Loading…
This event has ended. Create your own event → Check it out
This event has ended. Create your own
Don’t miss this great event - register now! 
View analytic

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

Wednesday, August 24
 

8:00am

10:30am

Coffee Break
Wednesday August 24, 2016 10:30am - 10:55am
Metropolitan Foyer

10:55am

Experience in Developing a Driver for a Complex New Hardware Device on Qemu/KVM - Knut Omang, Oracle
In this presentation Knut Omang will talk about how he has been deploying a patched version of QEMU/KVM as part of a framework to enable writing an almost functionally complete device driver for a complex new network device before any hardware was available. Different levels of simulation/emulation has different benefits and challenges. Detailed low level simulation of hardware is accurate but painfully slow. High level simulation/emulation is fast enough to run real workloads, but with somewhat lower precision. QEMU/KVM with device hot plugging allows recompiling/relinking/restarting different device models while keeping the guest(s) running. Even after availablity of hardware, the QEMU anchored models are valuable tools, as some testing can be run on the models, saving expensive server hardware, and also simplifying monitoring of traffic between device and driver.

Speakers
avatar for Knut Omang

Knut Omang

Principal Software Engineer, Oracle
Knut Omang has been using and championing the use of virtualization for testing for many years. He is currently working for Oracle as the lead developer of a Linux driver and user space library for Oracle's new Infiniband HCA. He has been an early user/adopter of the Q35/PCIe based Qemu machine type, where he also has contributed a few patches such as the initial work on a virtual Intel IOMMU and SR/IOV support. He holds a Ph.D in computer... Read More →


Wednesday August 24, 2016 10:55am - 11:45am
Harbour A

11:55am

Performant Security Hardening of KVM - Steve Rutherford, Google
Guest escapes and host information leaks in KVM are a causes for great concern. This talk covers a safer mode for KVM on x86 that is intended to reduce the frequency of such exploits, without decreasing performance. By removing complex, non-performance critical devices from KVM (namely, legacy interrupt controllers and the instruction emulator), the host kernel can expose less attack surface to the guest. This talk analyzes the guest exposed attack surface of KVM, as well as the performance and security implications of this new mode in production.

Speakers
SR

Steve Rutherford

Google
Steve is a Software Engineer on Google's Virtualization Security team, which maintains the security of Google Compute Engine. Steve's recent projects include KVM attack surface reduction (pulling legacy interrupt controllers out of KVM), which was merged into the 4.4 kernel.


Wednesday August 24, 2016 11:55am - 12:35pm
Harbour A

12:45pm

Lunch (Attendees On Own)
Wednesday August 24, 2016 12:45pm - 2:15pm
TBA

2:15pm

Securely Integrating QEMU with Open Source Virtualization Technology - Daniel Berrange, Red Hat
The QEMU projects provides the foundation for the overwhelming majority of open source virtualization deployments, being used for both KVM and Xen. This large deployment footprint makes QEMU an attractive target for exploitation and its wide variety of features offer many avenues for attack, whether from the guest or from other infrastructure on the managment LAN. The presentation will outline the technologies available that should be used in all deployments to protect QEMU and its communication channels from compromise. This will cover confinement of the QEMU process, security of its network services, security of its disk storage and future gaps in protection that remain to be addressed.

Speakers
DB

Daniel Berrange

Software Engineer, Red Hat
Daniel is a Principal Software Engineer, working in a variety of roles at Red Hat over the last 12 years. Since 2006, he has been specialized in the development of technologies related to virtualization management, as lead developer of Libvirt, GTK-VNC, Libvirt Perl, Libvirt GObject and Libvirt Sandbox, and contributor to the Xen, KVM, oVirt and OpenStack projects. Daniel is a passionate believer in the value of open source software and the... Read More →


Wednesday August 24, 2016 2:15pm - 3:05pm
Harbour A

3:05pm

Coffee Break
Wednesday August 24, 2016 3:05pm - 3:35pm
Metropolitan Foyer

3:35pm

An Introduction to PCI Device Assignment with VFIO - Alex Williamson, Red Hat
VFIO is a Linux kernel userspace driver framework used by QEMU to make devices directly assignable to virtual machines. This model replaces the now deprecated Legacy KVM device assignment driver. In this talk, Alex Williamson will give an overview of how VFIO works, how a device, in particular a PCI device, is decomposed and exported to userspace, and how a userspace application like QEMU recomposes the device into a virtual machine. Additionally, Alex will highlight the interfaces used to accelerate VFIO through KVM to achieve performance parity with legacy device assignment, while not creating hard dependencies to KVM.

Speakers
AW

Alex Williamson

Red Hat
Alex Williamson has been contributing to the Linux kernel and other open source projects for over ten years and is the maintainer of the VFIO userspace driver interface in the Linux kernel and VFIO-based PCI device assignment in QEMU. Alex currently works for Red Hat from his home in Fort Collins, Colorado.


Wednesday August 24, 2016 3:35pm - 4:25pm
Harbour A

4:35pm

Kernel Protection Using Hardware-Based Virtualization - Jun Nakajima & Sainath Grandhi, Intel
We propose that the Linux run in virtualization mode, activating hardware virtualization features to improve security and monitoring. Hardware enforced virtualization features can be used for hardening the kernel, by protecting key kernel data structures and locking the processor state when the processor is executing in guest mode. Security features from the latest processors can be added to virtual processors. Kernels running on platforms with processors from older generations are benefitted.

