The Xen Hypervisor was built for the Cloud from the outset: when Xen was designed, we anticipated a world, which today is known as cloud computing. Today, Xen powers the largest clouds in production.
This talk explores success criteria, architecture, trade-offs and challenges for cloudy hypervisors. It is intended for users and developers and starts with a brief introduction to Xen and XCP, their architecture and on common challenges for KVM and Xen.
I will introduce the concept of domain disaggregation as an approach to increase security, robustness and scalability: all important factors for building clouds at scale and show how advanced security features suchas Xen Security Modules and SELinux can help secure your cloud further.
The talk will conclude with exciting developments in the Xen community, such as Xen for ARM servers, a new virtualization mode for Xen, running applications without OS in a Xen guest and point out their implications for building open source clouds.
DevEX - reference for building teams, processes, and platforms
Scale11x : Virtualization with Xen and XCP
1. Virtualization in the Cloud:
Featuring Xen and XCP
Lars Kurth
Xen Community Manager
lars.kurth@xen.org
FREENODE: lars_kurth @lars_kurth
2. A Brief History of Xen in the Cloud
Late 90s
XenoServer
Project
3. A Brief History of Xen in the Cloud
Late 90s ‘03
XenoServer
Project
Xen 1.0
4. A Brief History of Xen in the Cloud
Late 90s ‘03 ‘06 ‘08
XenoServer Amazon EC2
Project and Slicehost
launched
Xen 1.0 Rackspace
Cloud
5. A Brief History of Xen in the Cloud
Late 90s ‘03 ‘06 ‘08 ‘11 ‘12
XenoServer Amazon EC2 XCP 1.x
Project and Slicehost Cloud Mgmt
launched
Xen 1.0 Rackspace
Cloud
XCP packages
in Linux
6. A Brief History of Xen in the Cloud
Late 90s ‘03 ‘06 ‘08 ‘11 ‘12
XenoServer Amazon EC2 XCP 1.x
Project and Slicehost Cloud Mgmt
launched
Xen 1.0 Rackspace Linux 3.0
Cloud
XCP packages
in Linux
7. A Brief History of Xen in the Cloud
Late 90s ‘03 ‘06 ‘08 ‘11 ‘12 ‘13
XenoServer Amazon EC2 XCP 1.x
Project and Slicehost Cloud Mgmt
launched
Xen 1.0 Rackspace Linux 3.0 Xen for
Cloud ARM servers
XCP packages
in Linux
10th
birthday
9. Xen.org
• Guardian of Xen Hypervisor and related OSS Projects
• Xen Governance similar to Linux Kernel
– Plus project lifecycle and Project Management Committee (PMC)
• Projects
– Xen Hypervisor
(led by 5 committers, 2 from Citrix, 1 from Suse, 2 Independent)
– Xen Cloud Platform aka XCP (led by Citrix)
– Xen ARM : Xen for mobile devices (led by Samsung)
10. Xen contributor community is diversifying
100%
90%
• The number of “significant”
80% active vendors is increasing
70%
60%
50% • New feature development driving
40%
30%
new participation
20%
10%
0%
2010 2011 2012
Citrix UPC
SUSE Amazon
University AMD
GridCentric Individual
NSA Intel
Fujitsu iWeb
Misc Oracle
Spectralogic University of British Columbia
12. Hypervisor Architectures
Type 1: Bare metal Hypervisor
A pure Hypervisor that runs directly on the
hardware and hosts Guest OS’s.
VMn
VM1
VM0
Guest OS
and Apps
Scheduler Hypervisor
Device Drivers/Models MMU
Host HW
I/O Memory CPUs
Provides partition isolation + reliability,
higher security
13. Hypervisor Architectures
Type 1: Bare metal Hypervisor Type 2: OS ‘Hosted’
A pure Hypervisor that runs directly on the A Hypervisor that runs within a Host OS and hosts
hardware and hosts Guest OS’s. Guest OS’s inside of it, using the host OS services
to provide the virtual environment.
