Transcript resume - Department of Computer Engineering

Zeinab Movahedi
Phare Team
Laboratoire d’Informatique de Paris6 (LIP6)
[email protected]
TCP/IP layers
 Application layer
 Mail transferring, P2P, applications, FTP, DNS, ARP
 Transport layer
 UDP, TCP
 Routing layer
 Link state routing, distance vector routing
 OSPF, BGP, DSR, …
 MAC layer
 Physical layer
More advanced concepts …
 Sensor networks
 Autonomic communication
 Green networking
 Cloud computing
 Virtualization
 Etc.
Outline
 Introduction
 Motivations et encouragements
 Definitions
 Architecture & conceptual model
 Challenges & related fields
 Conclusion
 AutoI Project
5
Introduction
 Explosion of computing systems
 Heterogeneity
 Complexity and cost of management
6
Motivations




7
between ⅓ to ½ of a company’s total IT budget spent for crashes
For each 1$ spent for storage, 9$ for its management
40% of failures caused by human errors
Huge impact of downtime on the economy
Solution
 Providing systems and networks with autonomic
behaviors, which means immigrating towards selfmanagement systems
8
Autonomous Nervous System
(ANS)
Autonomic Systems inspired
from Autonomous Nervous
System
9
IBM definition
 Proposed by IBM in 2001
 An autonomic system is a self-management system
 Fundamental properties:
 self-configuration
 self-optimization
 self-healing
 self-protection
10
purpose-driven definition
 An autonomic system is one that operates and serves its
purpose by managing its own self without external
intervention even in case of environmental changes
 Properties :
 Self-awareness & context-awareness
 Automaticity
 Adaptability
 Portable & openness
11
Architecture (1) : conceptual model
12
Architecture (2)
13
Challenges
 Relationship between autonomic elements
 Optimization & learning theory
 Robustness
 Trust
14
Relative fields
 Artificial Intelligence
 Multi-agent Systems
 Software Engineering
 Reliable Systems
 Etc.
15
Autonomic architectures
16
Architectures hiérarchiques: DRAMA
 YAP report, DRCP/DCDP for policies dissiminations
Hierarchical Architectures : DRAMA
Hierarchical Architectures: Cluster-based
Role: MN, CH, CN
Module: CM, TN
Hierarchical Architectures: Cluster-based
 CF = w₁.MEM(t)+w₂.PP(t)+w₃.BP(t)
/(w₄.MR(t) + w₅.CL(t))
 Replication and distribution of policies
 Nodes designated by Hyper Cluster


Based on network volacity
Activating the option in the module
Hierarchical architectures: AutoI
Distributed Architectures
Architecture Ginkgo
Cross-layering based architectures
 Take into account information from different layers and
not necessarily adjacent to obtain a system more
adaptable to its environment.
 Benefits
 Optimizing performance, creation of new applications,
avoid duplication of efforts, etc.
 A relevant approach for collecting information for
autonomic communications
 Security
 Two categories:
 Locale view
 Global view
Cross-layering based architectures (vue locale)
 Profile-based architecture
 Service-based architecture
Cross-layering based architectures (local view)
 MobileMan
Cross-layering based architectures (local & global view)
 Need a global view for optimization
 Load sharing, routing, energy consumption, etc..
 CorssTalk: uses both the global view and local view in
order to take local cross-layering decisions
Cross-layering based architectures (local & global view):
CrossTalk






The local view consists of cross-layering information
The local view is added to the end of data packets
Each node receiving a packet extract the information and adds
it to its global view
Only the source of packet adds some information to the
packet. Reasonable packet size
Setting the parameter of distance and time of the information
stored in the global view
Samples of the global view are aggregated to represent
relevant information (via some algorithms)
Architecture à base de cross-layering (vue globale & locale)

XLEngine
 La vue locale est communiquée en inondation
optimisée

POEM
 La vue locale est communiquée périodiquement aux voisins directs

MANKOP
 Plan de connaissance constitué de:



Networking-level Knowledge Plane
Application-level Knowledge Plane
Les informations de plan de connaissance est communiquée
périodiquement aux voisins directs (considération des besoins)
Autonomic architectures
Catégorie
Adaptation
Monitoring
Adaptation
Apprentissage
Non
Sécurité
Nosécurisé
No
Hierarchical
statique IMO
Monitoring
Politique
YAP statique
No
No
CAMANET
Hierarchical
Politique
XML-RPC
No
No
ADMA
Hierarchical
Politique
-
No
No
ANA
Hierarchical
Politique
MBF
No
No
INM
Hierarchical
Politique
Goosip et tree-based
No
No
Unity
Hierarchical
Fonction
d’utilité
Sentinel
No
No
CogNet
Distributed
Distribution
normal
aléatoire
-
Yes
No
XLEngine
Distributed
-
Flooding sélective
No
No
AutoI
Hierarchical
Drama
Politique
30
Conclusion
 An autonomic system is one that operates and serves its
purpose by managing its own self without external
intervention even in case of environmental changes
 Autonomic System is a novel and open research paradigm,
in relationship with several other fields
31
AutoI Project
32
AutoI Project
 STREP Project
 11 partners from 7 countries
 France, Germany, Greece, Ireland, Spain, USA, United
Kingdom
 3 industrial partners
 Motorola (USA), Ginkgo Networks, UCopia
Communication
 30 months project started at January 2008
33
AutoI Project: goals
 To improve the management of NGN
 Two principal axes :
 Autonomic Management
 Virtualisation for flexibility
•
A technology which allow coexisting of several virtual networks embedded
in a same physical network
 To design and develop a self-managing virtual resource overlay
that can span across heterogeneous networks, support service
mobility, quality of service and reliability.
34
Network virtualization
37
Virtual networks
38
Thank you for your attention
39