For the bare-metal, we have added a thin hypervisor to the kernel, and we have extended KVM for guest kernels so that they can identify this capability as a CPU feature, become enlightened and work with the hypervisor to lock and monitor kernel resources and processor state.

In this talk we will present the idea, its benefits and the work we have done in Linux/KVM.

Speakers
SG

Sainath Grandhi

Intel
Work for Intel in Open Source Virtualization group. Work on Xen and KVM kernel feature enabling. Currently working on a project that is a solution to run containers with a hypervisor underneath to provide security and resource isolation.
avatar for Jun Nakajima

Jun Nakajima

Senior Principal Engineer, Intel Open Source Technology Center
Jun Nakajima is a Senior Principal Engineer leading open source virtualization and cloud projects, such as, KVM, Xen, and OpenStack at the Intel Open Source Technology Center. Jun has been working on various virtualization projects for almost a decade, and NFV is one of his ongoing projects. Jun presented a number of times at technical conferences, including KVM Forum, Xen Summit, and USENIX. He has over 20 years of experience with operating... Read More →


Wednesday August 24, 2016 4:35pm - 5:25pm
Harbour A

6:15pm

Joint Event: 25th Anniversary of Linux 'Casino Royal' Gala (Buses Depart at 6:15 pm)
Prepare for an evening at Casino Royal to celebrate 25 years of Linux. Pack your tux, fancy dress or best suit (or stop at our onsite rental shop on Monday or Tuesday) and head over to Muzik on Wednesday evening for casino gaming, ice wine tasting, live entertainment and much more! 

Open to all KVM Forum attendees with an All-Access Conference Registration. Black tie optional. Dinner and drinks provided. 

Wednesday August 24, 2016 6:15pm - 9:30pm
Muzik
 
Thursday, August 25
 

8:00am

Breakfast & Registration
Thursday August 25, 2016 8:00am - 9:00am
Harbour Foyer

9:00am

Keynote: KVM Status Report - Paolo Bonzini, Senior Principal Software Engineer, Red Hat
Speakers
avatar for Paolo Bonzini

Paolo Bonzini

Sr. Software Engineer, Principal Software Engineer
Paolo Bonzini works on virtualization for Red Hat, where he is a Senior Principal Software Engineer. He is currently the maintainer of the KVM hypervisor and a contributor and submaintainer for QEMU.


Thursday August 25, 2016 9:00am - 9:15am
Harbour B

9:15am

Virtualization Developer Panel - Moderated by Stefan Hajnoczi, Principal Software Engineer, Red Hat
A panel discussion with current active contributors to (and maintainers of) KVM, QEMU, and libvirt (and related Virtualization infrastructure projects).

Topics will be chosen on the spot from areas including security, new hardware features, and development process.

Moderators
SH

Stefan Hajnoczi

Principal Software Engineer, Red Hat
Stefan Hajnoczi has contributed to QEMU since 2010. He currently co-maintains the QEMU block layer together with Kevin Wolf. Now at Red Hat and previously at IBM's Linux Technology Center, Stefan has worked on disk image formats, storage migration, multi-core device emulation, and I/O performance optimization in QEMU. He also maintains the QEMU tracing subsystem and co-maintains the QEMU net subsystem, and is organization administrator for QEMU's... Read More →

Speakers
DB

Daniel Berrange

Software Engineer, Red Hat
Daniel is a Principal Software Engineer, working in a variety of roles at Red Hat over the last 12 years. Since 2006, he has been specialized in the development of technologies related to virtualization management, as lead developer of Libvirt, GTK-VNC, Libvirt Perl, Libvirt GObject and Libvirt Sandbox, and contributor to the Xen, KVM, oVirt and OpenStack projects. Daniel is a passionate believer in the value of open source software and the... Read More →
avatar for Paolo Bonzini

Paolo Bonzini

Sr. Software Engineer, Principal Software Engineer
Paolo Bonzini works on virtualization for Red Hat, where he is a Senior Principal Software Engineer. He is currently the maintainer of the KVM hypervisor and a contributor and submaintainer for QEMU.
AH

Andrew Honig

Tech Lead - Cloud Platform Security, Google
DL

Den Lunev

Team Lead, Virtuozzo
Denis Lunev has started his work for Virtuozzo (ex Parallels, ex SWsoft) more than 15 years ago. From that time he has dealt with Linux namespaces for Virtuozzo, OpenVZ, Linux containers. The next big bit of work was devices emulation for Parallels Desktop and Parallels Server, proprietary virtualization products for Mac and Linux. Now he is the lead of the team in Virtuozzo supporting QEMU/KVM.


Thursday August 25, 2016 9:15am - 10:00am
Harbour B

10:00am

vGPU on KVM - A VFIO Based Framework - Neo Jia & Kirti Wankhede, NVIDIA Corporation
GPU virtualization becomes more and more important these days since its first public debut in 2013. To bring vGPU solutions of different GPU vendors onto KVM/QEMU platform, while maintaining a highly unified kernel and user space interface, a VFIO based vGPU framework is introduced in this presentation. In this talk, we will review the GPU vendor driver registration interface, the vGPU bus, vGPU core driver, VGPU TYPE1 IOMMU internals, walk through vGPU enabled VM life cycle, introduce the kernel and user space sysfs and the framebuffer interface for QEMU console VNC, discuss roadmap for future vGPU VFIO framework performance/scalability optimization.