VMn User-level VMM VMn
VM1 User
VM1
Apps
VM0 Device Models
VM0
Guest OS Guest OS
and Apps and Apps
Host OS
Scheduler Hypervisor
Ring-0 VM Monitor
Device Drivers/Models Device Drivers “Kernel “
MMU
Host HW Host HW
I/O Memory CPUs I/O Memory CPUs
Provides partition isolation + reliability, Low cost, no additional drivers
higher security Ease of use & installation
14. Xen: Type 1 with a Twist
Type 1: Bare metal Hypervisor
VMn
VM1
VM0
Guest OS
and Apps
Scheduler Hypervisor
Device Drivers/Models MMU
Host HW
I/O Memory CPUs
15. Xen: Type 1 with a Twist
Type 1: Bare metal Hypervisor Xen Architecture
VMn
VM1 VMn
VM0 VM1
Guest OS VM0
and Apps
Guest OS
and Apps
Scheduler Hypervisor
Device Drivers/Models MMU Scheduler MMU Hypervisor
Host HW Host HW
I/O Memory CPUs I/O Memory CPUs
16. Xen: Type 1 with a Twist
Type 1: Bare metal Hypervisor Xen Architecture
Control domain
(dom0)
VMn
VM1 Device Models VMn
VM0 VM1
Guest OS VM0
and Apps Drivers
Guest OS
Linux & BSD and Apps
Scheduler Hypervisor
Device Drivers/Models MMU Scheduler MMU Hypervisor
Host HW Host HW
I/O Memory CPUs I/O Memory CPUs
17. Xen and Linux
• Xen Hypervisor is not in the Linux kernel
• BUT: everything Xen and Xen Guests need to run is!
• Xen packages are in all Linux distros (except RHEL6)
– Install Dom0 Linux distro
– Install Xen package(s) or meta package
– Reboot
– Config stuff: set up disks, peripherals, etc.
More info: wiki.xen.org/wiki/Category:Host_Install
18. Basic Xen Concepts
Console
• Interface to the outside world
Control Domain aka Dom0
VMn
• Dom0 kernel with drivers
Control domain VM1
(dom0) • Xen Management Toolstack
VM0
Guest OS
Guest Domains
Dom0 Kernel and Apps • Your apps
Scheduler MMU XSM Hypervisor Driver/Stub/Service Domain(s)
• A “driver, device model or control
Host HW service in a box”
I/O Memory CPUs
• De-privileged and isolated
• Lifetime: start, stop, kill
Trusted Computing Base 18
19. Basic Xen Concepts
Console
Console • Interface to the outside world
Control Domain aka Dom0
VMn
• Dom0 kernel with drivers
Control domain VM1
(dom0) • Xen Management Toolstack
VM0
Toolstack
Guest OS
Guest Domains
Dom0 Kernel and Apps • Your apps
Scheduler MMU XSM Hypervisor Driver/Stub/Service Domain(s)
• A “driver, device model or control
Host HW service in a box”
I/O Memory CPUs
• De-privileged and isolated
• Lifetime: start, stop, kill
Trusted Computing Base 19
20. Basic Xen Concepts
Console
Console • Interface to the outside world
Control Domain aka Dom0
VMn
• Dom0 kernel with drivers
Control domain VM1
(dom0) • Xen Management Toolstack
One or more VM0
Toolstack driver, stub or
service domains Guest OS
Guest Domains
Dom0 Kernel and Apps • Your apps
Scheduler MMU XSM Hypervisor Driver/Stub/Service Domain(s)
• A “driver, device model or control
Host HW service in a box”
I/O Memory CPUs
• De-privileged and isolated
• Lifetime: start, stop, kill
Trusted Computing Base 20
22. Xen Variants for Server & Cloud
Hypervisor Xen
Toolstack / Console Default / XL (XM) Libvirt / VIRSH XAPI / XE
Increased level of functionality and integration with other components
Single Host Single Host
Basic Functions Additional Functionality
Multiple Hosts
Additional Functionality
22
23. Xen Variants for Server & Cloud
Hypervisor Xen XCP
Toolstack / Console Default / XL (XM) Libvirt / VIRSH XAPI / XE
Increased level of functionality and integration with other components
Single Host Single Host
