Loading…
Xen Developer Summit 2015 has ended
Session [clear filter]
Monday, August 17
 

9:00am PDT

Xen Project Weather Report - Lars Kurth, Citrix
In this talk, we will give an overview of the state of the Xen Project, trends that impact the project, see whether challenges that surfaced last year have been addressed and how we did it, and highlight new challenges and solutions for the coming year.

Speakers
avatar for Lars Kurth

Lars Kurth

Director Open Source / Project Chairperson The Xen Project , Citrix Systems UK Ltd.
Lars Kurth is a highly effective, passionate community manager with strong experience of working with open source communities (Symbian, Symbian DevCo, Eclipse, GNU) and currently is the community manager for the Xen Project. Lars has 12 years of experience building and leading engineering... Read More →


Monday August 17, 2015 9:00am - 9:45am PDT
Ravenna ABC Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

9:45am PDT

Xen 4.6 and Beyond - Wei Liu, Citrix
The talk is a status report for the latest release. It will cover the new features and important bug fixes (if any) in 4.6. It will also provide insight on what’s in the queue for 4.7. Retrospective on the release process will also be part of talk.

Speakers
WL

Wei Liu

Software Engineer, Citrix
Wei has worked on various aspects in Xen ecosystem for the past few years. His recent interest is hypervisor development and upstream CI systems.


Monday August 17, 2015 9:45am - 10:30am PDT
Ravenna ABC Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

11:00am PDT

Xen and OpenStack - Stefano Stabellini, Citrix
This talk will explain how to deploy OpenStack using the Xen hypervisor to run your virtual machines.

It will start by describing the OpenStack architecture and its most important components. It will go into details on how to use DevStack to setup an OpenStack development environment based on Xen and Libvirt. It will cover the current status of Xen in OpenStack, in particular with regards to the OpenStack CI-loop, and the latest plans to improve Xen support in Nova.

The audience will learn the most important OpenStack and Xen config options, the pitfalls to avoid in their deployment and how to tweak both of them to get the best performance and stability out of your OpenStack cloud.

Speakers
avatar for Stefano Stabellini

Stefano Stabellini

Principal Engineer, Xilinx
Stefano Stabellini serves as system software architect and virtualization lead at Xilinx, the world's largest supplier of FPGA solutions. Previously, at Aporeto, he created a virtualization-based security solution for containers and authored several security articles. As Senior Principal... Read More →


Monday August 17, 2015 11:00am - 11:30am PDT
Ravenna ABC Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

11:30am PDT

Migration (v2) - Andrew Cooper, Citrix System R&D Ltd
Migration of Virtual Machines is commonplace these days, but its support should never be taken for granted. Andrew Cooper will discuss why XenServer decided to rewrite Xen's migration support from scratch. He will cover the pre-existing issues, the surprises encountered along the way, the current status upstream, and the future areas to be explored.

Speakers
AC

Andrew Cooper

I have been at Citrix since graduating, working in the XenServer Ring0 team, largely at the hypervisor and kernel level.For the same duration, I have been a member of the Xen upstream community, and am now an x86 subsystem maintainer for the hypervisor.


Monday August 17, 2015 11:30am - 12:00pm PDT
Ravenna ABC Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

12:00pm PDT

Intel Graphics Virtualization Technology Update - Zhi Wang, Intel Corporation
Intel Graphics Virtualization Technology for full GPU virtualization (Intel GVT-g, a.k.a XenGT) is an open source project, published in 2013, to enable various GPU accelerated virtualization usages based on Xen. Since then lots of new enhancements are introduced. In this talk, the author will give an update of XenGT project, including upstreaming progress, new platform support (e.g. BDW), various optimizations (e.g. smart shadow GPU page table), and how different usages are effectively enabled with full GPU virtualization.

Speakers
L

智 王

Senior Software Engineer, Intel Corporation
Major contributor of the GVT-g upstream who have worked in this projectas a key developer for more than 3 years. In 2016, he worked with otherkernel maintainers and pushed whole project into Linux kernel. Now he is one of the maintainers of GVT-g.


