ETRI FI Projects & HiMang
2011. 8. 11.
Taesang Choi
ETRI
ETRI FI Projects Overview
5
Unified Architecture &
Coordination
Experiment
(MOFI)
6
Experiment
(HiMang)
Experiment
(CSON)
2
1
3
Physical
Substrate
OpenFlow
@FIRST
Virtualized
Network Device
4
Federation
CH
FH AMAM
Virtual
Network
GENI/AKARI/Europe
WLAN
Cellular
WLAN
1
Virtualized Programmable
Platform
3
OpenFlow
Switchnig
2
Control Framework for
Virtualized Testbed
4
Federation
2
5
Innovative
Architectures
6
Unified
Architectures
Research Area: Network Architecture
Mobile-Oriented Future Internet (MOFI)
To develop new naming, addressing & routing architecture for mobile oriented
environment
3
Research Area: Service Architecture
Context based Service Overlay Network (CSON)
To develop an adaptive service path configuration technology for Context
based Service Overlay network (CSON) which provides functional environment
for composite service of 3rd party service providers
4
Research Area: Management Architecture
Highly Manageable Network and Service Architecture (HiMang)
To develop a highly manageable network and service management architecture
that enables current and future network operators and providers to build and
manage economically viable and sustainable services.
Business Managment
Domain
Manager
Buisiness Policy & Rules AMF
Billing
Service Policy & Rules
Service Lifecycle ManagementAMF
Service Charging
Domain
Manager
Specific
Control Loop
Network Management
Network Policy & Rules
Network Provisioning, Fault
AMF
Performance, Accounting,
Routing, Traffic Engineering, QoS
Network Virtualization
Domain
Manager
EMS
AMF
NE AMF
EMS
AMF
NE AMF
EMS
Domain
Manager
Specific
Control Loop
AMF
NE AMF
5
Distributed
Knowledge
Specific
Control Loop
Element Management
Node Configuration
Physical & Virtual Resources
AMF
Routing Protocols,,
Traffic Managment, Forwarding
Ochestration Plane
Specific
Control Loop
Service Management
Research Area:
Virtualized Programmable Platform (FIRST)
FIRST@ATCA
ATCA standard module based programmable virtualized network platform that
provides convenient ways of building and controlling virtualized nodes with
strong resource isolation.
Platform-independence, developer-friendliness, extensibility, and JVM structure
Multi-core Octeon NP based Data Plane Virtualization and Programmability
3) FIRST Platform
Manager
VM
CPU*
Sliver
VM
CPU*
Memory
Memory
Storage
Storage
Net IF
Net* IF
2) Programmable Data Plane:
PacketvisorWorks
PacketvisorVM
1) Network Virtualization:
PacketvisorVM
6
Research Area:
Control Framework for Virtualized Test-bed
FiRSTProNET
Enables many users to share FIRST test-bed simultaneously
Control mechanism and subsystem that provides safe and dynamic
configuration of virtualized nodes and networks by managing ICT resources
Virtualized Network Control Solution
FIRST@ATCA

Virtualization based Future Internet test-bed control
solution

Structure for resource openness and sharing

FIRST@ATCA resource allocation,
virtualized nodes/networks configuration
Virtualized nodes
System Configurations and Features

Comprised of FH,CH,AM,CM,CM-Emulator

Resources : CPU, Memory, Virtual NIC, Bandwidth of
ICT devices (e.g., router/switch, server)

RSpec extensibility, intuitive use based on FH
ㄹ
FiRSTProNET
7
Virtualized
Network
Monitoring
Research Area: Standardization
FI Standardization
•
To develop the FI related standards such as FI reference model, network
virtualization, federation technologies, etc.
• ITU-T (Q.21/13, Future networks)
•
•
Y.3001, Future Network : Objectives and Design Goals
Y.FNvirt, Framework of Network Virtualization for Future
Networks
• ISO/IEC JTC 1/SC 6/WG7
•
PDTR 29181-1, Future Network, Part 1 – Overall aspects
•
Specification - Federation scenarios and requirements
•
draft-jeong-vnrg-virtual-networks-ps-00
• GENI Control Framework WG
• IETF/IRTF VNRG (Virtual Network Research Group)
8
HiMang Requirements
 Increasing Managed Entities – Services and Devices
 M2M, Internet of Things
 Increasing Management Tasks
 Green Networking, Energy Efficiency, Smart Computing & Networking
 Dynamic / On-demand Management
 Dynamic Process Management
 On-demand service and resource configuration/provisioning related to on-demand
subscribing or situation evolution
 Managing New Types of Services and Resources
 Virtual Resources (e.g., Cloud computing, Virtual networks, Virtual Storages, etc.)
 Dynamically provisioned or shared virtual infrastructure based on different types of
virtual machines (e.g., virtual routers, virtual service components, etc.)
 Interconnection, Interoperability and Extensibility
 Between network operators, service providers, and third parties
 Higher level of integration
 Orchestration of the closed management loops across federated domains
 Autonomic, Context-aware, and Knowledge Plane Focus
9
HiMang Design Goals
 To define a holistic autonomic management architecture that encompasses