Basic Functions Additional Functionality
Multiple Hosts
Additional Functionality
23
24. Xen Variants for Server & Cloud
Project Xen XCP
Toolstack / Console Default / XL (XM) Libvirt / VIRSH XAPI / XE
Increased level of functionality and integration with other components
Get Binaries from … Linux Distros Linux Distros Debian & Ubuntu
ISO from Xen.org
24
25. Xen Variants for Server & Cloud
Project Xen Hypervisor XCP
Toolstack / Console Default / XL (XM) Libvirt / VIRSH XAPI / XE
Increased level of functionality and integration with other components
Get Binaries from … Linux Distros Linux Distros Debian & Ubuntu
ISO from Xen.org
Products Oracle VM Huawei UVP Citrix XenServer
25
26. Xen Variants for Server & Cloud
Project Xen Hypervisor XCP
Toolstack / Console Default / XL (XM) Libvirt / VIRSH XAPI / XE
Increased level of functionality and integration with other components
Get Binaries from … Linux Distros Linux Distros Debian & Ubuntu
ISO from Xen.org
Used by …
More info: xen.org/community/ecosystem.html
xen.org/community/presentations.html
26
xen.org/products/case_studies.html
28. PV Domains
Technology:
Control domain Guest VMn
(dom0) • Paravirtualization
Apps Linux PV guests have limitations:
• limited set of virtual hardware
PV Back Ends PV Front Ends
Advantages
HW Drivers • Fast
Dom0 Kernel Guest OS
• Works on any system
(even without virt extensions)
Xen Hypervisor
Host HW
I/O Memory CPUs
28
29. PV Domains & Driver Domains
Technology:
Control domain Guest VMn Driver Domain
(dom0) e.g. • Paravirtualization
• Disk
Apps
• Network
Linux PV guests have limitations:
• limited set of virtual hardware
PV Back Ends PV Front Ends PV Back End
Advantages
HW Drivers HW Driver • Fast
Dom0 Kernel Guest OS Dom0 Kernel*
• Works on any system
(even without virt extensions)
Xen Hypervisor Driver Domains
• Security
Host HW • Isolation
I/O Memory CPUs
• Reliability and Robustness
*) Can be MiniOS
29
30. HVM & Stub Domains
Technology:
Dom0 Guest VMn
• Shows emulation using QEMU/Device
Model
IO Emulation
(SW Virtualization)
Device Model • In other situation HW can be used
Disadvantages
IO Event
• Emulation slower than PV
Dom0 Kernel VMEXIT (mainly I/O devices)
Xen Hypervisor Advantages
• No kernel support needed
Host HW
I/O Memory CPUs
30
31. HVM & Stub Domains
Technology:
Dom0 Guest VMn Stubdomn Guest VMn
• Shows emulation using QEMU/Device
Model
IO Emulation IO Emulation
(SW Virtualization)
Device Model Device Model • In other situation HW can be used
Disadvantages
IO Event IO Event
• Emulation slower than PV
Dom0 Kernel VMEXIT Mini OS VMEXIT (mainly I/O devices)
Xen Hypervisor Advantages
• No kernel support needed
Host HW Stub Domains
I/O Memory CPUs
• Security
• Isolation
• Reliability and Robustness
31
32. The Virtualization Spectrum
VS Virtualized (SW)
VH Virtualized (HW)
P Paravirtualized
Fully Virtualized (FV) VS VS VS VH
FV with PV for disk & network P VS VS VH HVM mode/domain
PVHVM P P VS VH
PVH Xen 4.3 P P P VH
PV mode/domain
Fully Paravirtualized (PV) P P P P
33. The Virtualization Spectrum
Optimal performance
Scope for improvement
Poor performance
Fully Virtualized (FV) VS VS VS VH
FV with PV for disk & network P VS VS VH HVM mode/domain
PVHVM P P VS VH
PVH Xen 4.3 P P P VH
PV mode/domain
Fully Paravirtualized (PV) P P P P
34. The Virtualization Spectrum
Important: Xen automatically picks the best
option based on HW & OS capabilities and
Optimal performance available drivers.
