Upload
phungdiep
View
221
Download
0
Embed Size (px)
Citation preview
1
Network Virtualization for Cloud Computingp g
Ruay-Shiung Chang ()Department of Computer Science and Information EngineeringNational Dong Hwa University ()
1
g y ( )June 29, 2010
Virtualization is hot! Cloud computing is hotter! Cloud computing is hotter! But right now, the hottest is
22010/6/29 at NTHU
2
Outlines Introduction What is network virtualization? What is network virtualization? Current systems in network virtualization Research directions in network virtualization Conclusions
32010/6/29 at NTHU
42010/6/29 at NTHU
3
Introduction Two key concepts in the title
52010/6/29 at NTHU
Virtualization(1/2) Virtualization: Make abstractions of the
resources esou ces Hide the physical hardware from the users Combine/Divide resources M-to-N mapping (M real resources, N virtual
resources) F l titi i th l i l di i i f
62010/6/29 at NTHU
For example, a partition is the logical division of a hard disk to create multiple separate hard drives
Greater resource utilization and flexibility
4
Virtualization(2/2) Time Sharing in Large Fast Computers, IFIP
Congress 1959 (by Christopher Strachey, 1916Co g ess 959 (by C s op e S ac ey, 9 61975, a British computer scientist)
Virtual memory (Tom Kilburn, 1921-2001, a British Engineer, developed Altas (paging) in1962)
Vi t l hi t ( 1980)
72010/6/29 at NTHU
Virtual machine concept (~1980) Virtual circuits in networks X25, ATM, Frame Relay, MPLS, GMPLS.
Key technology to build a cloud computing environment
Process of Virtualization
82010/6/29 at NTHU
Traditional Computer Architecture
Virtualized Computer Architecture
5
Hypervisor Virtual machine manager (monitor) Allow multiple operating systems to share a Allow multiple operating systems to share a
single hardware host Each guest operating system appears to have
the host's processor, memory, and other resources
92010/6/29 at NTHU
Make sure that the guest operating systems (called virtual machines) cannot disrupt each other
Hypervisor Two types of hypervisor Type 1 (or native, bare metal) hypervisors run directly yp ( , ) yp y
on the host's hardware to control the hardware and to monitor guest operating systems. A guest operating system thus runs on another level above the hypervisor. This model represents the classic implementation of virtual machine architectures; the original hypervisor was CP/CMS, developed at IBM in the 1960s
102010/6/29 at NTHU
Type 2 (or hosted) hypervisors run within a conventional operating system environment. With the hypervisor layer as a distinct second software level, guest operating systems run at the third level above the hardware.
6
Types of Virtualization(1/2) Server virtualization One physical machine is divided many virtual One physical machine is divided many virtual
servers VMware ESX, Citrix XenSever, MicroSoft Hyper-
V Storage virtualization The pooling of physical storage from multiple
112010/6/29 at NTHU
The pooling of physical storage from multiple network storage devices
Storage area networks (SANs)
Types of Virtualization(2/2) Network virtualization Presents a customized network to each user by Presents a customized network to each user by
splitting up the available resources in a network Virtual Local Area Network (VLAN) Virtual Private Network (VPN)
122010/6/29 at NTHU
7
What is cloud computing (1/2) A specialized distributed computing paradigm A pool of computing power storage platforms A pool of computing power, storage, platforms,
and services to be used remotely Abstracted Virtualized Dynamically-scalable M d
132010/6/29 at NTHU
Managed
What is cloud computing(2/2) Users use web service interfaces to demand
resources esou ces Pay only for the resources that one actually
consumes (May even be free for personal use!)
142010/6/29 at NTHU
8
Cost Shift
152010/6/29 at NTHU
Services of Cloud Computing (1/4) Software as a Service (SaaS) Who is offering on demand software Who is offering on demand software Salesforce.com Google NetSuite Taleo Concur Technologies
162010/6/29 at NTHU
Concur Technologies Nexgen Software Inc.
