Transcript Kaleidoscope 2009 - Powerpoint template for paper presentation

ITU-T Kaleidoscope 2010
Beyond the Internet? - Innovations for
future networks and services
Quality of Service in the
Future Internet
J. Carapinha, R. Bless, C. Werle, V. Dobrota,
A. Rus, H. Grob-Lipski, K. Miller, H. Roessler
Pune, India, 13 – 15 December 2010
The changing scenario of QoS
New challenges ahead:
Pervasive network-based applications –
ever-increasing number of applications rely
on the network
More stringent requirements –
predictability, flexibility, adaptability,
scalability
Increasingly dynamic network environment
Challenges call for fresh approaches
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
2
The 4WARD project
http://www.4ward-project.eu/
Holistic approach to shape the “Future
Internet” towards a consistent design to
satisfy requirements beyond 2015.
Innovations based
on a “clean slate”
design and new
network paradigms
Project ran from
January’08 to
June’10
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
3
Evaluating the impact on QoS
Three concepts developed and studied
by 4WARD are assessed from the point
of view of potential impact on QoS:
Network Virtualization
Generic Path Semantic Resource
Management
In-Network Management
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
4
Network Virtualization
Deployment Scenario
Business Model
Virtual Network Operator
Virtual Network Provider
Infrastructure
providers
Infrastructure
Provider A
Infrastructure
Provider B
Infrastructure
Provider C
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
Substrate
Node
Virtual
Node
End
User
Node
Substrate
Link
Virtual
Link
Virtual Last Mile
Link
5
Layered QoS
QoS provided by the Substrate
for isolation of different virtual networks
for creating virtual nodes and links with
deterministic virtual resource capacities
QoS provided inside the Virtual
Network
based on substrate QoS
homogeneous QoS solution inside possible
end-to-end QoS depends on the protocols
and mechanisms inside the Virtual Network
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
6
QoS and Network Virtualization
Network Virtualization needs QoS support in
the substrate
To become an attractive solution to deploy new
innovative network architectures
Atop, new innovative QoS solutions and
services can be rolled out and tested
But at a cost
Increased management overhead
Difficult monitoring / accountability
Standardization of interfaces, also with regard to
QoS is required
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
7
Definition of INM
In-Network Management (INM) = embedding
management intelligence in the network, enabled
through decentralization, self-organization,
embedding of functionality and autonomy.
Self-Managing Entity (SE) = a component of a
system that is self managed by objective and can
autonomously perform a series of managementrelated tasks, e.g. self-configuration and selfhealing.
INM is characterized by:
SEs are envisaged to run in all network nodes.
A dedicated SE implementing QoS-specific
tasks is also required
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
8
Self-Managing Entity Characteristics
Any Self-Managing Entity (SE) includes
two types of communication interfaces:
The organizational interface (ORG) is used by a
manager or another entity to send high level commands
to a specific INM entity.
The collaboration interface (COLL) is dedicated to
facilitate the communication between two management
entities residing either in the same or in different nodes
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
9
INM Cross-Layer QoS
The INM Cross-Layer is a SE entity
dedicated to perform QoS related tasks:
Accesses the hardware directly, through the
collaboration interface
Uses two approaches when exchanging QoS
information:
Bottom-up approach: will enable collecting traffic
parameters like: ATR, OWD, BER, and other information
that is able to characterize a specific physical link. This is
an objective way of evaluating a communication channel.
Top-down approach: will impose a specific set of
commited QoS parameters to the hardware using the
collaboration interface.
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
10
INM Cross-Layer QoS interactions
GP
Local database
ORG: publishQoSObject
ORG: publishQoSObject
ORG: publishCompositeMetricObject
ORG: committedQoSParams
INM CLQ
INM CompositeMetric
(ATR, OWD, BER, other)
COLL: committedQoSParams
COLL: requestQoSParams
COLL: getQoSParams
Legend
Measurement Netlet
HARDWARE
Publishing Netlet
(Physical Layer and MAC Sub-Layer)
Interaction between INM Cross-Layer QoS, INM Composite
Metric, hardware and other managed entities
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
11
INM Composite Metric Calculation
An immediate example of INM CLQ’s beneficiary would be the
real-time composite metric calculation.
The preliminary formula used for an overall perspective of the
links with the neighbors (for hop-by-hop data transport) is:
where k0 = 109 [bps], k1 = 10-5 [s] and k2 = 0 (additional tests
should be performed to tune the k2 parameter).
This CM could help the management as criteria for triggering
network-coding based GP activation, QoS-aware routing, etc.
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
12
INM Cross-Layer QoS and Network
Coding
QoS management element
(specialized software)
-> collects information about
available resources in each
strategic node (i.e. a node that
includes GP).