As a Xen user I chose a HVM or PV domain.
Scope for improvement
Poor performance
Fully Virtualized (FV) VS VS VS VH
FV with PV for disk & network P VS VS VH HVM mode/domain
PVHVM P P VS VH
PVH Xen 4.3 P P P VH
PV mode/domain
Fully Paravirtualized (PV) P P P P
36. XCP – Xen Cloud Platform
Complete stack for server virtualization
• Extends Xen to cover multiple hosts
• Adds further functionality and integrations
for cloud, storage and networking to Xen HV
• GPLv2
• XenServer is a commercial XCP distro
Two Flavours
• Appliance (ISO using CentOS Dom0)
• Packages in Debian & Ubuntu
(more distros to come)
37. Major XCP Features
• VM lifecycle: live snapshots, checkpoint, migration
• Resource pools: flexible storage and networking
• Event tracking: progress, notification
• Upgrade and patching capabilities
• Real-time performance monitoring and alerting
• Built-in support and templates for Windows and Linux guests
• Open vSwitch support built-in (default)
More info: wiki.xen.org/wiki/XCP_Release_Features
38. XCP 1.6
• New format Windows drivers:
installable by Windows Update Service
• Networking: Better VLAN scalability, LACP bonding, IPv6
• Storage XenMotion:
– Migrate VMs between hosts or pools without shared storage
– Move a VM’s disks between storage repositories while the VM is running
• Other: more templates, latest Xen, OVS, etc.
More info: xen.org/download/xcp/releasenotes_1.6.0.html &
More info: xen.org/download/xcp/index_1.6.0.html
42. System characteristics cloud users care about:
“Robustness, Performance, Scalability & Security”
Results XCP User Survey 2012 – 90% of users quoted these as most important attributes
43. Disaggregation
Split Control Domain into Driver,
Stub and Service Domains
– See: ”Breaking up is hard to do” @ Xen Papers
– See: “Domain 0 Disaggregation for XCP and XenServer”
Used today by Qubes OS and Citrix XenClient XT
Prototypes for XCP
See qubes-os.org
Different windows run
in different VMs
44. Benefits of Disaggregation
More Security
Increased serviceability and flexibility
Better Robustness
Better Performance Ability to safely restart parts of the system
(e.g. just 275ms outage from failed Ethernet driver)
Better Scalability
46. User VM User VM
NF BF NF BF
NB gntdev NB gntdev gntdev
Dom0 Network NFS/
Dom0 Qemu xapi Qemu Network NFS/ Local
drivers iSCSI drivers iSCSI storage
Domain
manager
drivers . drivers drivers
qemu
. qemu
healthd storaged storaged storaged
networkd . networkd
xenopsd tapdisk syslogd . tapdisk tapdisk
libxl vswitch blktap3 vswitch blktap3 blktap3
xapi xapi
eth eth
Dom0 eth eth scsi
Xen Xen
CPU CPU
NIC NIC RAM RAM NIC NIC
(or SR- (or SR- (or SR- (or SR- RAID
IOV VF) IOV VF) IOV VF) IOV VF)
47. User VM User VM
NF BF NF BF
NB gntdev NB gntdev gntdev
Dom0 Network NFS/ Qemu xapi D Logging Network NFS/ Local
driver iSCSI domain domain domain driver iSCSI storage
Domain o
manager
domain driver
m
. domain driver driver
domain domain domain
healthd storaged
qemu 0 . storaged storaged
networkd networkd
xenopsd tapdisk
. tapdisk tapdisk