Speakers
NJ

Neo Jia

Chief Software Architect, NVIDIA Corporation
Neo Jia works on GPU virtualization for NVIDIA Corporation. As a Chief Software Architect and a founding member of GPU virtualization team, he has been actively bringing NVIDIA GPU virtualization technology to different platforms. Nowdays, he is working on vGPU VFIO framework with Linux/KVM community.
KW

Kirti Wankhede

NVIDIA Corporation


Thursday August 25, 2016 10:00am - 10:45am
Harbour B

10:45am

Coffee Break
Thursday August 25, 2016 10:45am - 11:15am
Harbour Foyer

11:15am

VMBus (Hyper-V) devices in QEMU/KVM - Roman Kagan, Virtuozzo
Paravirtualized devices are a well-established way to achieve better performance of virtual machines. However, for Windows guests using VirtIO-based paravirtualized devices isn't easy, because the corresponding drivers aren't bundled with the OS.

Roman Kagan will present the ongoing work on the implementation of VMBus devices in QEMU/KVM. This is the paravirtualization provided by Microsoft Hyper-V virtualization platform, and is natively supported by Windows guests.

The presentation will cover the design of VMBus support in QEMU/KVM, the state of the implementation, the challenges, and the outlook.

Speakers
RK

Roman Kagan

lead developer, Virtuozzo


Thursday August 25, 2016 11:15am - 12:00pm
Harbour A

11:15am

QEMU Support for the RISC-V Instruction Set Architecture - Sagar Karandikar, UC Berkeley
RISC-V is a new instruction set architecture originally designed to support computer architecture research and education at UC Berkeley. RISC-V is now set to become a standard open architecture for industry implementations under the governance of the RISC-V Foundation.

This talk will summarize the benefits of RISC-V and open ISAs for the open-source systems-software community. The first section of the talk will focus on the draft RISC-V privileged specification, including the RISC-V vision for cleaner abstraction between hardware, hypervisors, and operating systems.

The talk will also discuss my experience bringing up RISC-V emulation support in QEMU, including adding architectural support, experimental devices, and fuzz testing against Spike, the “golden-reference” RISC-V simulator. The talk will conclude by outlining opportunities to contribute to RISC-V ISA support in QEMU.

Speakers
avatar for Sagar Karandikar

Sagar Karandikar

Graduate Student Researcher, UC Berkeley
Sagar Karandikar is a first-year graduate student researcher focusing in computer architecture in the Electrical Engineering and Computer Sciences department at UC Berkeley. His research focuses on hardware and software design of next-generation datacenters. He has also been involved in the RISC-V project, contributing to the development of infrastructure surrounding the open RISC-V Instruction Set Architecture.


Thursday August 25, 2016 11:15am - 12:00pm
Pier 3

12:00pm

Dirty Memory Tracking for Performant Checkpointing Solutions - Lei Cao, Stratus Technologies
There are bottlenecks and limitations in KVM that hamper the performance of a checkpoint/rollback system, namely use of large bitmaps to track dirty memory and lack of multi thread support due to mmu lock implementation. In this presentation, I will discuss a new dirty memory tracking mechanism that improves the performance of checkpoint/rollback implementations for commercial production. This new mechanism can also be used to improve the predictability of live migration of memory write intensive workloads.

Speakers
LC

Lei Cao

Stratus Technologies
I work for Stratus Technologies, a company that specializes in fault tolerant hardware and software solutions. I am currently working on our fault tolerant software product.


Thursday August 25, 2016 12:00pm - 12:45pm
Harbour A

12:00pm

Qemu Graphics Update 2016 - Gerd Hoffmann, Red Hat
This talk will give an overview about the graphic-related changes in qemu happened last year and the plans for the future. It will focus on virtio-gpu, spice, virgl and opengl. gpu device assignment and vgpu will be covered too.

Speakers
GH

Gerd Hoffmann

Red Hat
Gerd Hoffmann is working on virtualization. He started a few years back with user mode linux. Later the focus shifted to Xen. Nowdays he is working on qemu and kvm for the Red Hat. Currently he maintains spice and usb subsystems in qemu. Gerd gave various talks on virtualization-focuced conferences (Xen Summit, KVM Forum) and on german linux conferences (LinuxTag, Linux Kongress).


Thursday August 25, 2016 12:00pm - 12:45pm
Pier 3

12:45pm

Lunch (Attendees on Own)
Thursday August 25, 2016 12:45pm - 2:00pm
TBA

2:00pm

Quo Vadis Virtio? - Michael Tsirkin, Red Hat
Last version of virtio was published about a year ago: in october 2016. The virtio community has been idle since: implementations are constantly being improved, new features are in the works, or are being discussed. This talk will present an update on current status of the implementations and the challengest they face, and look forward to how we may address these challenges. It will highlight some of the proposed new features and the benefits they might bring. It will also suggest untested ideas for new applications for virtio on which (to presenter's knowledge) no one is currently working - some of these might turn out to be impractical, but most should serve as useful starting points for a discussion.