9
Services of Cloud Computing (2/4) Platform as a Service (PaaS) Active platform Active platform Google - Apps Engine Amazon.com - EC2 Microsoft - Windows Azure Terremark Worldwide - The Enterprise Cloud Salesforce.com - Force.com
172010/6/29 at NTHU
Salesforce.com Force.com Rackspace Cloud - cloudservers, cloudsites, cloudfiles Surge
Services of Cloud Computing (3/4) Infrastructure as a Service (IaaS) Infrastructure Vendors Infrastructure Vendors Google - Managed hosting, development environment International Business Machines - Managed hosting SAVVIS - Managed hosting Terremark Worldwide - Managed hosting Amazon.com - Cloud storage
182010/6/29 at NTHU
Amazon.com Cloud storage Rackspace Hosting - Managed hosting & cloud
computing
10
Services of Cloud Computing (4/4) Cloud Computing Consulting ServiceMesh Agile IT operating model ServiceMesh Agile IT operating model Cloud computing consultants I.T. simplified Booz Allen Hamilton Thomond Technology ENKI CloudTP
192010/6/29 at NTHU
CloudTP Appirio
202010/6/29 at NTHU
11
Why Network Virtualization? Ideally, all resources (compute, storage, and
networking) would be pooled, with services g) p ,dynamically drawing from the pools to meet demand.
Virtualization techniques have succeeded in enabling processes to be moved between machines.
Constraints in the data center network continue to
212010/6/29 at NTHU
create barriers that prevent agility, for example, VLANs, ACLs, broadcast domains, Load Balancers, Firewall/IPS Security settings and service-specific network engineering.
Forces Driving Network Virtualization Computing has always driven network design Mainframes drove SNA and analog multi-point Mainframes drove SNA and analog multi point
wide area networks (WANs) during the 70s. Mini-computers drove peer-to-peer networking
protocols like DecNet, OSI and TCP/IP in the 80s. Client-Server computing drove LANs and TCP into
the mainstream in the early 90s.
222010/6/29 at NTHU
The Web drove the Internet in the 2000s And now server virtualization and cloud computing
is once again changing fundamental networking requirements to make them more flexible.
12
Status Quo (1/4) Early virtualization is all about the servers. Innovation driven virtualization is holistic: Innovation driven virtualization is holistic: Servers Storages Networks
Network infrastructure must enable: Agility/elasticity
232010/6/29 at NTHU
Agility/elasticity Portability Replication
Inflexible and costly network infrastructure is the greatest barrier
Status Quo (2/4) In virtualized and cloud environments, its not
an issue of where the network is, its where it a ssue o w e e e e wo s, s w e eisnt.
The network must be workload aware (vs. dumb plumbing)
Workloads/VMs must express their policy i t d th t k t id
242010/6/29 at NTHU
requirements and the network must provide transit and enforcement regardless of physical or logical location.
13
Status Quo (3/4) The growing automation gap between network
and application infrastructurea d app ca o as uc u e
252010/6/29 at NTHU
Status Quo (4/4) The situation today: islands of management
Fully virtualized
with integrated
management
262010/6/29 at NTHU
management
14
Is Network Ready for Cloud Computing?
272010/6/29 at NTHU
Role of NV for Cloud Computing If you take a computing device or server and
run a virtualized server on it, without a properly u a v ua ed se ve o , w ou a p ope yvirtualized network, the network just sees that it is connected to a physical computer or a server. It doesn't have the ability to see the virtual machines that are on that computer or server.
Today with various applications we need a
282010/6/29 at NTHU
Today with various applications we need a network that is intelligent and can also virtualize itself so that we can apply the right resources to the right types of applications.
15
Role of NV for Cloud Computing Challenges in managing virtual networks When you virtualize, you don't have full visibility. When you virtualize, you don t have full visibility.
If you're a company and you've bought storage, they give you a box and it's got your name on it. You go to that data center and it's yours.
When you virtualize, you're essentially being given a service contract that says you have the same
f if h d l
292010/6/29 at NTHU
amount of storage as if you had your own personal box, but now it could potentially be sitting on many different machines.
Role of NV for Cloud Computing Challenges in managing virtual networks With that it becomes much more complex to have With that it becomes much more complex to have
visibility. The tools should be developed to enable better management.
As you evolve and get into things like virtual machine mobility, it becomes even more about how you keep track of where things are.