libxl vswitch blktap3 xapi syslogd vswitch blktap3 blktap3
dbus over v4v dbus over v4v
eth eth eth eth scsi
Xen Xen
CPU CPU
NIC NIC RAM RAM NIC NIC
(or SR- (or SR- (or SR- (or SR- RAID
IOV VF) IOV VF) IOV VF) IOV VF)
48. Xen Security Advantages
• Even without Advanced Security Features
– Well-defined trusted computing base (much smaller than on type-2 HV)
– Minimal services in hypervisor layer
• Xen Security Modules (or XSM) and FLASK
– XSM is Xen equivalent of LSM
– FLASK is Xen equivalent of SELinux
– Developed, maintained and contributed to Xen by NSA
– Compatible with SELinux (tools, architecture)
– XSM object classes maps onto Xen features
More info: http://www.slideshare.net/xen_com_mgr/
a-brief-tutorial-on-xens-advanced-security-features
50
49. User VM User VM
NF BF NF BF
NB gntdev NB gntdev gntdev
Dom0 Network NFS/ Qemu xapi D Logging Network NFS/ Local
driver iSCSI domain domain domain driver iSCSI storage
Domain o
manager
domain driver
m . domain driver driver
domain domain domain
healthd storaged
qemu 0 . storaged storaged
networkd networkd
xenopsd tapdisk
. tapdisk tapdisk
libxl vswitch blktap3 xapi syslogd vswitch blktap3 blktap3
dbus over v4v dbus over v4v
eth eth eth eth scsi
FLASK policy
restricting access Xen Xen
CPU CPU
NIC NIC RAM RAM NIC NIC
(or SR- (or SR- (or SR- (or SR- RAID
IOV VF) IOV VF) IOV VF) IOV VF)
51. Coming in Xen 4.3 (Q2 2013)
• PVH virtualization mode
• Extend scope of Xen Security Modules
• qxl Spice support for 3d acceleration
• Updated and improved libvirt drivers for Xen
• Lots of other stuff:
– scalability, performance, better NUMA support, …
More info: blog.xen.org/index.php/2013/02/11/xen-4-3-mid-release-roadmap-update
52. Xen 4.3 for ARM Servers
Fully functional for ARM v7 & v8
ARM v7: Versatile Express, Arndale &
Samsung Chromebook
ARM v8: Fast Model
53. Xen and ARM : a perfect Match
ARM SOC ARM Architecture Features for Virtualization
User mode : EL0
Device Tree describes …
Kernel mode : EL1
I/O
Hypercall interface :HVC
GIC 2 stage
GT
v2 MMU Hypervisor mode : EL2
54. Xen and ARM : a perfect Match
ARM SOC ARM Architecture Features for Virtualization
Dom0 Any Xen Guest VM (including Dom0)
only EL0
Device Tree describes … User Space
Kernel
EL1
I/O
HVC
GIC 2 stage
GT
v2 MMU EL2
Xen Hypervisor
55. One mode to rule them all
Optimal performance
Scope for improvement
x86: PVHVM P P VS VH HVM mode/domain
x86: PVH P P P VH PV mode/domain
ARM v7 & v8 P VH VH VH
56. Xen in CentOS 6.4+
Xen is coming back to CentOS
In semi-private beta
Planned release in CentOS 6.4
Include XAPI packages – aka XCP in CentOS
57. Xen Library Operating Systems
Application stacks only running on Xen APIs
Control domain Guest VMn
Works on any Xen based cloud or hosting service (dom0)
Apps
Examples
PV Back Ends
– ErlangOnXen.org : Erlang Library OS
embedded
– HalVM : Haskell HW Drivers in Language
run-time
– OpenMirage : Ocaml Dom0 Kernel
Benefits: Xen
– Small footprint Host HW
– Low startup latency
– Extremely fast migration of VMs
59. • Designed for the Cloud : many advantages for cloud use!
