Loading…
KVM Forum 2018 has ended
October 24-26, 2018 - Edinburgh, Scotland, UK
Click Here For Information & Registration

Sign up or log in to bookmark your favorites and sync them to your phone or calendar.

KVM Forum Track 1 [clear filter]
Thursday, October 25
 

09:15

Private vs. Common: Reflection on Cross-architecture Commonality - Christian Borntraeger, IBM
While starting out as an x86 hypervisor, the KVM module supports a variety of architectures. Over time a split between common code and per-architecture code has evolved. This talk will reflect on history, current state and potential future improvement regarding commonality.

Speakers
avatar for Christian Borntraeger

Christian Borntraeger

CPO Linux on IBM Z Development, IBM



Thursday October 25, 2018 09:15 - 09:45
Sidlaw Auditorium

09:45

Contributor Q&A Panel - Moderated by Stefan Hajnoczi, Red Hat
A technical Q&A panel discussion ranging on various KVM and QEMU-related topics. The discussion will be for about an hour. Topics will be chosen on the spot from a prepared list. And audience (live or remote) can add questions in the live Etherpad during the discussion.

Speakers
avatar for Christian Borntraeger

Christian Borntraeger

CPO Linux on IBM Z Development, IBM
PF

Peter Feiner

Software Engineer, Google
Peter has worked on virtualization at Google since 2014, focusing on memory management, performance and scaling.
avatar for Stefan Hajnoczi

Stefan Hajnoczi

Senior Principle Software Engineer, Red Hat
Stefan has been active in QEMU since 2010 and works in Red Hat's virtualization team with a focus on storage. He works on virtio drivers in Linux and helps maintain the block layer and tracing in QEMU. He also organizes and mentors in the Google Summer of Code and Outreachy internship... Read More →
CH

Cornelia Huck

Senior Software Engineer, Red Hat
Cornelia is a Senior Software Engineer at Red Hat, working mainly on virtualization and s390x related topics in QEMU and the Linux kernel. She has been acting as a maintainer for s390x virtualization topics in QEMU and the Linux kernel for several years.Previously, she has given talks... Read More →
avatar for Wanpeng Li

Wanpeng Li

Linux Kernel Contributor, Tencent Cloud
Wanpeng Li is a 7 years experienced Linux kernel/virtualization developer who works in Tencent Cloud currently. He mainly focus on KVM, scheduler and memory management. In KVM, he contributes a lot of features to improve performance and stability. He has experience worked in IBM LTC... Read More →
PM

Peter Maydell

Principal Software Engineer, Arm
Peter has been handling all things Arm in QEMU since 2010, including CPU architecture emulation, support for KVM virtualization and herding an ever-increasing number of board, SoC and device models


Thursday October 25, 2018 09:45 - 10:30
Sidlaw Auditorium

11:00

Intel SGX Virtualization on Linux and KVM - Sean Christopherson, Intel
Intel Software Guard Extensions (SGX) provides new hardware-based capabilities to software developers to protect an application's secrets. SGX allows for the creation of a trusted environment without drawing any dependencies on the integrity of code elsewhere in the software stack, which in modern cloud environments may be millions of lines of code across the application, kernel and VMM.

This talk will provide a brief overview of SGX and its (nefarious) history with respect to Linux, followed by an in-depth analysis of the design and status of SGX virtualization support in KVM.

Speakers
avatar for Sean Christopherson

Sean Christopherson

Software Engineer, Intel
Sean is a recent convert to Linux and KVM, having spent the first 10+ years of his career developing Intel-internal software used to validate the functional behavior of Intel processors.Sean's current foci are enabling Intel features in KVM and the never-ending saga of upstreaming... Read More →



Thursday October 25, 2018 11:00 - 11:30
Sidlaw Auditorium

11:30

Protect Data of Virtual Machines with Memory Encryption on KVM - Kai Huang, Intel Corporation
Today cloud data protection is a critical requirement, and it will be even more important in the future as we have more in-depth and sensitive data in the cloud for new types of workloads (such as IoT and machine learning). Since VM is the key container of such data, it is crucial to protect VM at-rest (as in storage), in-transit (as in network), and during execution. Encryption is considered as the foundation technology for VM protection, and there are established encryption technologies for VMs at-rest and in-transit. Intel® Multi-Key Total Memory Encryption (MKTME) is Intel's new memory encryption technology which supports VM memory encryption to complete VM protection in VM's entire lifecycle. In this presentation we give an introduction to Intel® MKTME, from background, to its hardware architecture, typical use cases in cloud, and our enabling work on Linux/KVM.