302010/6/29 at NTHU
16
Role of NV for Cloud Computing For good performance and efficiency, it is critical that
cloud services are delivered from locations that are the best for the current (dynamically changing) set of users.
To achieve this, we expect that services will be hosted on virtual machines in interconnected data centers and that these virtual machines will migrate dynamically to locations best suited for the current user population
312010/6/29 at NTHU
locations best suited for the current user population. A basic network infrastructure need then is the ability
to migrate virtual machines across multiple networks without losing service continuity.
Role of NV for Cloud Computing
Cl d S i P idCloud Service Provider
Network Virtualization1.Connectivity Services2.Network Infrastructure Services
322010/6/29 at NTHU
2.Network Infrastructure Services
Network and IT Resource Pool
17
Connectivity Services Provide connectivity services to virtual hosts in
Cloud computingC oud co pu g Burst up and turn down bandwidth on demand Provide low latency throughput among storage
networks, the data center and the LAN Allow for non-blocked connections between servers to
enable automated movement of virtual machines (VMs)
332010/6/29 at NTHU
Function within a management plane that stretches across enterprise and service provider networks
Provide visibility despite this constantly changing environment
Network Infrastructure Services Provide network infrastructures to users Customized topology Customized topology Network components Router ---routing algorithm, routing algorithm Links --- bandwidth on demand
342010/6/29 at NTHU
18
VMware Example
352010/6/29 at NTHU
Cisco Nexus 100V
362010/6/29 at NTHU
19
HP Network Automation
372010/6/29 at NTHU
Force 10
382010/6/29 at NTHU
20
Blade Network Technologies
392010/6/29 at NTHU
Arista Networks
402010/6/29 at NTHU
21
However But the problem gets bigger and more complex
when distance and cloud provider entities w e d s a ce a d c oud p ov de e esbecome engaged.
None of the solutions above address moving a VM from one physical server to another over large distance, be it around town, across state lines across the country or the globe
412010/6/29 at NTHU
lines, across the country or the globe. Also the problem of moving from one cloud to
a different cloud!
What is needed? So how can data center networks become more
flexible? e b e? A key element of the solution is the ability to
dynamically grow and shrink resources to meet demand and to draw those resources from the most optimal location.
T d th t k t d b i t ilit
422010/6/29 at NTHU
Today, the network stands as a barrier to agility and increases the fragmentation of resources which leads to low server utilization and prevents portable or mobile workloads.
22
43
VIOLIN Virtual Internetworking on OverLay
Infrastructure - Purdue Universityas uc u e u due U ve s y VIOLIN: A VN (Virtual Network) for VMs Independent IP address space Invisible from Internet and vice versa Un-tamperable topology and traffic control V l dd d t k i ( IP lti t)
442010/6/29 at NTHU
Value-added network services (e.g., IP multicast) Binary and IP compatible runtime environment
23
Architecture of VIOLIN
Two mutuallyIsolated VIOLINs VM
NMI
NM
NMI N
MI
NMIN
MI
NMINMI-based Grid
infrastructure
NMI:NSF Middleware Initiative
452010/6/29 at NTHU
Internet
I
Physicalinfrastructure
PlanetLab
Todays Network
Applications
Networks
Ask networks for a bit pipe from point A to point B; application logic runs at the edges
y
462010/6/29 at NTHU
24
PlanetLab
Futures Network
Applications
Networks
Ask networks for a logical subnet; application logic runs on them
472010/6/29 at NTHU
PlanetLab PlanetLab: an open, global network test-bed for
pioneering novel planetary-scale servicesp o ee g ove p a e a y sca e se v ces A model for introducing innovations into the
Internet through the use of overlay networks A common software architecture Distributed virtualization Sli k f i l hi
482010/6/29 at NTHU
Slice a network of virtual machines Isolation isolate services from each other protect the Internet from PlanetLab
25
Slices
492010/6/29 at NTHU
Slices
502010/6/29 at NTHU
26
Slices
512010/6/29 at NTHU
VINI
522010/6/29 at NTHU
27
VINI
