Upload
gloria-bowker
View
215
Download
0
Tags:
Embed Size (px)
Citation preview
Network Virtualization as a Mean for Service Convergence for Future Communication Systems –What can we learn from Federated Experimental Facilities?
K. Tutschku ([email protected])
Chair of Future CommunicationProf. Dr. K. Tutschku
Institute for Multimedia and Distributed SystemsFaculty of Computer Science
Future Internet?
?
Overview
The Internet under pressure
The success of the Internet
Network virtualization: virtual structures for convergent services
The GENI experimental facility
Performance issues of Transport Virtualization
Conclusion
Accessnetworks
Core networks
Internet under Pressure
Internet will become a network of applications, services und content
Services are the new central elements Convergence in usage
What changes hereof are anticipated for users, mechanisms and the future network architectures?
GSM
Teletext
Data service
Serviceprovider
Networkoperator
Services
Applications
class. national PTT
POTS Mobile
ISDN
Voice(wired)
Voice(cellular)
Reseller AReseller A
X.25 / FR
Networks under Change: Services
Limited convergence
GPRS
Web
IP service
Serviceprovider
Networkoperator
Services
Applications
A B C D E
IP Service Provider
POTS mobile
ATM/ MPLS
Limit convergence Internet Protocol (IP) is main converging layer
Networks under Change: Services
Deficiencies of the Current Internet
Performance (“World wide wait”)
However: No convergence; QoS islands with are available (depending on technology and provider)
Reliability:
Again: no convergence Availability of the Internet ´03: 93.2% − 99.6% Availability of POTS: 99.99% – 99.999% However: sophisticated resilience mechanisms available
at experienced ISP
Competition / business models:
J. Crowcroft: “… I can go on the web and get my gas, electricity, … changed , why is it not possible to get a SPOT price for broad-band internet?” (E2E-interest mailing list on April 26th, 2008); contracts prohibit change
No convergence; even technically infeasible
UMTS
Web. Unified communication appl.
IP Service
xDSL
WLAN
PSTN
Serviceprovider
Networkprovider
Services
Applications
Multi-Network Services
Voice
Overlays (e.g. Skype)
VideoMessa-
gingData
Limit convergence
Internet Protocol (IP) is main converging layer (but: hour glass model!)
Integration of different technical and administrative domains by virtual networks: Overlays Overcome deficiencies and implement new features Networks/overlays have to be (self-)organized for the services
A B C D E
IP Service Provider
Networks under Change: Services
Services will be offered and controlled from the edge („edge-based services“) Central services will be virtualized Boundaries between consumer and provider vanish (“prosumer”)
Symmetrical rolls require new architectures (ADSL?) and permit new business models („Peer productivity“)
Management of edge-based services? Optimal placement? Different user behavior? Dimensioning?
Which functions should be self-*?
?
?
provider at edge of network
Data/
Service
distributed centralized
Network-based provider (server)D
ata/ S
ervice
Data/
Service
Data/
Service
Data/
Service
Data/
Service
consumer at edge of network
Networks under Change: Services
Application-oriented and self-organizing overlays outperform current services
Support for resources contribution by arbitrary users: „Overlays for Cooperation/ Participation“
What is the performance of self-*? Scalability? Churn? Dynamical traffic patterns?
Networks under Change: Services
Head-quarter
ATME3
Management plane
Remote office
Servicerequest (FAX, Web)
„semi-manual“ provisioning
Networks under Change: Transport Systems
Control Plane
Head-quarter
IP layer
100GE layer
DWDM layer
EPON
auto. provisioning
Management Plane
State-of-the-art optical transport systems: Ultra-high transmission capacities; embedding of different transport network into one
physical network (multi-layer networks) Decay of CAPEX per Bit Increased automation self-* features (self-operation,
self-organization)
However: higher complexity („numerous overlays“?) How to achieve convergence?
