Lecture - 2 and 3 Virtualization, Orchestration and Design of Cloud Services
Salman Toor [email protected] Outline
• Introduction to virtualization • Cloud virtualization
• Importance • Demo
• Types • Virtualization beyond VMs
• Hardware support • Orchestration Tools
• Hypervisors • Design of Cloud services • libvirt API ?
2 Virtualization
• Nutshell: The abstraction of available resources
• Definition:
Virtualization technologies encompass a variety of mechanisms and techniques used to decouple the architecture and user-perceived behavior of hardware and software resources from their physical implementation.
• Whereas, resources can be either compute, storage, network..etc
http://www.computer.org/csdl/mags/co/2005/05/r5028.html 3 Virtualization Basic illustration
Virtualization layer
4 http://blog.cloudpassage.com/2011/07/22/the-difference-between-virtualization-and-cloud-computing/ Virtualization
• Large verity of platforms: https://en.wikipedia.org/wiki/Comparison_of_platform_virtualization_software
• Old concept, relaunched (article from 1974) http://cs.nyu.edu/courses/fall14/CSCI-GA.3033-010/popek-goldberg.pdf
• Properties of virtual machines (VM) • Efficiency • Resource control • Equivalance
5 Virtualization
• Virtualization Layer
Hypervisor or Virtual Machine Monitor (VMM) is a software that provides an interface between hardware and virtual operating systems.
• Types of Hypervisors
• Bare-Metal OS-1 OS-N
• Hosted OS-1 OS-2 OS-N Processes Hypervisor
Hypervisor Operating System
Hardware Hardware
Bare-Metal Hosted 6 Virtualization
• Virtualization could address following issues:
• Under-utilized resources
• Complicated system management
• Limited access to shared resources
• inefficient power consumption
• Tight coupling with underlying resources
• …
7 Virtualization
APP-1 APP-2 APP-N APP-1 APP-2 APP-N
Operating System Operating System
CPU CPU - 1 CPU - 2
Multi-tasking Multi-core or Hyper-threading
APP-1 APP-2 APP-N APP-1 APP-2 APP-N APP-1 APP-2 APP-N
Operating System Operating System Operating System
Virtual CPU - 1 Virtual CPU - 2 Virtual CPU -N
Virtualization 8 http://www.hardwaresecrets.com/everything-you-need-to-know-about-the-intel-virtualization-technology/ Virtualization Types • Platform virtualization
• Memory virtualization
• Application virtualization
• Storage virtualization
9 Virtualization Platform virtualization • Full virtualization
• Para virtualization
• Hardware assisted virtualization
• Operating-System (OS) level virtualization
• Hybrid virtualization
http://www.vmware.com/files/pdf/VMware_paravirtualization.pdf 10 Virtualization Full • Guest operating system (VM) is unaware of host OS
• Non-critical instructions run directly on hardware
• Runtime translation of critical non-virtualizable instructions in hypervisor
• Not best in performance
11 Virtualization Para • Thin layer interfaces between each guest OS and underlying hardware
• Need Guest kernel modification
• No need of runtime translation for critical instructions
• Superior in performance
• Requires expertise to patch the kernels
12 Virtualization Hardware assisted • Hardware provides support to run instructions independently
• No need to patch the kernels
• Runtime translations not required
• Better performance in comparison to other variants
• Greater stability
13 Virtualization OS level • Same OS for both Host and guest machines
• Userspace is completely isolated
• High performance
• Extremely light-weight
• Recently acquire lots of attention in Cloud world
14 Virtualization Hybrid • Combination of • para • hardware assisted virtualization
• Address the issues related to security and system stability
• Use patched guest OS along with hardware support
15 Virtualization Memory • Brings concepts of consolidation and cost effectiveness
• Managed with virtualizing physical memory by addition of an extra level of address translation VM - 1 VM - 2 CPU -0 CPU -1 CPU -1 CPU -1
VM main memory VM main memory
physical memory Question: Is “virtual memory” concept in a single operating system same as the concept of virtualization in terms of VMs? Answer: NO 16 Virtualization Desktop and Applications • Desktop and Applications run on servers
• Stateless thin clients connected to servers
• Efficient system management
• Requires high-end servers for system stability
17 Virtualization Network and Storage • Similar idea of providing an abstraction layer to the physical infrastructures
• In networks abstraction will be at the level of • Routers • Gateway • Switches • Firewalls • load balances • … • Storage abstraction allows single backends to be used for different requirements
• Ephemeral • Persistant • Specialize storage backends 18 Virtualization Network and Storage
• Aim is to provide provides:
• Efficient infrastructure utilization • Agility • Isolation • Security • ….
