io virtualization & tech blog

DevConf.cz 18

Introduction

New year and January are the two keywords that mark the annual Developer Conference (DevConf.cz) by Red Hat in Brno, Czech Republic. I’ve been given an opportunity to present (alongside great speaker from SPICE, Christophe de Dinechin) a topic that is close to me - GPUs and virtualization. The talk was called “Slicing a (v)GPU”, as figuring how single GPU can be used my multiple VMs was the main topic. The recording is available at youtube.

continue reading

Open Source Summit EU & KVM Forum '17

Introduction

October was the month of Open Source Summit (LinuxCon, CloudOpen and many more events under one umbrella) and KVM Forum. This year, I’m thankful to have been given the opportunity to present talk “Helping Users Maximize VM Performance” at the shared track of OSSummit and KVM Forum! Both slides and recording are public!

continue reading

vGPU in oVirt

Introduction

In oVirt 4.1.4, we added a new feature called vGPU passthrough. vGPU is a technology that allows us to “shard” a physical GPU (GRID capable, like NVIDIA Tesla M60) into number of smaller instances. Each instance can then be assigned to a VM, giving us the ability to run GPU-accelerated workloads. There is quite a bit of setup required to get the vGPU instance into the VM, and this is where oVirt comes to picture: it requires one time (per hypervisor and per VM) setup, but after that the vGPU creation and deletion is happening seamlessly. This post serves as a step-by-step guide to getting the vGPU feature up and running.

continue reading

cpuflags, oVirt and vCPU features

Introduction

A new hook is in the oVirt ecosystem: cpuflags. The cpuflags hook is a small but handy addition that allows fine tuning of vCPU characteristics. More specifically, oVirt users and administrators are now able to select which CPU flags are exposed to guest operating systems.

continue reading

Advanced Ansible for Development Infrastructure

Introduction

There is more to Ansible than just a plain playbook that describes the desired state of a bunch of machines. Let’s look at how it’s possible to reuse Ansible tasks and playbooks by taking advantage of advanced tools at our disposal.

continue reading