Transcript ppt - DPNM Lab

Towards Management of
Future Internet
Mi-jung Choi, Sungsu Kim
DP&NM Lab.
Dept. of Computer Science & Engineering
POSTECH, Korea
Email : [email protected], [email protected]
-1-
DPNM
Outline
 Why Future Internet?
 What is Future Internet?
 Status of Current Internet
– History of Internet Growth
– Merits and Demerits of Future Internet
 Summary of research effort of Future Internet
– FIND, GENI, FIRE, JGN2, etc.
 Challenges & Requirements of Future Internet
 Architecture of Future Internet
 Management Issues of Future Internet
– Management Requirements
– Management Operations
 Concluding Remarks
-2-
DPNM
Why Future Internet?
 A growing and changing demand
– For increasing user control of contents/services
– For interconnecting ‘things’-TV/PC/phone/sensor…
– For convergence: networks/devices/services
(video/audio/data/voice)
– Mobility
– Security
 Current technologies can be, and need to be improved
significantly
– For scaling up and more flexibility
– For better security
– For higher performance and more functionality
-3-
DPNM
What is Future Internet? (1)
 Need to resolve the challenges facing today’s Internet by
rethinking the fundamental assumptions and design
decisions underlying its current architecture
 Two principal ways in which to evolve or change a system
– Evolutionary approach (Incremental)
• A system is moved from one state to another with incremental
patches
– Revolutionary approach (Clean-slate)
• The system is redesigned from scratch to offer improved
abstractions and/or performance, while providing similar
functionality based on new core principles
 It is time to explore a clean-slate approach
– In the past 30 years, the Internet has been very successful using
an incremental approach
– Reaching a point where people are unwilling or unable to
experiment on the current architecture
-4-
DPNM
What is Future Internet? (2)
 Future Internet?
– Clean Slate design of the Internet’s architecture to satisfy the
growing demands
– Management issues of Future Internet also need to be considered
from the stage of design
 Research Goal for Future Internet
– Performing research for Future Internet and designing new
network architectures
– Building an experimental facility
-5-
DPNM
History of Internet Growth (1)
 Stage One: Research and Academic Focus (1980-1991)
– Debate about which protocols will be used (TCP/IP)
– The National Science Foundation (NSF) took a leading role in
research networking
• NSFNet1: “supercomputer net”
• NSFNet2: a generalized Internet (thousands of Internet nodes
on U.S campus)
– The Internet Engineering Task Force (IETF) created open
standards for the use of the Internet
• Request for Comments (RFC) standards documents
-6-
DPNM
History of Internet Growth (2)
 Stage Two: Early Public Internet (1992-1997)
– Federal Networking Council (FNC) made a decision to allow ISP
to interconnect with federally supported Internets
– The National Center for Supercomputing Applications (NCSA)
adopted Tim Berners-Lee’s work on the World Wide Web
– Mosaic, Netscape started us down the path to the browser
environment today
• It was watershed development that shifted the Internet from a
command-line, e-mail, and file-transfer kind of user interface to
the browser world of full-screen applications
– In the fall of 1996, a group of more than thirty University
Corporation for Advanced Internet Development (UCAID)
• Subsequently become known as Internet2
-7-
DPNM
History of Internet Growth (3)
 Stage Three: International Public Internet (1998-2005)
– The Internet achieved both domestic and international critical
mass of growth
– Fueled by giant bubble in Internet stocks that peaked in 2000 and
then collapsed
– Fiber-optic bandwidth Improvements to gigabit-per-second levels,
and price-performance improvements in personal computers
– The “bubble” years laid the foundation for broadband Internet
applications and integration of voice, data, and video services on
one network base
– In 1996, a group of more than thirty universities formed the
University Corporation for Advanced Internet Development
(UCAID)-became known as Internet2
-8-
DPNM
History of Internet Growth (4)
 Stage Four: Challenges for the Future Internet (2006-?)