19 Virtualization Hardware Acceleration • Aims:
• enhance the performance • reduce the complexity in the hypervisors/VMM
• Intel and AMD chips support VT
• Intel models http://ark.intel.com/Products/VirtualizationTechnology
• AMD model http://support.amd.com/en-us/kb-articles/Pages/GPU120AMDRVICPUsHyperVWin8.aspx 20 Virtualization Concept of overcommits • Process of allocating more than the available physical resources
• Common types:
• CPU Overcommits • Memory Overcommits • Storage Overcommits
• Strong requirement from infrastructure providers
21 Virtualization Concept of overcommits • Pros:
• Favorable economic model • Efficient resources utilization • Support green computing
• Cons:
• Performance loss or unstable system response • Complex system understanding • VM shutdown by the hypervisor (extreme cases)
22 Virtualization CPU overcommit
• Allows more virtual CPUs than physically available
• Example: In case of 8 physical cores
8 * overcommit-number = total-number-of-virtual-CPUs
• Open stack KVM allows:
overcommit-number = 16.0 (max) , 1.0 (no overcommits)
http://docs.openstack.org/kilo/config-reference/content/section_compute-scheduler.html 23 Virtualization Memory overcommit
• Allocate more than physical memory
• Open stack KVM allows:
overcommit-number = 1.5GB
• 1.5 GB is required by the instance • can run on any physical host having1GB free memory
http://docs.openstack.org/kilo/config-reference/content/section_compute-scheduler.html 24 Virtualization Hypervisors • Contribution from industry and academia
• Xen • Project from Cambridge Computer Laboratory
• VMware • Commercial product
• KVM (Kernel-based Virtual Machine) • A product of Open Virtualization Alliance (OVA)
• Qemu • Opensource machine emulator and virtualizer
• ….. http://www.xenproject.org/developers/teams/hypervisor.html http://www.vmware.com/ https://openvirtualizationalliance.org/what-kvm 25 http://wiki.qemu.org/Main_Page Virtualization KVM • Hypervisor for x86 solutions with complete hardware support
• Run multiple guest OSes with private virtualized hardware: network card, memory, disk etc
• Consist of Loadable kernel modules:
• kvm.ko for core virtualization • processor specific kvm-intel.ko or kvm- amd.ko
26 Virtualization Tools KVM • Opensource software
• Support Linux and Windows (limited)
• Peripheral support to the guest OS
• Wide variety of management tools
• Backend for compute resources in many Cloud suites
• Allow resources overcommits
27 Virtualization Tools KVM
• KVM architecture for x86 system
• support nested virtual machines
http://www-03.ibm.com/systems/kvm/whykvm.html 28 Virtualization Tools Qemu • Qemu runs in following modes:
• Emulator; hypervisor runs in the user space • Hypervisor; Hardware supported virtualization (KQemu)
• Support for multiple OSes
• Based on Xen or KVM, Qemu supports nested virtualization
http://wiki.qemu.org/Main_Page 29 Virtualization Tools Qemu • Opensource software
• Extensive peripheral support
• Diversity of management tools
• Backend for compute resources in many Cloud suites
• Recommended for testing and development environments
30 Virtualization Tools Qemu
Qemu Emulator Qemu with KVM support
VM-1 VM-2 VM-1 VM-2
QEMU QEMU QEMU QEMU
Host Operating System Host Operating System KVM module
Hardware Hardware
31 Virtualization Tools VMware • Commercial product by VMware.Inc
• One of the leading server virtualization system
• frontend
• Variety of system management components
• Backend for compute resources in number of Cloud software
32 Virtualization Tools VMware • VMware vSphere consists of:
• VMware ESXi • VMware vCenter Server • VMware VMFS • VMware Virtual SMP • VMware Clients (Desktop and Web access)
• ESXi is the virtualization platform
• vCenter Server is a service that act as a administrator to ESXi platform
http://pubs.vmware.com/vsphere-51/topic/com.vmware.vsphere.vcenterhost.doc/ 33 GUID-302A4F73-CA2D-49DC-8727-81052727A763.html Virtualization Tools VMware • Overview Architecture • Comprehensive solution for data center virtualization