– Resilience, Robustness & Scalability
– Security: Small surface of attack,
Isolation & Advanced Security Features
• Widely used by Cloud Providers and Vendors
• XCP
– Ready for use with cloud orchestration stacks
• Open Source with a large community and eco-system
– Xen is still on top of the game
– Exciting new developments and features in the pipeline
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The release ofLinux 3.0 was a key milestone for Xen. Linux 3.0 was the first Linux Kernel release that had …Full support for Xen as Guest as well as Domain 0 operation system (out-of-the-box) This was subsequently followed by distro support for Xen as Guest and Dom 0
The key point about the history run-down however is thatXen was designed from the outset with cloud computing in mindAnd the continued adoption and use of Xen by the largest cloud providers just underlines that point
Key points not on the slides:Governance: same roles, same license, DCO, …Vendors: the link shows a breakdown of contributions to Xen 4.2
Color code: green + bright green = trusted computing baseType 1 hypervisor: example ESXSimple architecture; Hypervisor replaces the kernel entirelyNo need to provide rich “process” semantics, like “user”, filesystems, etc.BUT: Device drivers and Device Models need to be rewritten for each hardware platformThis meansisolation and higher security by designis traded off against maintainability and re-use of driversType 2 is hosted: example KVM- The hypervisor is just a driver that works with user-level monitor and the host OS kernelHW access is intercepted by the ring 0- VM monitor passed to the User level Virtual Monitor, which passes requests to the kernelThis meansRe-use of device drivers and host OS kernel infrastructureis traded off against security and a large trusted computing base
Color code: green + bright green = trusted computing baseType 1 hypervisor: example ESXSimple architecture; Hypervisor replaces the kernel entirelyNo need to provide rich “process” semantics, like “user”, filesystems, etc.BUT: Device drivers and Device Models need to be rewritten for each hardware platformThis meansisolation and higher security by designis traded off against maintainability and re-use of driversType 2 is hosted: example KVM- The hypervisor is just a driver that works with user-level monitor and the host OS kernelHW access is intercepted by the ring 0- VM monitor passed to the User level Virtual Monitor, which passes requests to the kernelThis meansRe-use of device drivers and host OS kernel infrastructureis traded off against security and a large trusted computing base
Thinner hypervisorOnly Contains: Scheduler and Memory Management
Thinner hypervisorOnly Contains: Scheduler and Memory Management
Domain 0:Special Virtual Machine that co-operates with the HypervisorContains a Xen enabled Kernel: Linux 3.x or NetBSDUses Device Drivers and Device Models (QEMU)For Linux: Using Linux PV OPSExposes support for various virtualization modes: PV, PV on HVM and PVH modesSharing components with KVM (e.g. QEMU based Device Drivers)
As we talked a little bit about Xen and Linux already it is worth point out a few things to Xen and LinuxSome people argue that KVM is better than Xen, because KVM is in the kernelThis argument is bogusSoftware should be compared on the basis of features and system characteristicsNOT on where the source code livesSo, although Xen is NOT in the Linux KernelBUT from linux 3.0 onwards EVERYTHING that Xen and Xen Guests need to run isAnd all modern Linux distros (with the exception of RHEL6) contain Xen packagesTHIS MEANS:The user experience for Linux users for Xen and KVM is VERY similarYou choose your favourite distro and install (that is true for Xen and KVM)You download and install the Xen packages (KVM-QEMU package) via your package managerTHE only difference is that for Xen you reboot, after which Xen runs underneath your Dom 0Then you configure your systemThe point is: from a users perspective it doesn’t matter whether Xen is in Linux or not, as long as distros contain Xen packages
This is the architecture that we have explained beforeLet’s add a few more things!