Speakers
avatar for Kai Huang

Kai Huang

Software Engineer, Intel Corporation
Kai Huang @ Intel Corporation



Thursday October 25, 2018 11:30 - 12:00
Sidlaw Auditorium

12:00

Extending Secure Encrypted Virtualization with SEV-ES - Thomas Lendacky, AMD
AMD has enhanced the capabilities of its hardware-based memory encryption to further extend the security protections provided by Secure Encrypted Virtualization (SEV). This talk will focus on the Encrypted State (ES) technology which extends the SEV support to minimize exposure of the guest register state for added isolation and protection from the hypervisor. The presentation will discuss the technical details of this technology with a focus on how it is being integrated into KVM.

Speakers
TL

Thomas Lendacky

PMTS Software Engineer, AMD
Tom Lendacky is a member of the Linux OS group at Advanced Micro Devices. He is responsible for enabling and enhancing support for AMD processor features in the Linux kernel. He is currently working on extending the SEV support to enable SEV-ES (Secure Encrypted Virtualization - Encrypted... Read More →


Thursday October 25, 2018 12:00 - 12:30
Sidlaw Auditorium

13:45

Performance Optimization on Huawei Public and Private Cloud - Lei Gong & Jinsong Liu, Huawei
With the increasing demand for big data processing and faster memory databases, such as SAP HANA, the demand for large-size virtual machines is getting stronger. Meanwhile, for enterprise virtualization and private cloud scenarios (such as VSI), virtual machine density improvement is also an important means of saving cost, that is, CPU over-commitment.
In this session, we will share some optimization ways in virtualization on the KVM platform for public and/or private cloud. The first one is for the optimization of the spinlock holder of the guest, so that we can improve the synchronization efficiency between the VCPUs of the large-scale virtual machine. The second one is the "balance scheduler", as far as possible to ensure that different VCPUs of the same virtual machine can be scheduled at the same time to improve performance. The last one is optimization of the RTC clock compensation scheme is moved from userspace to kernel to reduce overhead of context switching and increase the density of Windows virtual machines in the private cloud.

Speakers
LG

Lei Gong

HUAWEI


Thursday October 25, 2018 13:45 - 14:15
Sidlaw Auditorium

14:15

Demand Paging Performance with 160 vCPUs - Cannon Matthews, Google
As physical machine memory sizes and CPU counts rise, so does the demand for large virtual workloads as well. One particular challenge of scaling large VMs is migrating them to new hosts for maintenance. This talk will look at the performance of the demand paging mechanism userfaultfd and how well it performs under stress, in particular how it scales with large number of vCPUs. Then, the talk will examine the guest memory and CPU performance during a demand paging migration of a VM on a 4-socket host with 160 vCPUs, and additionally compare userfaultfd to an alternate approach.

Speakers
CM

Cannon Matthews

Software Engineer, Google
Cannon is a Software Engineer working for Google Cloud on virtualization for Google Compute Engine.


Thursday October 25, 2018 14:15 - 14:45
Sidlaw Auditorium

14:45

Torwards a More Scalable KVM Hypervisor - Wanpeng Li, Tencent Cloud
Torwards a more scalable KVM hypervisor(Wanpeng Li) - The kvm hypervisor is at the core of cloud computing, Virtual Machine based approaches to workload consolidation, as seen in IaaS cloud as well as datacenter platforms, have long had to contend with performance degradation caused by synchronization primitives inside the guest environments. These primitives can be affected by virtual CPU preemptions by the host scheduler that can introduce delays that are orders of magnitude longer than those primitives were designed for. In this presentation, Wanpeng Li will introduce Paravirtual TLB Shootdown which can mitigate TLB Shootdown Preemption issue. In addition, he will also introduce Exitless IPIs to scale large VMs, and disable mwait/hlt/pause vmexits to improve latency for dedicated instances.