– The Internet has become a maturing, worldwide, universal
network
– Currently debated policy issues: net neutrality
• Two of the few surviving U.S. telcos intended to levy special
surcharges on broadband Internet traffic based on the
application and on the company
• Millions of Internet users
– Growth in functionality and value of the net could never
happened if there had been discrimination in managing
packet flow
– If the telco’s well funded campaign succeeds
• Then Progress toward universal and affordable broadband
access will be further delayed
-9-
DPNM
Merits & Demerits of Current Internet
 Merits
– The original Internet design goal of robustness
• Network architecture must not mandate recovery from multiple
failures, but provide the service for those users who require it
– Openness: low barrier to entry, freedom of expression, and
ubiquitous access
 Demerits
– “Nothing wrong – just not enough right”
– Pervasive and diversified nature of network applications require
many functionalities
• Current network architecture doesn’t support
– E.g., TCP variants for high bandwidth delay product networks [1],
earlier work on TCP over wireless networks [2], and current effort
towards cross-layer optimization [3]
- 10 -
DPNM
Research Institute for Future Internet
 US NSF
– Future Internet Design (FIND)
– Global Environment for Networking Innovations (GENI)
 European Commission
– Future Internet Research and Experimentation (FIRE)
– EIFFEL’s Future Internet Initiative
– EuroNGI & EuroFGI
 JAPAN
– NICT’s NeW Generation Network (NWGN)
– Japan Gigabit Network II (JGN2)
 KOREA
– Future Internet Forum (FIF)
- 11 -
DPNM
Research Roadmaps of Future Internet
 Roadmaps of Future Internet in EU, US and JAPAN
- 12 -
DPNM
Challenges of the Internet
 Security
– Worrisome to everyone (user, application developers, operators)
 Mobility
– Little support for mobile applications and services
 Reliability and Availability
– ISPs face the task of providing a service which meets user
expectations
 Problem analysis
– Toolset for debugging the Internet is limited
 Scalability
– E.g., routing system
 Quality of Service
– It is unclear how and where to integrate different levels of QoS
 Economics
– How network and service operators continue to make a profit
- 13 -
DPNM
Requirements of Future Internet
 Highly available information delivery
 Verifiably secure information delivery
 Support for mobility
 Interworking flexibility and extensibility
 Support for a scalable, unified network
 Explicit facilitation of cross-layer interactions
 Distribution of data and control
- 14 -
DPNM
Architecture
 Keywords
– Virtualization
• Virtualize network resources and provide customer-specific
services
– Service-oriented architecture (SOA)
• Define layer’s functions as services and converge the services
to support the network operations
• Register services, discover services in repository and acquire
necessary services
– Cross-layer design
• Divide network layers and support a cross-layer mechanism
- 15 -
DPNM
Virtualization - GENI
 Virtualize network resources and provide customer-specific services
Aggregate
Resource Controller
Slice Coordination
CM
CM
CM
Virtualization SW
Virtualization SW
Virtualization SW
Substrate HW
Substrate HW
Substrate HW
- 16 -
DPNM
SOA (1) – FIND’s SILOS
 Define layer’s functions as services and converge the services to
support the network operations
S2
Application
S1
S3
Control
Agent
M1,2
M1,1
M1,1
S4
M1,2
M4,4
M2,2
M5,3
M2,3
M5,1
S5
Policies and
Strategies
M3.2
M7.3
Physical Interfaces
Method
Precedence
Constraint
Service
- 17 -
DPNM
SOA (2)
 Register services, discover services in repository and acquire
necessary services
Service
Description
Discovery
Agencies
Service Repository
1. Publish
Service
2. Find
3. Interact
Service Requester
Service Provider
3.1 Invoke
3.2 Receive
Service
Description
Client
3.3 Reply
- 18 -
Service
Description
DPNM
Cross-Layer Design – JGN2
 Divide network layers and support a cross-layer mechanism
Application
Cross-layer Control Mechanism
Overlay Network
(IP + α) NW / Post IP NW
Underlay Network
Photonic NW
Mobile NW
- 19 -
Sensor NW
DPNM
Integrated Architecture
End Application (Content)
A
C
F
E
D
End
Application
Layer
G
B
Overlay Network
Cross-layer Control Mechanism
(Control Agent)
Content-based routing
User-based QoS
…
Application Layer
Service-Coordination Layer (SOA)
Reliable transmission
Service
Repository
TCP + Service +
Application Layer
Error detection
In-order delivery
…
Flow control
Transport Layer
Segmentation
Layer Functionalities 
Service Definition
IP + α
Forwarding
Header
error detection
QoS-guaranteed
Routing
Encapsulation
IP Layer
…
Underlay Network
Physical +
MAC Layer
Photonic NW, Mobile NW, Sensor NW, etc.  Resource Virtualization
- 20 -
DPNM
Research in Management (1)
 Research Efforts in USA
– Two FIND Projects
– Towards Complexity-Oblivious Network Management
• Current management architecture has two fundamental flaws
– The management plane depends on the data plane
– The complexity of the ever-evolving data plane disturbs
the management plane
• Propose an architecture that the management plane is irrelevant to
the data plane, and all data-plane protocols expose a generic
management interface
– Design for Manageability in the Next Generation Internet
•
•
•
•
•
Automated management
Intrinsic management support
Real-time change detection
Pervasive data sharing
Network management evaluation test-bed and methodology
- 21 -
DPNM
Research in Management (2)
 Research Efforts in Europe
– EuroNGI WP.JRA.1.5 Network Management: New trends and
Architectures