Added theToolstackExposes an RPC type interface such that the system can be controlled from the outside worldXen has a few options for toolstacks (see later)Added the ConsoleA CLIA UIOr an interface to a cloud orchestration stack Controls the system (for operations such as starting, stopping VMs, etc.)
A unique feature of Xen is called disaggregationThis means:I can take a Device DriverA Device ModelA service (e.g. the toolstack)That normally runs in Domain 0 and run it in a separate VM.That VM is de-priviliged (it is not part of the trusted computing base)It can be restarted (if needed) without affecting the rest of the systemThis is something unique to Xen’s architecture. More later.
On Xen.org, you will come by two main projects:The Xen HypervisorThe Xen Cloud PlatformAt a very high level, we will you show you the various options to build a system.
Let’s first look at choices of Xen Management Toolstacks:The Default with XL as CLI console Single Host, with basic functionalityLibvirt with VIRSH as CLI console (can also use virt-manager)Multiple Hosts, more functionalityXAPI (or Xen API) and XE as CLI (a number of UI options available)Multiple Hosts, even more functionality than LibvirtTypically when a Xen system contains XAPI, we call it Xen Cloud Platform
Let’s first look at choices of Xen Management Toolstacks:The Default with XL as CLI console Single Host, with basic functionalityLibvirt with VIRSH as CLI console (can also use virt-manager)Multiple Hosts, more functionalityXAPI (or Xen API) and XE as CLI (a number of UI options available)Multiple Hosts, even more functionality than LibvirtTypically when a Xen system contains XAPI, we call it Xen Cloud Platform
Now lets look at where you get beinaries these from :1) Xen (with default toolstack) you get from distros2) Xen, libvirt drivers, libvirt and VIRSH you get from distros3) XAPI, you can get from Debian and Ubuntu (other distros to follow) OR, you can download an ISO from Xen.org
A number of different commercial products use these combinations:Oracle VM based on Xen, the default stack with additionsHuawei UVP, which uses Xen + LibvirtCitrix XenServer which uses Xen + XAPI
Lots of services use various of these combinationsWe picked out a few for these different configurationsAWS uses Xen with DefaultSuse Cloud uses Xen with LIBVIRTRackspace Cloud Servers uses XAPIThere are loads more large users: check out en.org/community/ecosystem.html If you want to know more about AWS and Rackspace, check out xen.org/community/presentations.html We also have some user case studies
Remember the concept of disaggregation I mentioned earlier?With Driver Domains, I can take an individual driver (e.g. disk or network) and run it in its own domain (aka VM)In terms of how PV works: it is exactly the same as with drivers running in Dom0, only that Xen knows that the driver runs in its own driver domain
Device Model emulated in QEMUModels for newer devices are much fasterBut for IO operations PV is even faster
Remember the concept of disaggregation I mentioned earlier and Driver Domains?Stub Domains: Run an individual device model (QEMU instance) in its own domain Again, these work identical to Device Models within a Dom0.For scalability there is one stub domain per VM
VM lifecycle (start, stop, resume) Automation is the key pointLive snapshots: Takes a snapshot of a live VM (e.g. for disaster recovery or migration)Resource pools (multiple physical machines): live migration: VM is backed up while running, onto shared storage (e.g. NFS). disaster recovery: the key point is that I can back up the metadata for the entire VMFlexible storage: XAPI does hide details for storage and networkingApply generic commands for different backends (NFS, NETAPP, iSCSI ... once its created they all appear the same) Only need to know the storage type when I create storage and network objects (OOL)Upgrading a host to a later version of XCP (all my configs and VMs stay the same)Apply security patches and upgrade on running systems
Just one example of a survey, many morehttp://www.colt.net/cio-research/z2-cloud-2.htmlAccording to many surveys, security is actually the main reason which makes or breaks cloud adoptionBetter security means more adoptionConcerns about security means slowed adoption