-> monitors the substrate resources
between the neighboring nodes
available transfer rates
one-way delays
-> assists congestion control
mechanisms to get a global
perspective and to have
statistics on link status
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
13
Generic Path
New “clean slate” Internet architecture
for highly dynamic and mobile networks
between two or more end systems
leverages multiple routes, network coding, etc.
adapts transport and QoS procedures to the underlying
network
Overcomes the inflexibility of the OSI model by
introducing recursiveness
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
14
Generic Path architecture
Much richer class of data flows,
merge/
network code
beyond TCP, UDP
decode
State within the network, as necessary
but no more than necessary
Common management interfaces,
split/
balance
to set up and tear down flows and
code
to query their status
join
cooperatively
Explicit identification, notably to facilitate
Generic path 1
control of multi-flow applications like
Generic path 2
videoconferencing
Mechanisms for assured performance and efficient operation
to exploit techniques like network coding and cooperative
transmission
to choose the "best" paths for the considered transport
to ensure resource sharing is "fair" and meets application
requirements
to manage the mobility of users, networks and information
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
15
Generic Path Functional Architecture
Basic building blocks for IPC:
Compartment
Name space+admin border
Hooks (ports)
Entity=process of some form in an OS
Contains functional blocks
Base class with methods for
Compartment A
(e.g., TCP/IP)
Node
compartment
End
point
Entity
End
point
Entity
Management
Access control / name resolution
Reporting and management
Control
Resource management
Compartment B
(e.g., LAN)
Routing and mobility support
Endpoint = process in the data plane
Controlled by Entity, performs
Error, flow control
Mux/demux, forwarding
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
16
Generic Path Functional Architecture
The generic path is the base-class of a hierarchy of objects
equipped with basic functions common to all path types
All data transfer functionality is defined in derived classes
Error and flow control is specified in derived classes
Generic
Path
pt to pt
voice
file
transfer
pt to
multipt
multipt
to
multipt
torrent
Path instances are created using a “path factory”
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
17
Advertising of Network Resource
Information
Principle: Provide GP level-specific network resource information, e.g. congestion etc.
Resource advertising starts from the bottom-most ResourceObject GP-level (real physical
resources) to vertical and horizontal adjacencies.
The advertising process leads to an aggregation and/or concatenation of different and
possibly heterogeneous ResourceObjects at GP-level (n-1), which results in a more abstract
GP-level specific representation of the ResourceObject at GP-level(n)
Aggregation, concatenation and abstraction are supported by the network resource
ontology, e.g. by metrics or conversion formula.
At run time the advertising
is scalable as the process is
limited to the set of end
points (ep), which already
carry flows.
44
55
11
ep_A
ep
33
ep_G
22
55
44
11
The initialization process of
the advertising functionality
with resource information is
performed in the network
management plane during
network or cloud setup.
11
3
3
22
11
44
22
33
ep_B
33
33
11
ep_D
11
11 11
ep_J
node_CTA
22
22
33
ep_H
ep_E
44
22
22
22
33
11
11
55
22
ep_C
11
node_CTB
11
22
ResourceObject
33
33
11
ep_K
ep_F
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
22
ep_I
node_CTC
18
Resource Management based on Ontologies
Network Resource Ontology providing the generic
representation of the network resource information based on
the main classes
Network Resource Parameter
Metric
hasType
Type
ConversionFormula
Impact
Relationship
hasRelationship
Type
hasImpact
Network
Resource
Parameter
Impact
isAggregated
Aggregated
hasMetric
Aggregated
Relationship
convertsTo
Metric
hasConversionFormula
Conversion
Formula
hasUnit
MetricType
Value
Unit
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
19
Resource Management based on Ontologies
Benefits
Extends existing QoS and service ontologies
Bridges the network resource and service semantics
Increases the interoperability in heterogeneous networks
 provides useful conceptualizations of different
technology specific network resources
Aids to aggregate different network resource types
 enables E2E network resource provisioning
Enables to solve mismatches
Non-functional description of network resource characteristics
Defines a machine understandable network resource
vocabulary
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
20
Conclusions
QoS will constitute a crucial requirement for the
networks of the future; however, QoS challenges
and respective solutions will not remain unchanged;
4WARD proposed and explored novel networking
approaches for the Internet of the future, including:
Network Virtualization
In-Network Management
Generic Path Semantic Resource Management
Each of these approaches brings fresh ideas and
potential solutions to handle QoS, particularly taking
into account requirements of dynamicity, flexibility,
adaptability and scalability.
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
21
Thank you!
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
22
List of authors
Jorge Carapinha PT Inovação, Portugal
Roland Bless
Christoph Werle
Universität Karlsruhe (TH), Germany
Virgil Dobrota Technical University of Cluj-Napoca,
Andrei Bogdan Rus Romania
Heidrun Grob-Lipski
Horst Roessler
Alcatel-Lucent, Germany
Konstantin Miller Berlin Institute of Technology, Germany
Pune, India, 13 – 15 Dec 2010:
ITU-T Kaleidoscope 2010 – Beyond the Internet? Innovations for future networks and services
23