Monday August 17, 2015 12:00pm - 12:50pm PDT
Ravenna ABC Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:20pm PDT

ARM: Virtualization Extensions - Marc Zyngier & Thomas Molgaard, ARM Ltd
The ARM architecture is expanding its presence beyond the traditional embedded device market. Virtualization capabilities were added to the latest revision of the ARMv7 architecture (with processors like Cortex A7 and A15), and this was extended further with ARMv8 (64bit). The design of the virtualization extensions has made it possible to fairly easily port the Xen and KVM hypervisors.

This presentation will give a hypervisor agnostic view of the virtualization extensions from a CPU level, as well as some of the key peripherals developed as part of this. We will also describe some of the techniques that both Xen and KVM use, and present an overview of the ARM virtualization ecosystem.

Speakers
avatar for Thomas Molgaard

Thomas Molgaard

Senior Director, Arm
Director of Technology Management at Arm, based in Cambridge (UK). Working with a team of Software Technology Managers, setting the direction for Arm’s open source SW activities across all segments. Involved with the Linaro since 2013, currently serving on the TSC. In a previous... Read More →
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 →


Monday August 17, 2015 2:20pm - 2:50pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:20pm PDT

Latest News on Xen Support in libvirt - James Fehlig, SUSE Linux
A lot of progress has been made in the libvirt libxl driver since the last Xen Developer Summit. Looking back at the "Planned Improvements" slide from last year's presentation, most items have been implemented, committed upstream, and included in a release. In addition, there have been concurrency and stability improvements in libxl and libvirt to help elevate OpenStack+libvirt+Xen to supported status. But there are many more improvements and features waiting in the queue. This talk provides an update of the current status of libxl support in libvirt, highlights important missing features that might be interesting to community developers wanting to hack on libvirt, and discusses libxl improvements that could benefit libvirt.

Speakers
avatar for James Fehlig

James Fehlig

Software Engineer, SUSE
Jim Fehlig is a software engineer at SUSE Linux and has been working in the virtualization management space for several years. Since 2008, Jim has been a maintainer of the libvirt project, contributing primarily to the Xen drivers. He has also contributed to several other virtualization... Read More →


Monday August 17, 2015 2:20pm - 2:50pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:50pm PDT

Virtual Machine Introspection with Xen - Tamas Lengyel, Technische Universitat Muenchen
Over the last year Xen has received a lot of work and attention on its VMI capabilities. Starting with Xen 4.6, VMI is going to be natively supported on both Intel and ARM chips, bringing forth an unparalleled API on which users can build monitoring and security applications. In this talk we will take a look at the current state of affairs and look ahead to what features are coming. We will discuss the hardware issues and security vulnerabilities that were discovered over the last year as a result of the active development of new VMI features. We'll explore lessons learnt during the development process of the new and revamped VMI subsystem and discuss practical tools and methods that have been built to aid the process of developing new cross-platform Xen features. Finally, recommendations will be put forth to improve the experience of new developers contributing to Xen.

Speakers
avatar for Tamas Lengyel

Tamas Lengyel

Sr Security Researcher, Intel
Tamas is maintainer of several open-source projects, including the Xen hypervisor, DRAKVUF and LibVMI.


Monday August 17, 2015 2:50pm - 3:20pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:50pm PDT

Virtualizing the Locomotive: Ready, Set, Go! - Mark Kraeling, GE Transportation
Virtualization has so far been largely absent on locomotives in North America. As railroad customers move to view their locomotives as networked assets, and with the introduction of concepts such as mobile data centers, In 2015, standard platforms will be deployed by GE that utilize Xen for x86-based processors, and KVM for ARM-based processors. This end-user session will discuss the desired features that are needed for a hypervisor, the enhancements that could be considered in the locomotive real-time environment, and the applications that will utilize this technology. Finally, as an option, actual locomotive control systems equipment will be brought to the presentation to show a quick demonstration of the applications, how they communicate data off-board, and the industrial hardware that it runs on.