•
http://www.vmware.com/pdf/vi_architecture_wp.pdf 34 Role of LIBVIRT API
• Virtualization API
• Provide a common and stable layer to securely communicate with guest OS
• Supports all major hypervisors
• Open source project for secure VM management tasks
• Used to build applications based on virtual environment
http://libvirt.org/ 35 Virtualization Interesting Articles
• A quantitative comparison between xen and kvm (2010 J. Phys.: Conf. Ser. 219 042005)
• Performance Measuring and Comparing of Virtual Machine Monitors (2008 IEEE/IFIP International Conference on Embedded and Ubiquitous Computing)
• Recommendations for Virtualization Technologies in High Performance Computing (2nd IEEE International Conference on Cloud Computing Technology and Science)
• A Comparison of Software and Hardware Techniques for x86 Virtualization (Advance level) (ASPLOS XII Proceedings of the 12th international conference on Architectural support for programming languages and operating systems)
36 Virtualization and Clouds OpenStack • Open source platform for build public and private Clouds
http://www.openstack.org/ 37 OpenStack Compute • OpenStack Compute Project (NOVA)
Compute host
OpenStack Nova API operating system
KVM Qemu LXC
Hardware
38 Cloud Virtualization Storage • OpenStack volume (Cinder) • OpenStack object store (Swift)
SNIC Cloud Storage backend
Nova Glance Cinder Swift libvirt
lib-rgw lib-rbd
Ceph
Disk - 1 Disk - 1 Disk - 1 Disk - 1 39 Cloud Virtualization Network • OpenStack network components (Neutron)
40 Cloud Virtualization Network • OpenStack network components (Neutron)
http://docs.openstack.org/networking-guide/ 41 DEMO
42 DOES VIRTUALIZATION EFFECT THE SYSTEM PERFORMANCE?
43 Performance
• Yes performance loss may occur but it is highly dependent on
• Type of virtualization layer (Hypervisor) • Use case
• CPU bound application will perform differently than IO bound or network intensive applications
44 Performance
• In comparison with the physical node: • KVM perform 83.46% • Xen perform 97.28%
• Reason; Critical instruction test verses para-virtualization
In both cases, There is a performance different compare to physical machine.
Performance Measuring and Comparing of Virtual Machine Monitors (2008 IEEE/IFIP International 45 Conference on Embedded and Ubiquitous Computing) Performance
• Applica on Level • 4% performance loss evaluated with the • HEPSPEC-2006 (Thanks to Ulf Tigerstedt, CSC for help with HEPSPEC tests) • System Level • VM boot response both at local vs GlusterFS based setup
46 Article: A scalable infrastructure for CMS data analysis based on OpenStack Cloud and Gluster file system Images and formats
• Cloud images are customized disks images of OSes for private or public clouds
• Different formats are available: • raw: An unstructured disk image format (big in size) • vhd: VMware, Xen, Microsoft, VirtualBox, and others • vdi: Supported by VirtualBox, QEMU emulator. • iso: Archive format for the data contents of an optical disc • qcow2: Supported by the QEMU emulator that can expand dynamically and supports Copy on Write. • …
http://docs.openstack.org/image-guide/content/image-formats.html 47 Contextualization
In cloud computing contextualization means providing customized computing environment
Or
Allows a virtual machine instance to learn about its cloud environment and user requirement (the ‘context’) and configure itself to run correctly
http://www.journalofcloudcomputing.com/content/4/1/17 48 Contextualization
• Provide scalable solution
• No need to manage fat images
• Dynamic configuration
• Typically work in two layers • Meta-data : System information handled at cloud level • User-defined-data: User specific requirements/ settings
49 Beyond Virtual Machines
50 Containers
• OS level virtualization environment
• Kernelspace is shared • Userspace is separate for each linux system (container)
• A lightweight alternative to Virtual Machines (VM)
• Shared same resources as host OS
• A simple model for packaging applications in Linux.
