Bioinspired Framework for Autonomic Communication System

1
Bio-inspired Framework for Autonomic
Communication System

• S. Balasubramaniam, D. Botvich, W. Donnelly, M.
  Foghlu and J. Strassner
• Telecommunication Software and Systems
  Group,Waterfold Institute of Technology, Ireland
• Motorola Chicago Labs

2
Contents

• Introduction and Research Issue
• Bio-inspired network management system
• Self-Management Mechanism for Resource Management
• Self-Organization Mechanism for Route Selection
• Evaluation
• Conclusion

3
Introduction

• The telecommunications network environment has
  experienced major changes over the last ten years
• Internet Connectivity
• Emergence of new broadband wireless
  communications networks
• Future network infrastructure is expected to
  support greater volumes of traffic ranging from
  services such as VoIP and complex multimedia
  services (e.g., on-line collaborative work
  environment)

4
Research Issue

• As networks grow in complexity, new management
  solutions should have autonomy
• To minimize human intervention
• Network devices and software components are able
  to exhibit self-governance (self-management,
  self-organization and self-learning)

5

• Self-management
• Maintaining System Equilibrium (Balance)
• E.g., ability to maintain thermoregulation and
  blood glucose level
• Self-organization
• Formation
• growth process from an embryo into a mature
  structured organism
• Management
• Restructuring mechanisms
• E.g., some living organisms may loose specific
  body parts that they can regenerate
• Self-learning
• Changing the way to reach the equilibrium
• (The authors does not focus on self-learning in
  this paper)

6
Research Objective

• A number of self-governance principles can be
  found in nature, and one very good example is
  biological systems
• Various organisms have different biological
  processes to adapt to environmental changes
• The research objective is to
• investigate the various biological processes that
  exhibit self-governance, and
• combine different processes into a biological
  framework that can be applied to various
  communication systems
• (Towards a generic framework, but so far, they
  implement a set of bio-inspired algorithms for
  packet routing under various network load)

7
The proposed framework

• The proposed framework for autonomic network
  management consists of two layers
• System
• Monitoring a whole network and managing its
  resource in a centralized manner
• Self-management algorithm is applied to system
  layer
• (Self-management mechanism is always applied to a
  system level?)
• Device
• Monitoring and managing each device in a
  decentralized manner
• Self-organization is applied to device layer
• (Self-organization mechanism is always applied to
  a device level?)

8
Blood Glucose Homeostasis

• Self-management mechanism is designed after blood
  glucose homeostasis
• Homeostasis system a mechanism that living
  organisms use to maintain system equilibrium
• Monitoring the body performing different
  activities and responding through one or more
  feedback loops
• Glucose homeostasis
• Living organisms uses glucose in blood as their
  energy
• Balance the amount of glucose in blood
  autonomously
• Glucose can be stored as fat
• Motivation to use blood glucose homeostasis
• Balancing blood glucose lt-gt balancing network
  load

9

• Details of blood glucose homeostasis
• Glucose in blood is used as its energy
• When body intensity increases
• The glucose in blood get used, and the process of
  glycogen breakdown for glucose production is
  performed
• Glycogen is good for store glucose temporary (in
  muscle and liver), and its easy to transform
  to/from glucose
• Once the glycogen is used beyond a specific
  threshold, the breakdown of fat is performed to
  obtain glucose
• You have to do exercise at least 20 min to reduce
  your fat!!
• When body intensity decreases
• When the amount of glycogen is increased, the
  glycogen is transformed into glucose, and fat

Glycogen
Glucose
Fat
10

• Transformations happen by reacting bodys
  activities
• In average routine activity, a state transits
  between Glucose and Glycogen

11
The proposed algorithm

• For each source and destination pairs, a primary
  path (e.g., shortest path) is assigned, and its
  demand profile (average traffic load) is updated
  periodically
• When a traffic load exceeds the demand profile
  during a certain time, a resource manager will
  start using the spare capacity (different paths
  between the source/destination pair) in the
  network

12

• For new traffic t
• If the traffic t can be accommodated in a demand
  profile, use a primary path
• Otherwise, use another route between the source
  and the destination
• If the another route is used long time and many
  times, the route is added to the demand profile
• If a used bandwidth is lower than a demand
  profile long time, a demand profile is reduced

13
Route Management

• Bio-inspired self-organization mechanism for
  de-centralized route management
• The proposed mechanism combines a number of
  biological mechanisms
• Hormone Signaling
• A signal with an ability to change its state
  based on the environment condition
• (??)
• Reaction Diffusion
• Diffusion Reaction
• E.g., patterns on animal coat
• Chemo taxis
• Directing movements according to the
  concentration gradients of some chemicals in the
  environment
• e.g., bacteria finds food by swimming towards the
  highest concentration of food molecules (e.g.,
  glucose)

14

• Hormone signaling is used to determine the
  distance of intermediate nodes from a source node
• The hormone messages H, containing a hop count,
  are transmitted from the destination to the
  source
• Hs initial H value
• A back propagation is initialized once the source
  node received the final Hf value to normalize
  each hormone value
• h normalized hormone value

j
s
d
Hs 10
Hf 5
hj 1/2
15

• Once the hormone values are normalized, each node
  start deducing routes
• Based on reaction-diffusion and chemotaxis
• Each node calculates its own load, and spreads
  the information among its neighbors (diffusion)
• Once a node receives load information, it
  calculates a gradient value for each link
  (reaction)
• Once gradient values are calculated, packets will
  route hop by hop through the highest gradient
  value (chemotaxis)

16

• Capacity of link i of node n to node j is
  represented as ln,c,i?j
• Traffic of link i of node n to node j is
  represented as ln,i?j
• A gradient value of link i of node n to node j
  is represented as Gn,i?j
• Combination of a load, current traffic and a
  hormone value
• Large F ? large unused capacity
• a 0.2 in a simulation
• Large l ? high traffic
• ß 0.4 in a simulation
• Large H ? close to destination
• ? 0.4 in a simulation

Load on node n
Gradient Value
j
i
n
hj ?
17
Evaluation

• To evaluate the proposed algorithm, a simulation
  is performed with a certain topology and traffics
• Traffics
• S1 ? D1
• S2 ? D2
• S3 ? D3
• Show how paths are selected
• They just focus on the S1-D1 pair

18

• Before the traffic builds up within the network,
  the gradient calculation on each node shows the
  best path for S1-D1 is 1, 5, 6, 8, 12

19

• Simulated traffic patterns between each pairs
• Two types of traffics
• Data and Multimedia

20

• At time 0
• Gradient value for link 5-6 is greater than one
  for link 5-4
• Link 5-4 are used by other traffics
• At time 4
• As the traffic load increases in the network, at
  time 4, the link gradient for 5-6 reduces and is
  taken over by link 5-4

21

• As a result, new traffic between S1-D1 are sent
  through different path

22

• The amount of bandwidth consumed by each source
  and destination pair
• At time 14, the resource manager assigns spare
  links to S1-D1 pair

23
Conclusion

• The authors propose a bio-inspired network
  management system
• Self-Management Mechanism for Resource Management
• Self-Organization Mechanism for Route Selection