Speakers
MK

Mark Kraeling

GE
Mark is a product manager at GE Transportation developing architectures for onboard locomotive control and information systems. He is a regular speaker at the Embedded Systems Conference where he presents papers on embedded software, communications protocols, and safety-critical architectures... Read More →


Monday August 17, 2015 2:50pm - 3:20pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

3:20pm PDT

QEMU Security Vulnerability Automatic Discovery by Symbolic Execution - Chunjie Zhu, Citrix
Fuzz testing is a popular software testing technique, often used to automatically discover severe bugs. Fuzz testing is equally useful in a security context, where any code fault that compromises memory safety or privilege separation is likely to be a security vulnerability. We have successfully used this approach to harden QEMU, where it identified several severe errors (e.g. buffer overflows) with a potential security impact. However, fuzzing has an inherent disadvantage, that is, the time required for reaching a sufficient confidence level is hard to determine. Symbolic execution on the other hand is a more systematic approach to automatic testing, focusing on generating inputs that achieve high code coverage. We apply this technique to QEMU using the state-of-the-art KLEE symbolic execution engine (https://klee.github.io).

Speakers
CZ

Chunjie Zhu

I am working on XenServer product in Citrix company.


Monday August 17, 2015 3:20pm - 3:50pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

3:20pm PDT

VM Introspection: Practical Applications - Steven Maresca, Zentific LLC and Russell Jancewicz, Zentific LLC
Xen possesses uniquely powerful capabilities for monitoring virtual machine behavior. Collectively known as virtual machine introspection (VMI), these capabilities are often described from a security perspective, yet they are incredibly useful far beyond that sphere. VMI enables observation of both OS and software from an external vantage point, even when a VM has crashed, when a customer has locked himself out of the OS, when a guest is so minimalistic that it lacks management tools, and when a VM is inaccessible via network. We will explore how VMI complements traditional systems monitoring tools and aids issues encountered in daily IT operations. A variety of perspectives will be considered from unikernels to automotive/embedded, from performance metrics to audit/compliance, from debugging to security. Finally, we will review dev tools and libraries for using these fantastic features.

Speakers
RJ

Russell Jancewicz

Senior Security Engineer, Zentific LLC
Russell Jancewicz is a security researcher and software developer at Zentific LLC. He has experience with large scale enterprise environments supporting a wide range of services, including identity and access management for a top fifteen research university. His focus of research... Read More →
SM

Steven Maresca

Lead engineer and security researcher, Zentific LLC
Steven Maresca is lead engineer and security researcher at Zentific LLC. In this role, he builds flexible tools to manage virtual machines and reduce complexity in systems security.  He possesses deep knowledge of hypervisor internals, systems design, and information security. Most... Read More →


Monday August 17, 2015 3:20pm - 3:50pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

4:20pm PDT

Hyper: Make VM Run Like Containers - Xu Wang, Hyper
Hyper is an open source project on app-centric hypervisor. it launches app-centric images, e.g. docker image, on hypervisor without a full guest OS, which makes it have much less overhead than traditional virtual machine. On the other hand, hyper provides better isolation over containers, and better compatibility to VM based infrastructure. Hyper is based on Xen or KVM/Qemu currently.

Speakers
avatar for Xu Wang

Xu Wang

Senior Staff Engineer, Ant Financial
Xu Wang is a senior staff engineer at Ant Financial and an initial member of Kata Containers Architecture Committee. He was the CTO and Cofounder of hyper.sh and created hypervisor-based open source container runtime runV (secure as VM, fast as container). runV merged with clear containers... Read More →


Monday August 17, 2015 4:20pm - 5:10pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

4:20pm PDT

xSplice - Patching Hypervisor - Konrad Rzeszutek Wilk, Oracle
The Xen hypervisor in various installations needs to be updated - either due to Xen Security Advisories or other patches. xSplice offers a method to do live patching without having to reboot the system. In this talk Konrad will discuss the design of it, how to use it and how it works.