• Location management
• Mobility management
• Management architecture
– Special Joint Specific Research Project (JRA.S.06): Design and
Evaluation of Distributed, Self-Organized QoS Monitoring for
Autonomous Network Operation (AutoMon)
• Specify a distributed, self-organizing and autonomic IP QoS
monitoring framework which is based on Distributed Hash Tables
• Evaluate the performance of the peer-to-peer mechanisms for
maintaining the monitoring overlay
– European network on MANagement solutions for the Internet and
Complex Services (EMANICS)
• Joint Research: Scalable management, Economic management,
Autonomic management
- 22 -
DPNM
Management Requirements (1)
 Information Model
– Need to define high-level, goal-directed specification of network
properties & policies
– Need to specify the management objects (from HW resources to
business goals) and management functionalities
– Must be extensible
 Communication Model
– Basic management operations: get, set, create, add, delete,
action, notify
– Need to define a unified, generic management interface for all
data plane protocols  simple, interoperable, and scalable
– Must be interoperable
- 23 -
DPNM
Management Requirements (2)
 Functional Model
– Basic management functionality: FCAPS
– Management functionalities are also defined as services based on SOA
– Network nodes such as terminal, intermediate, core need to be
programmable
– Perform management functions operationally independent of data plane
– Support network discovery and selection
– Guarantee QoS
– Support generalized mobility
 Non-functional Requirements
–
–
–
–
–
Robustness (primary and backup NMs) = Reliable
Scalability
Flexibility
Interoperability
Autonomic (Automatic & Intelligent & Self-*)
- 24 -
DPNM
Management Operations (1)
 Fault Management
– Various & numerous network devices  scalable fault management
solution
– A possible solution: autonomic (self-detection, self-healing, …)
 Configuration Management
– Automatically configured
– Self bootstrapping without pre-configuration
 Accounting Management
– Authentication, Authorization and Accounting (AAA), charging, and billing
 Performance Management
– At a lower level: network performance monitoring is required
– At an upper level: service quality management is required
 Security Management
– Ensure the integrity, authenticity, confidentiality of communication with
any given peer
- 25 -
DPNM
Management Operations (2)
 Mobility Management
– Horizontal and vertical handoffs, and roaming
 Identity & Addressing Management
– Provide seamless ubiquitous support to various services in a larger service
provider environment
 Terminal Management
– Terminal location & trace management
 Service Management
– Dynamic service registration & fast discovery
 Resource Virtualization Management
– Common, well published, interoperable interfaces
– Easier integration of mgmt interfaces across virtualized resources
– Abstraction independent of underlying topology
 Cross-domain Control Management
– Define cross-domain interfaces
– Provide management capabilities based on SOA
- 26 -
DPNM
Concluding Remarks
 Future Internet
– Clean slate design of Internet architecture considering security, scalability,
mobility, robustness, identity, manageability, etc.




Summarize current research efforts for Future Internet
Summarize challenges & requirements of Future Internet
Propose an integrated architecture of Future Internet
Propose management requirements & operations of Future Internet
 Investigate possible research topics towards management of Future
Internet
– In a design phase, we can imagine all possible mechanisms to solve the
drawbacks of current Internet
– How can we validate our proposed architecture and management issues?
– What topic can we focus on?
- 27 -
DPNM
Question and Discussion
- 28 -
DPNM
Example GENI Substrate
- 29 -
DPNM
Abstractions (1)
 Three major abstractions that the GMC defines
– Components
– Slices
– Aggregates
 Components
– A collection of resources
• Physical resources (e.g., CPU, memory, disk, bandwidth)
• Logical resources (e.g., file descriptors, port numbers)
– E.g., Programmable edge node (PEN) (i.e., a conventional compute server),
Programmable core node (PCN) (a customizable router, i.e., a backbone router),
Programmable access point (PAP) (e.g., for wireless connectivity)
– Uniquely identified using GGIDs (GENI global identifiers)
• E.g., geni.us.backbone.nyc
– Each component is controlled via a component manager (CM), the entity
responsible for allocating resources at a component
 Sliver
– A distinct partition of the component’s resources
– Each component must include HW or SW mechanisms that isolate sliver from
each other
– E.g., virtual server, virtual router, virtual switch, virtual access point
- 30 -
DPNM
Abstractions (2)
 Slices
– A distributed, named collection of slivers that collectively provides the
execution context for an experiment, service, or network architecture
– Slices are uniquely identified by GGIDs (GENI global identifiers)
• E.g., geni.us.princeton.codeen
 Aggregates
– A GMC object representing a group of components, where a given
component can belong to zero, one, or more aggregates
• Example aggregate might correspond to a physical location
(components co-located at the same site), a cluster (components that
share a physical interconnect), an authority (a group of components
managed by a single authority), a network (a group of components
that collectively implement a backbone network or a wireless subnet)
– Researcher portal
• Coordinate resource allocation
• Manage set of components
- 31 -
DPNM