Speakers
MS

Michael S. Tsirkin

Senior Principal Software Engineer, Red Hat
Michael S. Tsirkin works on KVM at Red Hat. He is the chair of the virtio technical committee; virtio is the basis of most para-virtualized solutions in use with KVM today. He is the maintainer of virtio and vhost subsystems in Linux as well as PC and PCI subsystems in QEMU. On several occasions he has been recognized as one of the most prolific reviewers and contributors of code to Linux and QEMU. Michael's past speaking experience includes... Read More →


Thursday August 25, 2016 2:00pm - 2:30pm
Harbour A

2:00pm

Jobs and Unemployment in the New QEMU Economy - John Snow, Red Hat
QEMU has a legacy blockjobs system with a somewhat tenuous and poorly designed interface. The current state of Blockjobs is that you can only run one at a time, and the permissions system underpinning when you are allowed to run them is opaque to the user. This can make managing long-running background jobs somewhat difficult, and the lack of persistent handles to jobs that modify the storage tree can make it even more difficult. In this talk I will discuss the future of QEMU Block Jobs, including the abstraction of the "Job" concept to encompass all of QEMU to present a unified object representing long-running tasks to management APIs such as libvirt.

Speakers
avatar for John Snow

John Snow

Software Engineer, Red Hat
John Snow is a UMass Lowell Alumni and (increasingly non-recently) a recent hire for Red Hat. He works for the virtualization team as a Software Engineer and has worked on the AHCI emulator, legacy IDE and FDC support, and gave a talk at KVM Forum 2015 on QEMU's new incremental backup features.


Thursday August 25, 2016 2:00pm - 2:30pm
Pier 3

2:30pm

Design of Vhost-pci - Wei Wang, Intel
Current vhost-user based backend designs for virtio-net devices present scaling challenges, as communication intensive applications (e.g. virtual network functions) running in VMs start to stress this centralized design and resources assigned to it. vhost-pci offers a protected and point-to-point based inter-VM communication mechanism and we present a comprehensive design of vhost-pci in this talk. A VM with a vhost-pci device is able to copy data to another VM’s memory directly. The vhost-pci design is further extended to support communicating to multiple VM destinations.
The vhost-pci based inter-VM communication framework consists of a vhost-pci device, its driver, and an extension to the vhost-user/DPDK to set up the path between VMs. Initial results demonstrate desired scaling advantages due to vhost-pci’s decentralized approach to inter-VM communication.

Speakers
avatar for Wei Wang

Wei Wang

Virtualization Software Developer, Intel
Wei is currently a virtualization software developer at Intel. He earned a Master degree from the University of Ottawa, Canada, and he was a research assistant there working on hardware/software co-design solutions to virtualize FPGA accelerators.


Thursday August 25, 2016 2:30pm - 3:00pm
Harbour A

2:30pm

I/O Prefetch Cache as QEMU Block Filter Driver - Pavel Butsykin, Virtuozzo
Prefetch (aka readahead) is a well-known technique to improve sequential data read performance. However, making it generally useful is challenging, because real-world workloads are typically dominated by random access patterns, which defeat the prefetch efficiency for sequential accesses, and vice versa, latency of random requests may degrade due to prefetch.

In this presentation, Pavel presents an implementation of the prefetch cache as a QEMU block filter driver which improves the I/O performance while maintaining the adverse effects under control. He demonstrates the results of the tests that measure QEMU I/O performance in different configurations of the QEMU block subsystem including the one using virtio-block data plane scheme.

Speakers
PB

Pavel Butsykin

Software Engineer, Virtuozzo
Pavel is a member of a team at Virtuozzo working on QEMU/KVM-based opensource hypervisor for the next generation of Virtuozzo virtualization platform.


Thursday August 25, 2016 2:30pm - 3:00pm
Pier 3

3:00pm

Coffee Break
Thursday August 25, 2016 3:00pm - 3:30pm
Harbour Foyer

3:30pm

AMD’s Virtualization Memory Encryption Technology - Thomas Lendacky, Advanced Micro Devices
AMD recently disclosed the new Secure Memory Encryption (SME) and Secure Encrypted Virtualization (SEV) x86 technologies which leverage hardware-based memory encryption to provide additional security protections. This talk will focus primarily on the SEV technology which supports encrypted virtual machines for extra isolation and protection from the hypervisor itself. The presentation will discuss the technical details of this technology with a focus on how it can be integrated within the KVM infrastructure.

Speakers
TL

Thomas Lendacky

Advanced Micro Devices
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 on x86 and arm64 platforms.


Thursday August 25, 2016 3:30pm - 4:00pm
Harbour A

3:30pm

Backups with QEMU - Max Reitz, Red Hat
QEMU has various different ways of creating backups of disk images while the VM is running: External snapshots, internal snapshots, a “backup” block job, and a “mirror” block job. In turn, each of these may have sub-varieties. After explaining what these things are, this talk will give a high-level user-oriented overview of the characteristics and distinctive features of each approach to serve as a guideline on when and maybe even more importantly on how to use which.

Speakers
MR

Max Reitz

Red Hat
Max Reitz is a computer science student at TU Dresden and a Red Hat intern co-maintaining QEMU's block layer core. Together with Kevin Wolf, he has held talks at KVM Forum 2014 and 2015.


Thursday August 25, 2016 3:30pm - 4:00pm
Pier 3

4:00pm

TCG Enhancements for PowerPC - Nikunj Dadhania, IBM Linux Technology Center
QEMU allows for emulating different target architectures using the TCG mode. TCG
mode has been effectively used for validating pre-silicon systems. Developers
have often used QEMU TCG to develop application stack before actually deploying
them on real target architectures. Even though application development with TCG
can be slow, the non-availability of early hardware is often a compelling reason
to use TCG.