Speakers
KR

Konrad Rzeszutek Wilk

Software Director, Oracle
Konrad Wilk is a Software Director at Oracle. His group's mission is to make Linux and Xen Project virtualization better and faster. As part of this work, Konrad has been the maintainer of the Xen Project subsystem in Linux kernel, Xen Project maintainer and had been the Release Manager... Read More →


Monday August 17, 2015 4:20pm - 5:10pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

5:10pm PDT

Rump Kernel Based Upstream QEMU Stubdom - Wei Liu, Citrix
This talk will start with introduction to unikernel and QEMU stubdom in general, then gives a basic rump kernel introduction and the reason we choose it as foundation for QEMU stubdom work. Finally we will go over the progress we make and the plan going forward.

Speakers
WL

Wei Liu

Software Engineer, Citrix
Wei has worked on various aspects in Xen ecosystem for the past few years. His recent interest is hypervisor development and upstream CI systems.


Monday August 17, 2015 5:10pm - 5:40pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

5:10pm PDT

Status Update on Xen-COLO HA/FT Solution - Hongyang Yang, Fujitsu
COLO is a VM replication technique which provides application-agnostic software-implemented hardware fault tolerance "non-stop service". 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 patches for Xen have already been sent to the dev maillist and are under review. In this talk, we will talk about the COLO implementation in Xen, the new designed block replication, discussing on problems we've met while developing COLO. Apart from the technical part, we will also present the latest progress from Fujitsu and Intel.

Speakers
HY

Hongyang Yang

Fujitsu
Yang Hongyang is a software engineer at Fujitsu for 7 years. Has been working on a wide range of software projects at fujitsu, and has contributed to opensource projects such as Linux, Xen and QEMU. Currently focus on Virtual Machine (VM) replication projects (Remus on Xen, COLO on... Read More →


Monday August 17, 2015 5:10pm - 5:40pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

5:40pm PDT

Faster, Improved Guest Model for Xen - Elena Ufimtseva, Oracle
PVH guest has become a number one priority in recent years for XenProject community. This type of guest capitalizes on the best component from PV and HVM guests and achieves better performance and maintainability. This talk is about fascinating PVH features and how the Xen hypervisor and Linux kernel benefit from this guest model. The talk will also cover challenges, existing problems and future work.

Speakers
avatar for Elena Ufimtseva

Elena Ufimtseva

Software Engineer, Oracle
I am a former OPW intern of 2013 program and my project was vNUMA work for Xen hypervisor.  After OPW I realized that I like to continue to contribute to Xen Project and Linux kernel. I had given a talk at 2013 LinuxCon in Europe and at Xen Project summit of the same year about my... Read More →


Monday August 17, 2015 5:40pm - 6:10pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

5:40pm PDT

Live Migration at AliCloud - Benefits, Challenges, Developments, and Future Works - Liu Jinsong, AliCloud
This slides is both a technical and an end-user talks to introduce live migration at AliCloud. From end-user point of view it introduces what fields live migration benefits AliCloud, and what challenges it met when develop live migration at AliCloud in practice. It also gives an example how AliCould fix h/w issues of ~70 Shuguang X86 servers via live migration. From technical point of view it summarizes what problems it solved in the fields of virtualization, network, and storage to make live migration work at AliCloud. Especially, it demonstrates how live migration performance, reliability, and fault-tolerance were improved at AliCloud, with 20+ enhancement. This slides finally introduces the problems not solved yet (i.e., heterogeneous migration issue), and what future works to-do for AliCloud live migration (i.e., further improvement of performance and fault-tolerance, live migration).