Speakers
avatar for Wanpeng Li

Wanpeng Li

Linux Kernel Contributor, Tencent Cloud
Wanpeng Li is a 7 years experienced Linux kernel/virtualization developer who works in Tencent Cloud currently. He mainly focus on KVM, scheduler and memory management. In KVM, he contributes a lot of features to improve performance and stability. He has experience worked in IBM LTC... Read More →



Thursday October 25, 2018 14:45 - 15:15
Sidlaw Auditorium

15:45

CloudAtlas: Ways to Make Live Migration Safe and Expectable - Chao Zhang & Feng Xie, Alibaba Cloud
The public cloud users, especially enterprise users, are increasingly concerned about the stability of the infrastructure provided, therefore the performance impact of the live migration itself can no longer be ignored. In order to perform migration, the following needs to be considered: 1)How to classify the workload and evaluate the performance impact. 2)When should we migrate to reduce the impact to a minimum. 3)How to select params of the migration to balance the resource consumption and migrate speed. The ClouAtlas we build rely on KVM hypervisor to collect data like CPU usage, memory/TLB access and IO. Based on a statistical model and Deep Learning Algorithm, we classify the workload of the VM, and predict proper time to perform migration. After tens of thousands of tests, our system shows an obvious alleviation of the migration impact and a boost of the migration performance.

Speakers
FX

Feng Xie

technical expert, Alibaba Cloud
Feng has been working on virtualization since 2012. He's worked on several areas such as GPU virtualizaton and ARM virtualization. Now Feng is employed by Alibaba Cloud, and is focus on Artificial Intelligence Operation and Maintenance based on Machine Learning.
CZ

Chao Zhang

technical expert, Alibaba Cloud
Chao, as a initiated participant of the Live migration team at Alibaba Cloud, has completely witnessed how this technology implanted itself into the Ali-Cloud system along a tough way. Before join Alibaba, Chao is a kernel developer engineer working at Huawei EulerLinux team for... Read More →


Thursday October 25, 2018 15:45 - 16:15
Sidlaw Auditorium

16:15

Distributed QEMU - Yubin Chen & Zhuocheng Ding, Shanghai Jiao Tong University
It is difficult for a single physical machine to satisfy applications which require massive resources such as big data analysis, while a virtual machine spanning multiple physical nodes can do it. In this presentation, Yubin will demonstrate their work of distributed QEMU which allows a guest OS running on single virtual machine to take advantage of resources (CPU, memory, I/O, etc.) from multiple physical nodes. They extend QEMU to enable forwarding of PIO, MMIO, Interrupts so that vCPUs and I/O devices on different machines can communicate with each other. Furthermore, in order to run a cross-node virtual machine, they also have implemented an RDMA-based distributed shared memory by extending the access control of EPT, which makes it possible for aggregating physical memory transparently. The preliminary experiment shows near-native performance for NUMA-friendly benchmarks on sv6 OS.

Speakers
YC

Yubin Chen

Researcher, Shanghai Jiao Tong University
Yubin Chen is a second-year graduate student researcher at School of Software, Shanghai Jiao Tong University. His research focuses on building distributed system by taking advantage of virtualization. The main topics of interests are QEMU/KVM, containers, and distributed systems... Read More →
ZD

Zhuocheng Ding

Researcher, Shanghai Jiao Tong University
Zhuocheng Ding is a second-year graduate student researcher at School of Software, Shanghai Jiao Tong University. His research interests include parallel and high-performance computing, system virtualization, distributed system and system security, and currently focuses on building... Read More →


Thursday October 25, 2018 16:15 - 16:45
Sidlaw Auditorium

16:45

Adaptive Live Migration - Xiao Guangrong & Yulei Zhang, Tencent Cloud
Pre-copy migration could fail for many cases, e.g, if there are memory intensive workloads in VM. Fortunately, QEMU/KVM gains some features to improve it, however these features require the user need to pre-know the workloads in VM, e.g, enable compression only if the data is compressible and system has enough resource to do compression that is not friendly to the public cloud providers. Post-copy migration improves the situations indeed, however, it suffers some shortcomings, e.g, it is unrecoverable, poor performance, etc.

We will present adaptive live migration which speculates VM's workload from host side then enables pre-copy features and adjusts its parameters dynamically, during live migration, it detects the tendency and try next feature if current tendency shows live migration is impossible to success. Try post-copy if we have used all ways but pre-copy still can not success.