https://linuxcontainers.org/ 51 Basic Illustration of Containers
Linux based host OS Non Linux host OS
Ubuntu Windows or MAC
Container Controller Container Controller
Linux VM Container Engine Container Engine
Con -1 Con -1 Con -N Con -N Con -2 Con -2 52 VM and Containers
Its is important to understand that VMs and Containers should not be viewed as competitors Virtual Machine Containers • Complete isolation • Application level abstraction • Big in size • Lightweight • High overhead • Works well with Linux, limited support for Windows • Flexible support of multiple OSes • Weak security • Significant management Overhead • Greater stability both for hypervisors and VMs • Not well suited for large applications • Important for micro-services design • Better security
53 LXC
• LinuX Container (LXC) is an open source software
• Virtual environment based on separate memory, CPU, network, io etc
• Similar to the concept of chroot
• Used in most of the container based orchestration tools
• LXD is a newer version of LXC, advanced and stronger support for cloud plugins http://blog.scottlowe.org/2013/11/25/a-brief-introduction-to-linux-containers-with-lxc/ 54 Container Softwares
• OpenVZ
• Virtuozzo (Linux and Windows)
• Solaris Containers (Solaris)
• Spoon (Windows)
• VMware ThinApp (Windows)
55 ORCHESTRATION TOOLS
56 Orchestration
• Orchestration is a process of resource contextualization based on the automation available in the cloud systems.
• A process required for • rapid application deployment • scalability • management • high availability • Agility
• Essential for large complex applications
• A process at the level of Platform as a Service (PaaS) Orchestration
• Available tools
• CoreOS https://coreos.com/ • OpenShift https://www.openshift.com • Docker https://www.docker.com/ • Kubernetes http://kubernetes.io/ • Molns (used in Lab-2) • Heat https://wiki.openstack.org/wiki/Heat
58 HEAT
• OpenStack based orchestration software to create a human- and machine-accessible services
• Consist of template base orchestration engine
• Enables simple or composite cloud applications based on existing services
• Compatible with AWS Cloud Foundation
https://wiki.openstack.org/wiki/Heat 59 HEAT heat_template_version: 2014-10-16 Template example description: A simple server. parameters: ImageID: type: string description: Image use to boot a server NetID: # heat stack-create -f test-stack.yml -P type: string "ImageID=cirros-0.3.4-x86_64;NetID=$NET_ID" testStack description: Network ID for the server resources: server: type: OS::Nova::Server properties: image: { get_param: ImageID } flavor: m1.tiny networks: - network: { get_param: NetID } outputs: private_ip: description: IP address of the server in the private network value: { get_attr: [ server, first_address ] } 60 http://docs.openstack.org/kilo/install-guide/install/yum/content/heat-verify.html Docker
• Docker package an application together with all its dependencies in the container • Guarantees that it will always run the same regardless of the environment • Container based orchestration tool • Docker Hub, container registory • Open source
https://www.docker.com/whatisdocker 61 Docker
• Docker Hub, container registry
• Image repository
• Automated Builds
• Webhooks
• Organizations
• Github and Bitbucket integrations
https://docs.docker.com/docker-hub/ 62 Kubernetes
• Google supported orchestration software
• Open source
• Supported in public, private and multi-cloud environments
• Currently supported for Linux platform
• Support Docker containers
63 Cloud Services Design
• Cloud service architecture should adhere with following features:
• Scalability • Loosely coupled components
• Security • ….
• Fault tolerant
• Vendor agnostic
• Elasticity
• Leverage different backends
64 Cloud Services Design
• Scalability
• Horizontal scaling • Vertical scaling
• Security
• Multi level security • Authorisation • Authentication
65 Cloud Services Design
• Fault-Tolerance
• Recover from failover and unexpected short interruptions
• Vendor Agnostic
• No dependency on vendor specific tools
66 Cloud Services Design
• Elasticity
• Dynamically adopt according to the requirements
• Support for different backends
• Flexible enough to accommodate different users requirements
67 Cloud Services Design
• Loosely coupled service model
• Components should not tightly connected with each other • Work as micro and macro services
• ….
68