This talk will focus on some of the enhancements that are in progress to support
PowerPC architecture better in QEMU TCG mode. This includes the new PowerNV
platform, support for upcoming POWER9 processor and its ISA. We will also look
at some of the performance enhancements like multi-threaded TCG support for
PowerPC.

Speakers
ND

Nikunj Dadhania

Advisory Software Engineer, IBM Linux Technology Center
Nikunj A. Dadhania works on KVM virtualization as part of IBM Linux Technology Center, India. He has been the maintainer of SLOF (Slim-line Open Firmware) fornKVM on Power. He has experience in libvirt, scheduler and embedded systems.


Thursday August 25, 2016 4:00pm - 4:30pm
Harbour A

4:00pm

Microsoft Failover Clustering on KVM - Vadim Rozenfeld, Red Hat
Thinking about failover clustering as one of the features which helps to make high availability more achievable for KVM based Windows VM setups, Vadim is going to talk about adding Microsoft Failover Clustering compatibility to the existing Windows virtio-scsi miniport device drive, as well as about the relevant changes that need to be done at QEMU level in order to support Microsoft Failofer Clustering.

Speakers
VR

Vadim Rozenfeld

Software Engineer, Red Hat
I'm a software engineer, working at Red Hat for almost nine years, and mostly specializing on developing para-virtualized device drivers for Windows platforms. I'm currently working on making Windows virtio-scsi miniport driver compatible with Microsoft Failover Clustering requirements. Last time I participated as a speaker at KVM Forum was in 2012 when I was speaking about Hyper-V enlightenment features, recently added to KVM... Read More →


Thursday August 25, 2016 4:00pm - 4:30pm
Pier 3

4:30pm

4:45pm

Lightning Talks
Thursday August 25, 2016 4:45pm - 5:30pm
Harbour B

5:30pm

Birds of a Feather (BoF) Discussion
A sign up sheet for suggested Birds of a Feather topics will be availble throughout the day.

Thursday August 25, 2016 5:30pm - 6:00pm
Harbour A

5:30pm

Birds of a Feather (BoF) Discussion
A sign up sheet for suggested Birds of a Feather topics will be availble throughout the day.

Thursday August 25, 2016 5:30pm - 6:00pm
Harbour B

5:30pm

Birds of a Feather (BoF) Discussion
A sign up sheet for suggested Birds of a Feather topics will be availble throughout the day.

Thursday August 25, 2016 5:30pm - 6:00pm
Pier 3

7:00pm

Joint Attendee Party with Xen Project Developer Summit (Badge Required)
Join fellow attendees of KVM Forum along with Xen Project Developer Summit attendees for food, drink, and a memorable hockey experience.

Party guests will have the opportunity to view historic hockey memorabilia, visit a replica of the Montreal Canadien's dressing room, and play a variety of interactive hockey games.

The party is located on the Concourse Level of the Hockey Hall of Fame - a 10 minute walk from the Westin. Walking maps will be provided.

Badge required for entry to party.

Thursday August 25, 2016 7:00pm - 9:00pm
Hockey Hall of Fame 30 Yonge St, Toronto, ON M5E 1X8, Canada
 
Friday, August 26
 

8:00am

Breakfast & Registration
Friday August 26, 2016 8:00am - 9:00am
Harbour Foyer

9:00am

Keynote: QEMU Status Report - Alexander Graf, Upstream Maintainer, SUSE
Speakers
AG

Alexander Graf

Upstream Maintainer, SUSE
Alexander Graf | Upstream Maintainer - SUSE | Alexander started working for SUSE about 9 years ago. Since then he worked on fancy things like SUSE Studio, QEMU, KVM and openSUSE on ARM. Whenever something really useful comes to his mind, he tends to implement it. Among others he did Mac OS X virtualization using KVM, nested SVM, KVM on PowerPC and a lot of work in QEMU for openSUSE on ARM. He is the upstream maintainer of KVM for PowerPC, QEMU... Read More →


Friday August 26, 2016 9:00am - 9:15am
Harbour B

9:15am

QEMU Community Growth Through Open Source Internships - Stefan Hajnoczi, Red Hat
QEMU began participating in the Google Summer of Code internship program in 2010 and has since expanded to Outreachy, a program aimed at underrepresented groups in open source. As QEMU enters its sixth year of open source internships, it's time to explain why it makes sense to participate and how community volunteers can successfully mentor remote interns.

Mentors from the QEMU community help interns with a 12-week full-time remote work project. Most project ideas are proposed by mentors and have multiple applicants. What makes a good internship project idea and how to select applicants who will become long-term members of the community are two fundamental questions that this presentation answers.

The aim of the presentation is to encourage even more participation and show how the time, effort, and money in these programs benefits interns, QEMU, and sponsors.

Speakers
SH

Stefan Hajnoczi

Principal Software Engineer, Red Hat
Stefan Hajnoczi has contributed to QEMU since 2010. He currently co-maintains the QEMU block layer together with Kevin Wolf. Now at Red Hat and previously at IBM's Linux Technology Center, Stefan has worked on disk image formats, storage migration, multi-core device emulation, and I/O performance optimization in QEMU. He also maintains the QEMU tracing subsystem and co-maintains the QEMU net subsystem, and is organization administrator for QEMU's... Read More →


Friday August 26, 2016 9:15am - 10:00am
Harbour B

10:00am

Painless Switch from Proprietary Hypervisor to QEMU/KVM - Den Lunev, Virtuozzo
A year ago Virtuozzo (ex Parallels) has made a decision to switch from proprietary hypervisor to QEMU/KVM/libvirt stack. This hard made decision was taken after serious investigations. In this presentation, Den Lunev will highlight possible integration options and architectural decisions made during this switch, challenges for the team and the results obtained.