Speakers
LJ

Liu Jinsong

Liu Jinsong, an experienced developer and PM maintainer/ RAS maintainer at Xen community. During 2004 – 2014, Jinsong worked at Intel Opensource Technology Center, contributed 3 Xen sub-systems: Xen Power Management, Xen RAS, CPU and Memory hotplug, with many other x86 features... Read More →


Monday August 17, 2015 5:40pm - 6:10pm PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101
 
Tuesday, August 18
 

9:00am PDT

Achieving QoS in Server Virtualization: Intel Platform Shared Resource Monitoring/Control in Xen - Chao Peng, Intel
As more and more VMs run simultaneously on a single platform, the platform shared resources(eg. LLC, Memory bandwidth) become a key contention between VMs. The increasing resource contention and lack of control for it result in poor determinism and QoS for each individual VM. This is, however, a key requirement for cloud infrastructure, data center apps and communications workloads(NFV/SDN).This talk describes the QoS challenges for Server Virtualization. It then introduces Intel platform shared resource monitoring/control technologies exist on recent Intel server platforms and examine how QoS can be achieved with them. After that it outlines the implementation in recent Xen, such as CMT,MBM and CAT, etc. Particularly, it focuses on how these technologies can be used to improve QoS in modern data centers and cloud platforms. Finally, it introduces the potential future work in this area.

Speakers
CP

Chao Peng

I'm a software engineer from Intel. Recently I enabled Platform QoS features for Xen project.


Tuesday August 18, 2015 9:00am - 9:30am PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

9:00am PDT

PCI Passthrough Support in Xen Hypervisor for ARMv8 - Cavium ThunderX Soc - Manish Jaggi, Cavium Networks & Vijaya Kilari, Cavium Networks
ARMv8 architecture provide the support for running server class applications. One of the requirement is assigning devices directly to a domain with minimum interrupt injection latency.
With ARM IPs like GICv3 / ITS / SMMUv2 (IOMMU), assigning a PCI device (MMIO + MSIX) directly to a domU is possible.
This talk will focus on the architecture of these IPs on the 48core ARMv8 Cavium ThunderX SoC and the support added in Xen hypervisor to provide PCI passthrough and SRIOV functionality. Also discussed is the GICV3 and ITS emulation support added for running ARMv8 guest domain in Xen.

Speakers
avatar for Manish Jaggi

Manish Jaggi

ManishJaggi currently working as a Technical Lead at Cavium Networks for Virtualisation software. Currently working on Xen and KVM for ARM v8.x ThunderX Soc family.  In past Founded FreeTechMind an arm virtualization software company that developed type 1 hypervisor for arm cortexA15... Read More →
VK

Vijaya Kilari

Vijaya Kilari currently working as a Technical Lead at Cavium Networks for Virtualisation software. Currently working on Xen and KVM for ARM v8.x ThunderX Soc family. Vijay has been an active contributor to gicv3 its code in xen.


Tuesday August 18, 2015 9:00am - 9:30am PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

9:30am PDT

Improve ARM Guest Performance with 64KB Pages - Julien Grall, Citrix
The ARMv8 architecture adds support of 64KB granularity allowing only two levels of page tables and faster TLB look-up. While ARMv8 still support 4KB granularity, some major distributions will only ship their kernel with 64KB pages.

During this session, we will explain the meaning of 64KB granularity from the hardware and software perspective. We will also explain the impact to the Xen hypercall interface and a first approach to allow a 64KB pages kernel running on the current interface.

Last but not least, we will discuss changes to take advantage of 64KB granularity in PV drivers interfaces and gain better performance.

Speakers
avatar for Julien Grall

Julien Grall

Citrix
Julien Grall is a Software Engineer at Citrix, working on the Open Source Xen Platform team. He has been working on Xen since 2012, focusing at the beginning on x86 port, then on ARM port. Before joining Citrix, he was a student on Embedded System at Epita.


Tuesday August 18, 2015 9:30am - 10:00am PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

9:30am PDT

XenServer Power Saver - Cheng Zhang, Citrix
Power, memory and a CPU are the keys to maintaining high performance for XenServer. By creating a reasonable power management policy for Xen Hypervisor we are able to reclaim resources in idle VMs to improve both power saving and increase resources to other VM’s. In this presentation, Cheng Zhang will discuss the current status of power management in XenServer and ideas on how to use power more efficiently to increase the performance of higher priority VM’s.