532010/6/29 at NTHU
CoreLab
542010/6/29 at NTHU
28
CoreLab
552010/6/29 at NTHU
CoreLab Deployment (on-going)
562010/6/29 at NTHU
29
Comparisons
572010/6/29 at NTHU
Global Environment for Network Innovations
582010/6/29 at NTHU
30
Global Environment for Network Innovations
GENI, a virtual laboratory for exploring future Internetse e s
Experiments in end-to-end virtualized slices
592010/6/29 at NTHU
MANTICORE II
602010/6/29 at NTHU
31
FEDERICA
612010/6/29 at NTHUhttp://www.fp7-federica.eu/
FEDERICA
622010/6/29 at NTHU
32
FEDERICA
632010/6/29 at NTHU
Cabo Cabo: Concurrent Architectures are Better than
One
642010/6/29 at NTHU
33
65
A Virtual Network in Cloud Computing
662010/6/29 at NTHU
34
Virtual Network Components Virtual Server Virtual link Virtual link Virtual switch/bridge Virtual router Resource monitor Virtual network controller
672010/6/29 at NTHU
V ua e wo co o e User interface
Server Virtualization Full virtualization KVM KVM VMware
Paravirtualization (guest host OS may need to be modified) XEN D li
682010/6/29 at NTHU
Denali Performance issues Hardware utilization Instruction parallelism for multi-core CPU
35
Link Virtualization Time-division multiplexing (TDM) Multi-Protocol Label Switching (MPLS) Multi-Protocol Label Switching (MPLS) Tunneling Generic Routing Encapsulation (GRE) Performance issues Simple
692010/6/29 at NTHU
p Fast Flexible Isolated
Switch/Bridge Virtualization OpenFlow switch Ethernet switch with flow-table Ethernet switch with flow table Run experimental protocols in real networks Decrease the work load of the router
Embedded in Hypervisor or OS
702010/6/29 at NTHU
36
Router Virtualization Logical routers (Cisco/Juniper) Run several logical routers in parallel Run several logical routers in parallel Application Specific Routing
Advantages Reconfigurability Mobility N t k C t i ti
712010/6/29 at NTHU
Network Customization
Routing Issues Addressing Non IP routing Non IP routing Virtualized object addressing
Routing policy Multiple routing paths Energy aware routing F lt t l
722010/6/29 at NTHU
Fault tolerance Multicast
Routing protocol Customized routing protocol
37
Virtual Network Controller Virtual resource management Virtual resource allocation Virtual resource allocation Virtual network provision Issues Security (Authentication, Authorization, Accounting) QoS
732010/6/29 at NTHU
Non-blocked connections (Fault tolerance) Visibility Resource utility rate (Load balance)
Virtual Network Provision Issues Isolated Resource utility rate (load balance) Resource utility rate (load balance) Non-block connections (Fault tolerance) Extendibility Energy aware (Green)
742010/6/29 at NTHU
38
75
Conclusions Virtualization is a key-technology to build
cloud computingc oud co pu g Network Virtualization can support on
demand, customizable networks for cloud computing
Design Issues C l i Vi l M hi Vi l N k
762010/6/29 at NTHU
Complexity:Virtual Machines x Virtual Networks Performance, security, privacy, policies, stability,
scalability, mobility, interface, heterogeneity, resource discovery, OAM
39
Conclusions Networks are an essential part of business,
education, government, and home communications. , g ,Many residential, business, and mobile IP networking trends are being driven largely by a combination of video, social networking and advanced collaboration applications, termed visual networking.
772010/6/29 at NTHU
The Cisco Visual Networking Index (VNI) is the company's ongoing effort to forecast and analyze the growth and use of IP networks worldwide.
Conclusions
782010/6/29 at NTHU
40
Conclusions By 2014, annual global IP traffic will reach
almost three-fourths of a zettabyte (767 a os ee ou s o a e aby e (767exabytes). A zettabyte is a trillion gigabytes.
By 2014, the various forms of video (TV, VoD, Internet Video, and P2P) will exceed 91 percent of global consumer traffic.
792010/6/29 at NTHU
Conclusions By 2014, global online video will approach 57
percent of consumer Internet traffic (up from 40 pe ce o co su e e e a c (up o 0percent in 2010).
Globally, mobile data traffic will double every year through 2014, increasing 39 times between 2009 and 2014.
802010/6/29 at NTHU
41
Conclusions What can we say about the Internet?
812010/6/29 at NTHU