Speakers
DL

Den Lunev

Team Lead, Virtuozzo
Denis Lunev has started his work for Virtuozzo (ex Parallels, ex SWsoft) more than 15 years ago. From that time he has dealt with Linux namespaces for Virtuozzo, OpenVZ, Linux containers. The next big bit of work was devices emulation for Parallels Desktop and Parallels Server, proprietary virtualization products for Mac and Linux. Now he is the lead of the team in Virtuozzo supporting QEMU/KVM.


Friday August 26, 2016 10:00am - 10:45am
Harbour B

10:45am

Coffee Break
Friday August 26, 2016 10:45am - 11:15am
Harbour Foyer

11:15am

QEMU Coroutines, Exposed - Jeff Cody, Red Hat
Coroutines are an integral part of the QEMU block layer, providing a cooperative multi-tasking framework that largely obviates the need for callback functions. This allows the QEMU block layer to implemented a sequential, multi-threaded workflow, without the added complexity of preemption. This presentation will go over the QEMU coroutine implementation, how coroutines are used in the block layer, and the advantages of coroutines. And -- despite these advantages -- coroutines are not without their drawbacks, such as their own code complexity and unique debugging challenges. These drawbacks will also be discussed in the presentation.

Speakers
JC

Jeff Cody

Sr. Software Engineer, Red Hat
Jeff is a Senior Software Engineer with Red Hat, currently working in the virtualization group. He currently works on the QEMU block layer, with a focus on live block job operations, and block image formats. Jeff was a presenter at the 2012, 2013, and 2014 KVM Forums. He works remotely from his home in Fort Wayne, Indiana.


Friday August 26, 2016 11:15am - 12:00pm
Harbour A

11:15am

KVM on System z: The Good, the Bad and the Weird - Cornelia Huck, IBM
System z aka s390x has been the second architecture supported by KVM, and it has been quite a way since then. This talk will examine how we dealt with the advantages and the pecularities of System z: What we have an instruction for, where we'd like to travel back in time to fix the interface, and where we just have to be different. You may also learn some new acronyms along the way.

Speakers
CH

Cornelia Huck

IBM
Cornelia has been working at IBM for 15+ years on the System z aka s390 platform. She has been doing work in the Linux kernel and virtualization (KVM, qemu), being one of the s390 maintainers for KVM and qemu. Her main projects include virtio and s390 channel I/O, on which she has presented at KVM forum before.


Friday August 26, 2016 11:15am - 12:00pm
Pier 3

12:00pm

Atomic.h weapons: The C11 Memory Model and QEMU - Paolo Bonzini, Red Hat
C gained its popularity because of its flexibility and the ability to function as a "high-level assembler". However, due to the growing complexity of programming both languages and code bases, compiler writers need to extract more and more information from the source code, causing behavior that the programmer did not expect---the dreaded "undefined behavior".

This is even more true for multithreaded programs. As soon as the comfort of pthread APIs is left, compiler writers and users must speak a common language (a "memory model") in order to express the semantics of a multithreaded program. In this talk I will explain: some key concepts of the C11 memory model; how they form the basic API in QEMU's "atomic.h" header file; how a subset of C11 atomics was chosen, and how it might grow in the future; how a theoretical understanding of the model makes code easier to write and understand.

Speakers
avatar for Paolo Bonzini

Paolo Bonzini

Sr. Software Engineer, Principal Software Engineer
Paolo Bonzini works on virtualization for Red Hat, where he is a Senior Principal Software Engineer. He is currently the maintainer of the KVM hypervisor and a contributor and submaintainer for QEMU.


Friday August 26, 2016 12:00pm - 12:45pm
Harbour A

12:00pm

Light Weight Virtualization with QEMU/KVM - Chao Peng, Intel
Comparing to traditional full-blown VM based virtualization,light weight virtualization like Linux container boots much faster and runs more efficiently.But it also has drawback:the software-level isolation makes it not as secure as traditional hardware-enforced hypervisor.In this talk Anthony/Chao will share their work on tailoring and improving the traditional hypervisor to run a light weight VM, to gain both performance and scalability, while at the same time maintaining hardware-based isolation. Specifically, you will see how the open-source and industry-proven QEMU/KVM can be optimized to achieve near native container performance/scalability (e.g. fast boot and small memory footprint). Firstly,you will see how QEMU boot time can be shortened from ~300ms to ~50ms. Secondly, you will also see how these improvements be used in the related projects like Intel Clear Container and Docker.

Speakers
CP

Chao Peng

Sr. Software Engineer, Intel
Chao is a senior software engineer from Intel and his job content includes hardware virtualization feature enabling for open source hypervisors like KVM/Xen, as well as the performance tuning for virtualization and cloud products. Recently he focuses on light-weight virtualization solutions like Linux container and try to improve the security of container using hardware isolation technologies.