Speakers
avatar for Cheng Zhang

Cheng Zhang

Citrix
Cheng Zhang has more than 9 years of experience working in software engineer. He is now working for XenServer for more than 3 year. Cheng has very good understanding on both software engineer and architecture and currently working for XenServer Livepatch integration and new packaging... Read More →


Tuesday August 18, 2015 9:30am - 10:00am PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

10:30am PDT

Getting U-Boot FIT for Xen - Robbie VanVossen & Paul Skentzos, DornerWorks
The Flattened Image Tree (FIT) image is a single formatted binary that supports multiple images and configurations that U-Boot understands. This format allows U-boot to perform integrity checks on the images, load a single image, and select from multiple configurations. The FIT format originally only allowed for loading a single kernel, a single ramdisk, and a single Flattened Device Tree (FDT) from the items in the image. But Xen requires two kernels: one for the Xen hypervisor and one for the Dom0 OS. So until now Xen users could not take advantage of the multiple benefits of the FIT image. Robert VanVossen will give an overview of the modifications DornerWorks made to U-Boot, allowing the FIT format to support loading generic images. He will further explain the benefits this brings to the Xen community and provide some examples of generating and using a FIT image to boot a Xen system.

Speakers
PS

Paul Skentzos

Paul Skentzos is currently a software engineer in DornerWorks’ hypervisor group. While at DornerWorks, he also spent time developing the technical and business goals for the hypervisor. Previously he was the lead engineer at InterMet Systems and Applied Analytics, where he developed... Read More →
RV

Robbie VanVossen

Embedded Systems Engineer, DORNERWORKS
I am an embedded systems engineer at DornerWorks in Grand Rapids, Michigan. I have done work with ARINC653 extensions for the Xen Hypervisor. I am also involved with providing support for Xen on the Xilinx Zynq Ultrascale+ MPSoC.I co-presented at the 2014 Xen Developer's Summit.


Tuesday August 18, 2015 10:30am - 11:00am PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

10:30am PDT

Shared Virtual Memory (SVM) Support in Xen - Feng Wu, Intel
For devices supporting appropriate PCI-Express capabilities, OS may use the DMA remapping Hardware capabilities to share virtual address space of application processes with I/O devices. Shared Virtual Memory (SVM) along with support for I/O page-faults enables application programs to freely pass arbitrary data-structures to devices such as graphics processors or accelerators, without the overheads of pinning and marshalling of data.

For a guest with an Intel GPU assigned, we need to advertise the SVM capability so that it can use the feature and gain performance improvement by sharing the GPU and CPU page tables.

In this presentation, Feng will talk about the mechanism of SVM support in Intel VT-d hardware and its advantages, as well as high-level design options for SVM in Xen.

Speakers
FW

Feng Wu

Intel
I am an employee of Intel. Intel Corporation is an American multinational semiconductor chip maker corporation headquartered in Santa Clara, California. Intel is one of the world's largest and highest valued semiconductor chip makers, based on revenue. It is the inventor of the x86... Read More →


Tuesday August 18, 2015 10:30am - 11:00am PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

11:00am PDT

How to Passthrough Your Integrated Device to a VM on ARM - Julien Grall, Citrix
The upcoming Xen release will add support for ARM device assignment.

Device passthrough allows the user to give control of physical devices (NIC, graphic card, GPS, etc) to a virtual machine, giving it full and direct access to the device. This has several potential uses, including avoiding overhead of the device para-virtualization.

On ARM based SOCs, most devices (NIC, graphic card, etc) are directly integrated on the SOC and not discoverable. This make passthrough less intuitive than with PCI devices.

During this session we will explain what is the approach taken and show the different steps to passthrough an integrated device. We will also discuss possible improvements for a more generic solution.