This is information from our own XCP survey (Q4 2012).Basically users were asked to rate system attributes according to importance on a scale of 1-5 90% of all users that use XCP for the cloud ratedRobustness, Performance, Scalability and Security at a 5Many other system attributes ranged from 3-4
We introduced disaggregation earlier under architectural concepts. To remind you, in a nutshell it means distributing system components running in Dom0 into separate VM’s.If you want to know more about the concept: There are the two papers listed above.The first takes disaggregation to the extreme!Disaggregation is in use by OSS and commercial products for Desktop Was originally developed for NSA:Every person with a specific security clearance has to have one terminal (desktop computer) per clearance levelTypically every employee had several TerminalsProducts using disaggregation are deemed secure enough by the NSA (and similar orgs) to Prototypes for server and cloud products
Good for SecurityReduces Trusted Computing BaseBetter defined interfaces (better for security audits)Can kill-start domains on a timer to wipe security risksIt is also good for robustness:You can basically restart parts of the system without affecting the rest of the system Latency for this is low (see the graph – in the order of 100 of ms)Incidentally this is also good for serviceability: in other words I can upgrade parts of my system while leaving others untouched and runningGood for performance and scalability:Removes Dom 0 as a bottleneck when many VM’s are runningPlaces more emphasis on the Xen scheduler vs. the Dom0 scheduler, which tends to lead to better performanceYou can chose better performing OS+driver combinations (e.g. BSD)
This shows a diagram of the type of services running in a normal XCP instance.The picture shows:1) Two Hosts2) User VMs per Host (on top instead of on the side due as in due to space constraints on the slide) 3) Dom0 with Xen per host (green boxes – also Trusted Computing base)4) Services and processes running within each Dom0Note: that inter-host communication is handled by the XAPI component Also note: this is an approximation as there is not enough space to cram everything in! There is stuff missing!NOW LETS LOOK WHAT DISAGGREGATION DOES!
This shows a diagram of the type of services running in the XCP disaggregation prototype.What is different:Key services are running in their own domainsEach host has a fast inter-domain communication mechanism called DBUS that allows direct VM to VM communicationThe Trusted Computing base has become a lot smaller (look at the green)
Now looked at disaggregation:It is worth to remind ourselves of existing Xen Security advantages: more modular, smaller TCBThere are also additional features, the most noteworthy is XSMDeveloped and maintained by the NSA (who contribute 3-4% of code per year to Xen)Follows the same architecture as SELinux (including compatibility with policy checking and dev tools for SELinux)Can be used together with SELinuxThe main difference is that XSM exposes objects that map to Xen featuresAlthough XSM alone is useful, it becomes REALLY INTERESTING when used with disaggregation!See WHY IN THE NEXT SLIDE!
The RED BOX, shows how we can use XSM and SELinux to secure our system.Per domain, I can restrict what can be started in a domainI can restrict what hypervisor interfaces (PV or hypercalls) a class of domain can access. In this example, I can ensure that only Linux and Xen functionality that is necessary to operate the Network Driver Domain can be accessed or launched
A unique feature of Xen is called disaggregationThis means:I can take a Device DriverA Device ModelA service (e.g. the toolstack)That normally runs in Domain 0 and run it in a separate VM.That VM is de-priviliged (it is not part of the trusted computing base)It can be restarted (if needed) without affecting the rest of the systemThis is something unique to Xen’s architecture. More later.
A unique feature of Xen is called disaggregationThis means:I can take a Device DriverA Device ModelA service (e.g. the toolstack)That normally runs in Domain 0 and run it in a separate VM.That VM is de-priviliged (it is not part of the trusted computing base)It can be restarted (if needed) without affecting the rest of the systemThis is something unique to Xen’s architecture. More later.
Performance : similar to other hypervisorsMaturity: Tried & Tested, Most Problems that are Problems are well knownOpen source: Good body of Knowledge, Tools