Friday August 26, 2016 12:00pm - 12:45pm
Pier 3

12:45pm

Lunch (Attendees on Own)
Friday August 26, 2016 12:45pm - 2:00pm
TBA

2:00pm

Machine-type Introspection and Configuration: Where Are We Going? - Eduardo Habkost, Red Hat
In the past years we have developed multiple solutions to try to make QEMU configuration more introspectable and machine-friendly (e.g.: -readconfig/-writeconfig, qdev, QOM, QMP, and others). Those solutions have addressed many problems, but don't cover all aspects of QEMU configuration and introspection. Some areas, such as machine-types and other command-line options lack machine-friendly configuration/introspection interfaces that effectively help management software.
In this talk, we will discuss existing deficiencies and possible approaches to cover them using our existing configuration and introspection mechanisms. The talk will touch some topics discussed at previous KVM Forum 2015 talks: "Rethinking machine types" by David Gibson, and "QEMU interface introspection: from hacks to solutions" by Markus Armbruster.

Speakers
avatar for Eduardo Habkost

Eduardo Habkost

Software Engineer, Red Hat Inc
Eduardo Habkost is a Software Engineer in the KVM team at Red Hat, who lives in Curitiba, Brazil. Having worked maintaining kernel-xen in Fedora, now he works mostly in KVM and QEMU.Before joining Red Hat in 2007, Eduardo worked for 6 years for Conectiva (which later become Mandriva), in multiple areas, including kernel code, device drivers, and packaging for the Conectiva Linux distribution.


Friday August 26, 2016 2:00pm - 2:30pm
Harbour A

2:00pm

Real Time & Fast Live Migration Update for NFV - Liang Li, Intel
NFV usages poses stringent requirements on KVM in terms of real time and live migration. This presentation is an update about the progress we have made for these requirements.
We have to reduce the interrupt latency so as to speed up packet forwarding. By utilizing and enhancing the real time Linux kernel, reducing the VMX cost, we reduced the cyclic test result from 27 to 11 us, device interrupt handling latency from 49us to less than 10 us.
We have come up with a bunch of innovation approaches that improve the performance of live migration significantly. We shortened the live migration duration from 2.3s to 700ms while keeping the VM down time < 7ms, and reduced the network traffic by about more than 60% in some typical cases.
In this talk, Yunhong & Liang will share their experience for real time optimization, and introduce the technical details for live migration optimization.

Speakers
LL

Liang Li

Sr. Software Engineer, Intel
Yunhong is a senior software engineer works for Intel, and he has more than 10+ years experience in the area of system virtualization and open source.nLiang is a software developer works for Intel. He have 7+ years’ experience of embedded software development and about 2 years’ experience of system virtualization. He had a topic about KVM live migration optimization in KVM forum 2015.


Friday August 26, 2016 2:00pm - 2:30pm
Pier 3

2:30pm

CPU Hotplug Support in QEMU - Bharata Rao, IBM Linux Technology Center & David Gibson, Red Hat
CPU hotplug allows virtual CPUs to be added to, or removed from a running KVM guest. This can be used to scale the guest's compute capacity on demand in cloud environments.

CPU hotplug has been available for some time on x86 systems. However, the approach used there doesn't generalize well to other platforms. There has been an ongoing effort to design a general infrastructure for CPU hotplug in QEMU and, in particular, common management interface semantics usable on all platforms.

This talk will cover the challenges of supporting CPU hotplug in qemu, and highlight which areas do and don't yet have consensus on design. In particular it will discuss the proposed scheme implemented on Power and x86 based on a combination of generic device_add/device_del based hotplug and a new QMP interface for use by management layers.

Speakers
DG

David Gibson

Senior Software Engineer, Red Hat
David is a qemu and Linux kernel developer, currently working at Red Hat on KVM and qemu for IBM Power servers. Previously, while employed by IBM, he was the main author of qemu's 'pseries' machine type. He also wrote the Device Tree Compiler, and has worked on low-level memory management for Power machines, bringup for embedded PowerPC machines and the 'orinoco' wireless driver.
BB

Bharata B. Rao

IBM Linux Technology Center
Bharata B Rao works on KVM virtualization as part of IBM Linux Technology Center, India. Earlier he has worked in the area File Systems, Scheduler, Debuggers etc. He has graduated from the University of Mysore. He has spoken at linux.conf.au and Linux Plumbers earlier.


Friday August 26, 2016 2:30pm - 3:00pm
Harbour A

2:30pm

Vhost with Guest vIOMMU - Peter Xu & Wei Xu, Red Hat
Vhost is the technique used in QEMU/KVM to accelerate IOs for virtio devices, so that guest IOs can be handled without trapping into QEMU userspace. While vhost boosts guest block/network performance, it also brings security issues considering it is (before this work) bypassing protection of vIOMMUs. This becomes a more critical issue for NFV scenarios, in which user-space drivers are vastly used. A problematic user space driver inside guest can directly corrupt guest kernel in no time. In this presentation, Jason Wang (Wei Xu as presenter) and Peter Xu will propose a solution to add guest vIOMMU protection for vhost devices, and some performance numbers.

Speakers
avatar for Peter Xu

Peter Xu

Software Engineer, Red Hat
Software engineer at Red Hat virtualization team.
WX

Wei Xu

Red Hat


Friday August 26, 2016 2:30pm - 3:00pm
Pier 3

3:00pm

Nesting KVM on s390x - David Hildenbrand, IBM
The way virtualization is built into IBM z Systems hardware allows for a quite neat and clean implementation of nested virtualization. This session gives an introduction to nested virtualization on IBM z Systems, explaining where we stand, how it works and what is missing. Also included is a short discussion of security aspects. Be prepared for some memory management, interrupt handling and virtualization insights.