Speakers
XG

Xiao Guangrong

Senior Software Engineer, Tencent Cloud
Xiao Guangrong is a Linux Kernel developer working on Ftrace, MM, Btrfs but his main interest is KVM. As a active contributor, he was invited to give some presentations at some conferences: Japan LinuxCon 2011, Japan LinuxCon 2012 China CLK 2012, KVM Forum 2016, 2017, 2018. He is... Read More →
ZY

Zhang Yulei

Senior Software Engineer, Tencent Cloud
More than 10 years experienced software developer working in Virtualization area. Used to design and implement Intel GPU virtualization technology(a.k.a Intel GVT-g). Recent presentation was: as technique speaker presented "Enable media cloud with intel Graphics virtualization technology... Read More →


Thursday October 25, 2018 16:45 - 17:15
Sidlaw Auditorium
 
Friday, October 26
 

09:30

Arm Timers; and Fire! - Christoffer Dall, Arm
A key feature that is missing in KVM/ARM compared to x86 is the support for stolen time reporting to virtual CPUs. Stolen time allows a virtual CPU to make more informed scheduling decisions on oversubscribed systems by factoring in CPU time used by the hypervisor for other tasks instead of running the virtual CPU. The Arm architecture defines two separate timers: The virtual and physical timers, which are both available to the OS, and further provides separate timers for the hypervisor. While the intention behind these architecture features is to provide OSes a mechanism for telling physical and virtual time apart, this does not work in practice. This talk explains why we cannot easily use the architected mechanism to provide stolen time on Arm, and why we must instead use a paravirtualized solution based on a shared page and how this approach differs slightly from the one used on x86.

Speakers
avatar for Christoffer Dall

Christoffer Dall

Principal Engineer, Arm
Christoffer Dall is principal engineer in Arm's architecture and technology group. Christoffer is a long-time open source contributor, and original author of KVM/ARM. Christoffer has a PhD in computer science from Columbia University, specializing in operating systems and virtualization... Read More →


Friday October 26, 2018 09:30 - 10:00
Sidlaw Auditorium

10:00

ARM virt 3.0 and Beyond: Towards a Better Scalability - Eric Auger, Red Hat
We start to see AArch64 servers with huge amount of physical processors, memory and IO devices. Virtual machines must scale with this trend. The ARM virt machine model had limitations with respect to virtual resource sizing. This talk aims at describing recent changes and work ongoing to achieve a better scalability.

Improvements related to vcpu, PCIe bus and memory sizing will be detailed. The talk will cover the support of larger guest physical address space and extension of existing RAM with DIMM slots. While reconsidering the RAM partitioning, we will discuss the challenge of assigned device IOVA reserved regions.

This is also an opportunity to draw a comparison between the ARM virt machine model and the PC Q35 machine model, in terms of virtual resources and vertical scalability. What bridges are left to cross to get feature parity?

Speakers
EA

Eric Auger

software engineer, Red Hat
Eric has been involved in KVM/QEMU since March 2014. He started as a Linaro assignee and currently works for Red Hat, in the Virtualization team. Eric has contributed to VFIO, KVM/ARM and QEMU on topics such as platform/PCIe device assignment, virtual interrupt controllers, added... Read More →


Friday October 26, 2018 10:00 - 10:30
Sidlaw Auditorium

11:00

L1TF and KVM - Alexander Graf, SUSE
Recently a new speculative execution side channel was unvealed, which
could potentially result in leakage of arbitrary memory contents into
unprivileged virtual machines on most recent Intel CPUs. This
presentation will give insights as to what the L1 Terminal Fault (L1TF)
Spectre vulnerability is. It will show how it can be exploited and based
on that knowledge it will take a look at how KVM mitigates those
issues. It will also show performance penalties these mitigations incur.

On top of that, the presentation will present an alternative work in
progress approach to mitigate L1TF that may recover some of the
performance penalties by leveraging unrelated CPU features.