auto. Signaling
Remote office
Multi-Layer-Networks
Networks under Change: Transport Systems
Success of the Current Internet
Efficient P2P-based, self-organizingcontent distribution networks
Ratio of data traffic types at
public access node
Data traffic by IP TV
P2P, 67,3%
eMail, 1,2%
FTP, 0,3%
other, 23,3%Web, 7,9%Quelle: Telefonica (2003)
Terrabytes per month
YouTube − world wide (Cisco est., May 2008) 100.000
P2P Video streaming in China (Jan. 2008) 33.000
YouTube − USA (Mai 2008) 30.500
US. Internet back bone at year end 2000 25.000
US. Internet back bone at year end 1998 6.000
Quelle: CISCO (2008)
Multi-Source Download (eDonkey, BT)
Publish X
Publish X
Que
ry X
Transfer of segment A
Offers file X
Offers file X
Peer
Index server
Looking for X
Que
ry X
Transfer of segment B
Publish X
Offers file X
P2P: two overlays (virtual structures) with different application layer functions (two basic P2P functions: searching / content exchange); each with different topology, addressing, and routing
Search function: able of self-contained re-organization of search mechanism
Downloading peer: self-initiated selection of providing peer (parallel routing of content) based on resource quality (throughput) select the best (multi-)path for the content
→ Self-operation of basic P2P functions among networks convergence is possible
Diversity I: Multi-Provider Environment
High diversity wrt. paths: Three North-american nation-wide ISPs
Tier1 (AS 3967 Exodus, AS3356 Level3, AS6467 Abovenet; M. Liljenstam et al., 2003)
Multiple routes for increased
resilience and compe-
tition are (theoretically) readily
available!
Network selection not available in
current IP no convergence
Any way: autonomous identi-fication
of available resources needed (Thanks to Michael Menth für vsualization)
East coastWest coast
Diversity II: Multi-Quality Environment
25% of paths violate the triangle inequality (wrt. packet delay) Measurements in PlanetLab by
S. Banerjee et al. (2004)
➞ Internet routing is far from optimal ➞ Better paths exist; capazity is readily
available ➞ Can be offered (competition)
➞ Again: autonomous identification of available resources needed
! „Multi-homing“ not really available current IP protocols
A
Triangle Inequality (TI): D(A,C) ≤ D(A,B) + D(B,C)
B C
direct connection
Using an intermediate
Virtualization of Operating Systems
One hardware executes multiple systems
Safe: Strong isolation of resources, e.g. for testing and debugging
Individual and powerful: User see whole computing center as his own computer
Efficient: reduction of CAPEX (consolidation of multiple machines in a single
physical one) and OPEX (operational issue) Convergence of operating systems
Virtual Networks for Convergent Services
Build a „personal network (PN)” for an application (PN PC) Integration of different technologies and administrative domains Re-use of generic infrastructure on small time scale Push application-layer mechanisms safely down the stack
☝ Avoid “multi-layer” trap autonomic/self-* operation; particularly smart resource mgmt
Convergence by Network Virtualization
Diversity Exploit diversity of resources by smart
localizationProvide optimal resources
OS virtualization Strong isolation of resources Consolidation and efficient
operation Enables local convergence
Overlays Overlays: application-oriented topology,
addressing, and routing Multi-Network Services Self-operation of functions Enables global convergence
Transfer von Segment A
Stellt X zurVerfügung
Stellt X zurVerfügung
Peer
Index server
Sucht X
Transfer vonSegment B
Stellt X zurVerfügung
Share Virtual Machine
Guest OS
Virtual CPU
Virtual Memory
Virtual I/O
CPU Memory I/O
Virtual Machine Monitor
Guest OS
Virtual Machine
Service Service
Aggregation Load BalancerService
Logical Virtual Server
Load Balancer
Switch
Physical Server
A Formal Description for Virtualization
Virtual resources
Generation of logical resources Sharing: one physical, multiple logical resources Aggregation: one logical, multiple physical
Transport Virtualization (TV)
Example: Virtual Memory OS integrates disconnected physical memory, even disk space,
into continuous memory location of physical memory doesn’t matter
Transport Virtualization (Tutschku, Nakao, 2008): abstraction concept for data transport resources
Physical location of transport resource doesn't matter (as long resource is accessible)
Achieved by: abstract data transport resources combined from one or more physical/overlay transport
resources, e.g. leased line, wave length path, an overlay link, MPLS path, or an IP forwarding capability
physical resources can be used preclusive or concurrently basic resources can be located in even different physical
networks or administrative domains
A. Nakao
T. Zinner, P. Tran-Gia
Concurrent Multi-Path Transfer
Physical topology
Overlays of provider I
Overlays of provider II
Aim: Very high and reliable throughput between two end hosts
Aim: Very high and reliable transmission between two end hosts
Solution: Transport Virtualization:Combine multiple paths (even from different
overlays)
pooled transport pipe
POP
1
Path Source
2
3
SORA Router (One-(Overlay)-Hop)
Internet Router
4
Path oracle One-hop Source Router (SOR)
Routing Overlay (= P2P Multi-Source Download)
Implementation: routing overlays
Gummadi et al (2004): Scalable “One-Hop” (= intermediate) routing overlays
Nakao, Tutschku, Zinner: Consideration of multiple paths
(2008)
! May be inefficient Reduction of overhead (since edge-based) Placement of NV router in core
Application: Transport System Virtualization for high-capacity transmissions, e.g. for HD TV
How can we test it?