Speakers
avatar for Julien Grall

Julien Grall

Citrix
Julien Grall is a Software Engineer at Citrix, working on the Open Source Xen Platform team. He has been working on Xen since 2012, focusing at the beginning on x86 port, then on ARM port. Before joining Citrix, he was a student on Embedded System at Epita.


Tuesday August 18, 2015 11:00am - 11:30am PDT
Ravenna C Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

11:00am PDT

PVHVM Linux Guest: Why Doesn't Kexec Work? - Vitaly Kuznetsov, RedHat
Kexec (and kdump) are valuable multi-purpose mechanisms in Linux which are currently unavailable to Xen guests. What makes Xen guests so special, how is it different from other hypervisors? Can we solve these issues in a foreseeable future? This presentation will give an overview of several different approaches to making kexec possible for PVHVM Linux guests and describe difficulties we face while trying to implement these approaches. I will show which changes should be made to the hypervisor, toolstack, and Linux kernel in order to make things work. While the talk mainly focuses on PVHVM x86 guests I will also try to mention PVH and ARM implications.

Speakers
avatar for Vitaly Kuznetsov

Vitaly Kuznetsov

Principal Software Engineer, Red Hat
Software Engineer


Tuesday August 18, 2015 11:00am - 11:30am PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

11:30am PDT

Xen Containers: Better Way to Run Docker Containers - Sainath Grandhi, Intel
The container technologies, including Docker containers are being widely adopted for packaging and deploying applications. Limitations with today’s bare-metal containers are security and performance isolation as containers share the same host operating system.
We propose “Xen containers” to run Docker container applications sandboxed in a small VM as an alternative to bare-metal containers. Xen containers in PVH, for example, provide tighter security and resource isolation. By using minimalist VMs that spin up in sub-seconds and consume couple of tens of MBs of memory, the advantages of the Xen containers outweigh (negligible) overhead of virtualization. We compare the bare-metal containers and Xen containers in various aspects in the presentation.

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.


Tuesday August 18, 2015 11:30am - 12:20pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:00pm PDT

Scheduling in Xen: Past, Present and Future - Dario Faggioli, Citrix & Meng Xu, University of Pennsylvania
The Xen-Project hypervisor provides a set of schedulers, each one suited for specific use cases. In this talk, we will discuss what they are, how  they work and what features they offer. The unmatched flexibility that Xen 4.5 offers when it comes to allocate physical processor to virtual processors will be showcased, and insights about the ongoing development
aimed at making the Credit2 scheduler ready for being used in production will be given.

We will also focus on the benefits (better average performance) and the drawbacks (less predictable behavior) of shared cache in multicore and multisocket systems. Proposals for hardware and software based approaches for partitioning the cache, and hence avoid the various virtual machine to interfere among each others, will be illustrated and discussed.

Speakers
avatar for Dario Faggioli

Dario Faggioli

Virtualization Engineer, SUSE
Dario is a Virtualization Software Engineer at SUSE. He's been active in the Open Source virtualization space for a few years. Within the Xen-Project, he is still the maintainer of the Xen hypervisor scheduler. He also works on Linux kernel, KVM, Libvirt, and QEMU. Back during his... Read More →
avatar for Meng Xu

Meng Xu

PhD Student, Apple Inc.
Meng Xu is a 3rd year PhD student in the PRECISE Center at the University of Pennsylvania. He is working on the Real-Time Xen Project, which aims to bridge the gap between real-time scheduling theory and virtualization technology. He is the contributor of the RTDS real-time scheduler... Read More →


Tuesday August 18, 2015 2:00pm - 2:50pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:00pm PDT

Joint Xen/KVM Hackathon
Tuesday August 18, 2015 2:00pm - 6:00pm PDT
Virginia Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

2:50pm PDT

Project Raisin - Stefano Stabellini, Citrix
As part of the process of building Xen, the current build system clones multiple external git trees and compile them. For example Seabios and QEMU are imported using this mechanism. The process is unfriendly to Linux distributions, which have the requirement of building each component just once, and is unfriendly to users, who often find it confusing and inflexible. A new tool was created to solve this problem.