Speakers
avatar for Alexander Graf

Alexander Graf

Principal Software Engineer, SUSE :)
Alexander started working for SUSE about 10 years ago. Since then he worked on fancy things like SUSE Studio, QEMU, KVM, openSUSE and SLES on ARM and U-Boot. Whenever something really useful comes to his mind, he tends to implement it. Among others he did Mac OS X virtualization using... Read More →


Friday October 26, 2018 11:00 - 11:30
Sidlaw Auditorium

11:30

What Did Spectre and Meltdown Teach about CPU Models? - Paolo Bonzini, Red Hat
During the past year, several architectural flaws in CPUs were disclosed, allowing user processes to read privileged memory or registers through cache side-channel attacks. For those flaws that were fixed through microcode updates, the effect of the updates may need to be visible in the guest, requiring changes in QEMU, libvirt and the rest of the virtualization stack. Even if no updated microcode was needed, as was the case for "Meltdown", the fixed kernel used a previously irrelevant CPU features in order to avoid an even more substantial performance drop.

All in all, these experiences showed that even higher-level management, such as OpenStack, need to be taught about the details of CPU configuration. This talk will explain how this conclusion was reached, what exactly is expected from management tools, and why.

Speakers
avatar for Paolo Bonzini

Paolo Bonzini

Sr. Principal Software Engineer, Red Hat, Inc.
Paolo is a long-time KVM contributor and co-maintainer; he also co-maintains some subsystems in QEMU. He works at Red Hat.


Friday October 26, 2018 11:30 - 12:00
Sidlaw Auditorium

12:00

KVM/arm Meets the Villain: Mitigating Spectre - Marc Zyngier, ARM
The recent discovery of the Spectre and Meltdown vulnerabilities has
been an unfortunate disruption in the otherwise rapid pace of KVM/arm
development. Significant efforts have been put into analysing how CPU
implementations are affected, and in developing a mitigation framework
that covers both KVM and VMs by allowing VMs to call hypervisor
services to mitigate the vulnerabilities in the virtual machine.

This talk will briefly explain how the vulnerabilities affect ARM
systems and describe the range of techniques introduced in KVM/arm to
mitigate Spectre v2, v3, v3a and v4. We will explain how our
mitigation framework hides as much of the diversity of CPU
implementations and system integration while trying to keep the
overhead as low as possible.

Speakers
avatar for Marc Zyngier

Marc Zyngier

Kernel Nacker, ARM
Marc has been working on the Linux kernel since an unexpected encounter with 0.99pl13 in 1993. His first contribution was merged in 1996 in the form of the original version of the MD driver. Having played with fault tolerant systems at Bull, worked on exotic (and ultimately doomed... Read More →


Friday October 26, 2018 12:00 - 12:30
Sidlaw Auditorium

13:45

A Driver Framework for qtest - Laurent Vivier, Red Hat
Unit tests for QEMU's virtual device currently are built upon the glib test framework. In order to ease the implementation of tests, QEMU provides the qtest protocol as well as a library of helper modules, libqos. libqos provides helper code for common device kinds (USB, PCI or virtio), but test needs to know how to start QEMU (machine type and parameters).

This talk will present an extension to libqos that wraps the glib test framework and augments it with knowledge of QEMU machine types and devices. Based on a description of the machines and devices, and on generic code to access the device, libqos generates all possible combinations of machines and architectures, and takes care of creating a "device object" for use in the tests.

This work was sponsored by Google, as part of GSoC, with mentoring from Laurent Vivier and Paolo Bonzini of Red Hat.

Speakers
avatar for Laurent Vivier

Laurent Vivier

Senior Software Engineer, Red Hat
Laurent is contributing to QEMU and KVM since 2007, after a détour into the world of HPC, he has joined in 2015 the Red Hat Virtualization team, focusing his efforts on KVM for POWER.



Friday October 26, 2018 13:45 - 14:15
Sidlaw Auditorium

14:15

The Functional Test Beast: Tame it, Bring it Home and Make it your Pet - Cleber Rosa, Red Hat, Inc.
Functional tests have traditionally been kept outside. These poor beasts get beaten by heavy work, but are not allowed to sit in the couch by your side, nor are they given the freedom to run inside your $HOME.

There have been valid reasons for this separation, but what if a far more comprehensive (yet safe and manageable) functional testing environment could be added to the regular developer's work-flow, pretty much for free?

In this talk we'll present QEMU's brand new functional testing architecture and how it lets developers easily write and run advanced functional tests.