1 Divert selected endhost packets
2 Request Paths for Diverted Packets
3 Encapsulated, send using path
4 Decapsulate, egress to destination
Started in 2007
Original agenda Research:
○ Identify fundamental questions; Drive a set of experiments to
validate theories and models Experiments & requirements
○ Drives what infrastructure and facilities are needed
Currently One very rough blueprint; Five different control
architecture
Major ideas infrastructure operation: Clearing house: settles usage request Lifetime for resources: has to be returned at prede-
fined lifetime
GENI: The Global Environment for Network Innovation
Appealing Idea: Federation
Backbone #2ComputeCluster #1
Backbone #1
ComputeCluster #2
Wireless #1
Wireless #2
Access #1
CorporateGENI suites
Other-NationProjects
Other-NationProjects
My experiment runs across the evolving GENI federation.
NSF parts of GENI
My GENI Slice
(Slide by Chip Elliot)
What resources can I use?
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
Offer
GENIClearinghouse
Researcher
Aggregates publish resources, schedules, etc., via clearinghouses
Resource Discovery
(Slide by Chip Elliot)
GENIClearinghouse
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
Create my slice
Clearinghouse checks credentials & enforces policyAggregates allocate resources & create topologies
Slice Creation
(Slide by Chip Elliot)
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
Experiment – Install my software,debug, collect data, retry, etc.
GENIClearinghouse
Researcher loads software, debugs, collects measurements
Experimentation
(Slide by Chip Elliot)
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
Make my slice bigger !
GENIClearinghouse
Allows successful, long-running experiments to grow larger
Slice Growth & Revision
(Slide by Chip Elliot)
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
Make my slice even bigger !
GENIClearinghouse
Components
Aggregate DNon-NSF Resources
FederatedClearinghouse
Growth path to international, semi-private, and commercial GENIs
Federation of Clearinghouses
(Slid
e b
y C
hip
Elli
ot)
Components
Aggregate AComputer Cluster
Components
Aggregate BBackbone Net
Components
Aggregate CMetro Wireless
GENIClearinghouse
FederatedClearinghouse
Components
Aggregate DNon-NSF Resources
Always present in background for usual reasonsWill need an ‘emergency shutdown’ mechanism
Oops
Stop the experimentimmediately !
Operations & Management
(Slid
e b
y C
hip
Elli
ot)
Routing Overlay
usedpath
Routing Overlay
pooledressource
Routing Overlay I
Routing Overlay II
pooledressource
Federation for Transport Virtualization
Path selection
Path selection for concurrent use
Path selection in federated networks convegence of networks
Transmission Model
p1,1
dst
Assumption: use k parallel paths on m overlays
p1,n1
pm,1
pm,nm
src k pooled paths
m
i ink1With paths
Data stream divided at router into segments with k parts
1
k
2
k parts have arrived
k parts are send in parallel at time t
k-1
each provider will offer a set ni of parallel paths(i = 1…m)
1
k
overlay 1
overlay m
Buffer occupancy?
Reassemble data stream from obtained parts
Re-sequencing buffer of size L
Scheduling?
So far: Simulation Experiment
Input:
Number of paths
Scheduling
Output: Re-sequencing buffer occupancy distribution
Search for path selection strategies; future on-line selection for convergence
Path delay distributions
Path capacity
Source Destination
Impact of Type of Delay Distribution I
Types of distributions:
Uniform: artificial behavior
Truncated Gaussian: mathematical tractability
Bimodal: two modes of a path
Investigation of different influence factors
Delay
Impact of Type of Delay Distribution II
Two synchronous, equal capacity paths Three synchronous, equal capacity paths
Buffer
Highly non-linear careful and complex path selection
Buffer
Current Work: Perform Real-World Measurements
Measurement set-up
Gain realistic parameters and strategies
Conclusion
Expected features of the Future Internet Faster, more reliable, more business cases, increased interaction
with users: symmetric rolls, „Architecture for Participation“ Forming of applications-specific overlays
Network virtualization: Consolidation of multiple (virtual) network into one physical
infrastructure
Making data transport independent from resource locations transport virtualization
Integration/convergence of different transport systems und operator domains by overlays and network virtualization
Design networks for applications (rather than designing applications for networks)
Experimental facilities: Federation: blue print for future network operation and
convergence Resources with limited lifetime significant challenges in
resource management
Thanks for your
attention!
Questions?