This talk will introduce Raisin: a source distribution system for the Xen Project. Raisin offers a generic framework to deploy any Xen related projects from source. Currently it supports QEMU, SeaBIOS, OVMF, Linux, Libvirt and blktap, but is very easy to extend. Raisin also helps developers by providing quick validation tests that can be run against local changes.

The presentation will explain the goals of the project, the current status, and the best ways to use it.

Speakers
avatar for Stefano Stabellini

Stefano Stabellini

Principal Engineer, Xilinx
Stefano Stabellini serves as system software architect and virtualization lead at Xilinx, the world's largest supplier of FPGA solutions. Previously, at Aporeto, he created a virtualization-based security solution for containers and authored several security articles. As Senior Principal... Read More →


Tuesday August 18, 2015 2:50pm - 3:20pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

4:00pm PDT

Deploying Real-World Software Today as Unikernels on Xen with Rumprun - Martin Lucina
Rump kernels were named by the Xen project as one of the "7 Unikernel Projects to take on Docker in 2015". On top of rump kernels we have built the rumprun software stack, which runs real-world software such as Nginx, PHP, and MySQL as unikernels on Xen, today.

In his presentation, Martin will demonstrate the ease of deploying unmodified, existing applications as unikernels on Xen with rumprun.

He will present his experience and lessons learned over the last few months from "dogfooding" rumprun by migrating existing production services from a traditional monolithic Linux system to a set of cooperating unikernels, and discuss the challenges involved and future work needed to address them.

Going forward, he will make the case that more collaboration between unikernel projects will enable interesting deployments with benefits to users and developers.

Speakers
avatar for Martin Lucina

Martin Lucina

rumpkernel.org
Martin has been programming since before programming was trendy, eating Sharp SC61860A machine code for breakfast since before it was healthy, and installed his first Linux system from scratch in 1992 back when it was just Linus Torvalds' glorified terminal emulator.Currently, as... Read More →


Tuesday August 18, 2015 4:00pm - 4:50pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

4:50pm PDT

Keeping Up with the Hardware: Challenges in Scaling I/O Performance - Jonathan Davies, Citrix
Commodity disks and NICs are set to make order-of-magnitude increases in performance in the near future. As this happens, the limitations and overhead of virtualising these devices becomes increasingly dominant.

This presentation will discuss the areas of network and storage I/O performance in which the Xen Project hypervisor is weakest. In each area,
* reasons will be suggested for why performance is poor;
* ideas to improve performance will be surveyed (including existing prototypes and future development work); and
* measurements for prototypes that exist will be presented.

Speakers
avatar for Jonathan Davies

Jonathan Davies

XenServer Performance Team, Citrix
Jonathan is a Principal Software Engineer at Citrix where he leads XenServer's Performance Team. This team has oversight of the performance and scalability of all aspects of XenServer. Current projects are focussed on improving XenServer's storage and network performance.


Tuesday August 18, 2015 4:50pm - 5:20pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101

5:20pm PDT

A Fault Tolerant Virtualization Server Based on Xen - Jürgen Groß, SUSE
A fault tolerant virtualization server based on Xen (Juergen Gross, SUSE) - Virtualization servers are subject to high availability demands as a malfunction would influence many systems at once. This requests virtualization solutions with as few single points of failure as possible. In this technical talk Juergen Gross will show how this aim can be reached with Xen by making use of Xen's architectural benefit of being a type 1 hypervisor. He will suggest a system architecture being much more tolerant against hardware or software failures compared to currently available solutions.

Speakers
avatar for Jürgen Groß

Jürgen Groß

Principal Developer, SUSE
Contributing to Xen since 2007, maintainer of several Xen components.


Tuesday August 18, 2015 5:20pm - 5:50pm PDT
Ravenna AB Sheraton Seattle, 1400 6th Ave., Seattle, WA 98101
 
Filter sessions
Apply filters to sessions.