Because anything functional demands to be practical, we'll write a test for a real QEMU bug and its fix, running it in different targets and configurations from the very same development tree. Finally, we'll give the test a definitive home alongside that fix, while putting the bug on a permanent tight leash.

Speakers
CR

Cleber Rosa

Sr. Software Engineer, Red Hat
Cleber is a Software Engineer working for Red Hat on the virt team. He helps to develop the testing stack for virtualization technologies, such as KVM, QEMU and Libvirt. He is a maintainer of the Avocado test framework and sub projects such as Avocado-VT.


Friday October 26, 2018 14:15 - 14:45
Sidlaw Auditorium

14:45

Testing in QEMU and libvirt: Beyond Make and Make Check - Yash Mankad, Red Hat
Automated testing in the QEMU and libvirt communities is largely focused on running unit tests on various environments using the `make check` command, but has never developed beyond that into a centralized continuous integration environment. In this session, Yash Mankad will give an overview on the initiative undertaken by the Virtualization group at Red Hat to look beyond `make check`, by leveraging Ansible and the Avocado testing framework to expand automated test coverage executing across multiple architectures, operating systems, and testing environments. Yash will also share his ideas on integrating testing initiatives already in place, like patchew and travis CI, with this common platform. This initiative will result in QEMU and libvirt developers gaining three times the platform coverage while providing a central point for continuous integration for the QEMU and libvirt projects.

Speakers
YM

Yash Mankad

Software Engineer - Virtualization, CI, Automation, Red Hat
Yash works in the Virtualization group at Red Hat where he is leading the efforts to implement a continuous integration environment that implements multi-architecture testing for qemu-kvm, and libvirt on RHEL, Fedora, and CentOS. Yash has previously given a talk titled "Avocado and... Read More →


Friday October 26, 2018 14:45 - 15:15
Sidlaw Auditorium

15:45

Facilitating Incremental Backup - Eric Blake, Red Hat
As any good sysadmin will tell you, collecting data backups are essential for future recovery needs. But making them more efficient and flexible has been an ongoing quest, involving multiple improvements throughout the virtualization stack. In this talk, Eric Blake will give a demonstration of the newest technique available through the qemu+libvirt stack, which now support the ability to orchestrate online incremental backups where third-party clients can grab just the dirty clusters at their own speed, rather than waiting for qemu to push a full backup image to a destination location.

Speakers
avatar for Eric Blake

Eric Blake

Software Engineer, Red Hat
Eric Blake is a software engineer at Red Hat, working on block device management in virtualization. He has contributed extensively to qemu and libvirt. He has spoken in past KVM Forums, most recently about incremental backup through libvirt at KVM Forum Oct 2018.


intro cast
full cast
incr cast

Friday October 26, 2018 15:45 - 16:15
Sidlaw Auditorium

16:15

Qemu Backup Status - Vladimir Sementsov-Ogievskiy, Virtuozzo
Large work is underway in Qemu around backups, which involves dirty-bitmaps management, NBD protocol extensions and external backup API. The presentation will describe what was done and alredy available and what is upcoming.

Speakers
VS

Vladimir Sementosov-Ogievskiy

software developer, Virtuozzo
Vladimir works in Virtuozzo and deals in the main with Qemu block layer for more then 3 years, developing features around backup, qcow2 and NBD protocol. He also attended KVM Forum 2017 as a speaker with a talk about Qemu backup.


Friday October 26, 2018 16:15 - 16:45
Sidlaw Auditorium

16:45

Background Snapshots - Denis Plotnikov, Virtuozzo
The VM snapshot is a widely-used feature to save a VM state. Currently, when you make a snapshot you are forced to pause your virtual machine until the snapshot is done. This is undesirable behavior because it's seen for the user as a temporary VM shutdown which is something you don't want when your goal is to provide your end-users with good quality of service.
I would like to introduce background snapshots: the feature allowing not to pause your virtual machine when making a VM snapshot. It keeps your virtual machine up and running until the end of the snapshotting process. I would like to explain some technical details on how we have achieved that and very short overview on how to use the feature.

Speakers
avatar for Denis Plotnikov

Denis Plotnikov

Software developer, Virtuozzo
A member of Server Virtualization Team at Virtuozzo



Friday October 26, 2018 16:45 - 17:15
Sidlaw Auditorium