both the Evolutionary Approaches and the Revolutionary/Clean-Slate
approaches
To accommodate revolutionary clean-slate thinking to overcome the
legacy management limitations
To be applicable incrementally (evolutionarily) and “instantiated” for
autonomic management of today’s protocols and architecture
Smooth Migration from the current Internet to Future Internet can be
realized by applying Autonomicity principle
To define Standardizable Specifications of the architectural functional
entities and interfaces that guarantee interoperability
To develop a highly manageable network and service management
architecture that enables current and future network operators and
providers to build and manage economically viable and sustainable services
10
Our Assumption on Future Internet
Pault
Performance
Accounting
Configuration
Security
CE
CE
Control Plane B
Control Plane A
CE
CE
CE
Virtual Network A
CE
Virtual Network B
Network C + Control Plane C
Virtualization capable Network
Virtualization capable Network
11
Legacy Network
Cognitive Network Management
Monitoring Data
(learning and reasoning)
Future Internet
12
Decision Making
Main Principle: Cognitive Network Management
Knowledge-Centric View
 Data
 Raw values
 Meaningless unless they are
undergrounded in context
 Produces Information
 Information
 Data applied to the current situation
 Uses short- and long- term memory
 Produces Knowledge
 Knowledge
 Can be measured, calculated, and/or
inferred
 Can be bound to specific objects as well
as to generic situations
13
Cognition Model for HiMang
Learn
Compare
Normalize
Finite State
Machine
Model and
Reasoner
Observe
Plan
Short-Term
Memory
Long-Term
Memory
Environment
Decide
Act
14
Proposed High-level Architecture
Inference
Plane
Management
Plane
Orchestration
Manager
Distributed
Inter-domain
Knowledge
Domain Manager
Domain Knowledge
Specific
Control
Loop
…
Autonomic
Aggregation
Manager
Physical &
Logical (Virtual)
Data Plane
Autonomic
Managed
Element
15
Virtual Resource
Knowledge
HiMang Instantiation to Cloud DC Management
Orchestration Plane
Distributed Knowledge
Network-level Domain Manager
Service-level
Failure Management
Objectives & Policies from
Network-level to EMS-level
Usage & Cost
Management
Security
Management
Traffic Monitoring
Performance
Management
Core
Router
Physical Capacity
Management
End-of-Row
Switch
…
Top of Rack
Switch
Switch/Server
Configuration
Link Management
AMF
Border
Router
EMS-level Domain Manager
Device Failure
Management
Virtual Capacity
Management
Server
Rack
AMF
…
Cloud Data Center
16
Changes to
User’s
specification
User’s Initial
Request
Periodic Quality
Assurance Check
Retrieve Cloud Resource
Information
Retrieve Cloud Resource
Information
Compute Optimal Resource
Availability (Theoretical)
Look for Faults
(SLA Violation)
N
Achievable?
N
Y
Fault
Detected?
Y
Configure Resource
Provision
Fault Recovery
Failed
Validation
Validation
Success
Success
Inform Admin
Inform Admin
Inform User
Inform User
End of
Provisioning
End
17
Proposed Implementation Architecture
CSI (Common Service Interfaces)
GUI
TMF
Measured
Traffic Data
Configuration
Package
Measurement
Package
Fault
Package
Misc.
Package
Traffic Adaptation
Interface
RMF
PF
FMF
Provisioning Interface
Fault Event Interface
GUI – Graphical User Interface
TMF – Traffic Management Function
PF – Provisioning Function
FMF – Fault Management Function
RMF – Resource Management Function
Fault
Event
TMF
Agent
FMF
Agent
PF
Agent
Domain to Aggregation
Adapting Interface
PF/RMF/TMF/FMF
Node
Type 1
Node
Type 2
18
Node
Type n
Autonomic
Aggregation
Manager
Resource
Status
19