Speakers
avatar for David Hildenbrand

David Hildenbrand

Software engineer, IBM R&D Germany GmbH
David has been working as software developer at IBM on QEMU/KVM for Linux on z Systems for ~1.5 years. His projects include nested virtualization, hardware support for guest debugging, cpu models and architecture compliance.


Friday August 26, 2016 3:00pm - 3:30pm
Harbour A

3:00pm

Status Update on KVM-COLO FT - Hailang Zhang, Huawei
COLO is an ideal Application-agnostic Solution for Non-stop service in the cloud. Under COLO mode, both primary VM (PVM) and secondary VM (SVM) are run in parallel. They receive the same request from client, and generate response in parallel too. If the response packets from PVM and SVM are identical, they are released immediately. Otherwise, a VM checkpoint (on demand) is conducted. COLO prototype has been realized, and most of the patches has been reviewed in QEMU community. Same as last year, in this talk, we will talk about COLO implementation in QEMU, and report the latest progress from Huawei, Fujitsu, Intel, Since COLO is still in the early stage, some future works need to be done to make it commercial use, we will talk about that too.

Speakers
HZ

Hailiang Zhang

Software Engineer, Huawei
Hailiang Zhang is a software engineer at Huawei for 4 years. Currently he focus on Virtual Machine (VM) replication projects (COLO on QEMU


Friday August 26, 2016 3:00pm - 3:30pm
Pier 3

3:30pm

Coffee Break
Friday August 26, 2016 3:30pm - 4:00pm
Harbour Foyer

4:00pm

As Time Goes By, Analysing Where We Spend Our Cycles During Exits - Christian Bornträger, IBM Germany
KVM exits spend several hundreds or thousands of cycles running hypervisor code. This talk will should where we spend our time, how to analyze that, what has been done and what can still be done to improve things.

Speakers
CB

Christian Bornträger

Maintainer KVM on z (s390), IBM Germany
Christian Borntraeger is a software engineer at IBM Germany and is the s390 maintainer for KVM and QEMU/KVM


Friday August 26, 2016 4:00pm - 4:30pm
Harbour A

4:00pm

Qemu as a USB-MTP Responder - Bandan Das, Red Hat
Media Transfer Protocol (MTP) is an extension to Picture Transfer Protocol for handling transfer of media files. MTP exposes a file oriented protocol where the MTP responder is in control of the file operations and capabilities. Atomicity guarantees by the MTP server thereby help with eliminating data corruption which are seen with block based file sharing. This talk is an overview of the Media Protocol Server implementation in Qemu. By exposing itself as a MTP server, file sharing between hosts and guests is as simple as choosing a guest file browser with an integrated MTP client. We will delve into implementation details such as file operations and monitoring, usage scenarios, and unimplemented features such as write support. We will also discuss the pros and cons with respect to other host-guest sharing mechanisms available in Qemu.

Speakers
avatar for Bandan Das

Bandan Das

Software Engineer, Red Hat
Bandan Das is a Software Engineer in the Virtualization group at Red Hat. He likes working on the Linux kernel, he usually picks up work that everyone else has forgotten about or don't want to work on.


Friday August 26, 2016 4:00pm - 4:30pm
Pier 3

4:30pm

NVDIMM Virtualization - Xiao Guangrong, Intel
NVDIMM is the key feature introduced to Intel Skylake Servers which has the density characteristics of NAND FLASH but performance characteristics comparable to DRAM. This opens up the opportunity of introducing a new level in the memory hierarchy that enables both the traditional volatile system memory and a new persistent memory accessed via the same synchronous, load/store semantics.

Virtualization plays a key role in cloud infrastructure, introducing NVDIMM into the virtualization field brings IT revolution and infrastructure changes.

As the contributor of NVDIMM virtualization, we will talk about the implementation of vNVDIMM and how it can be used in tradition cloud computing and why it bring signification benefit for lightweight virtualization such as Intel's Clear-Container.

Speakers
XG

Xiao Guangrong

Linux Kernel Developer
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 and China CLK 2012nHe is the maintainer of NVDIMM in Qemu’s community who designed and implemented NVDIMM in KVM and Qemu.


Friday August 26, 2016 4:30pm - 5:00pm
Harbour A

4:30pm

Libvirt Admin API - A Different Kind of Management for libvirt - Erik Skultety, Red Hat
Libvirt has been well known for providing a stable abstract middleware layer that hides hypervisor specifics from its users in order to provide a uniform way of management, while having most of its features available through libvirtd daemon. Although libvirtd can be configured through a configuration file, there has not been a way to change the configuration nor manage the daemon itself during runtime. And that is exactly where libvirt Admin API comes in.
This presentation will explore this new administration interface the goal of which is to support runtime daemon management. During the presentation, scenarios like tweaking the log settings to make the daemon more or less verbose and, thus tailoring it to one's linking, threadpool management, as well as looking at clients currently connected to the daemon will be addressed and demonstrated by using the virt-admin utility.

Speakers
ES

Erik Skultety

Associate Software Engineer, Red Hat
Erik Skultety is an associate software engineer at Red Hat working on libvirt project. At the same time he's finishing his studies in computer science at Brno University of Technology.


Friday August 26, 2016 4:30pm - 5:00pm
Pier 3

5:00pm

Closing Session & Photo
Friday August 26, 2016 5:00pm - 5:30pm
Harbour B