Title: Increasing robustness in the BioNetworking architecture: A Distributed Approach
1Increasing robustness in the Bio-Networking
architecture A Distributed Approach
- Vishakh
- UC Irvine School of Information Computer Science
2Introduction
- Problem Centralized means of providing network
services wont cope. - Solution Bio-Networking architecture (Bionet)
- Problem Bionet needs efficient mechanism to be
robust. - Solution Price Propagation
3Todays Internet The putative news site
- The CNN web server offers web pages with news.
- Users connect to this server and request pages.
4Todays internet Usage patterns and wastage of
resources
- Server bandwidth and capacity based on usage
patterns. - Resource wastage at most times
5Todays internet Another problem
- Spurts in service demand.
- Servers may not cope.
- E.g. online Starr Report in 98
- Failure rates
- CNN.com- 32
- MSNBC.com- 47
- Current methods arent robust!
6Todays Internet The Larger Problem
- Internet expanding and changing.
- Current centralized methods of providing services
wont cope. - We need a way to make services on the internet
more dynamic
7A Solution!- The Bio-Networking Architecture
- The Bio-Networking Architecture (Bionet) is being
developed by Dr. Tatsuya Suda and his team at
UCI. - Decentralized way to provide network services.
- Adaptive, scalable and evolvable.
- Based on biological systems such as ant colonies.
8Bionet Cyber-Entities
- Several agents knows as Cyber-Entities (CEs)
instead of one central server. - CEs travel across the network, providing services
to users. - CE populations adapt to number of users.
9Bionet The Problem
- Ideally
- CE population should constantly change to reflect
user demand. - CE distribution should constantly change to
reflect user locations. - In other words, need to maintain robustness.
- We need a mechanism to regulate reproduction and
migration among CEs in an efficient manner.
10Using Economics
- Economics- how and where to allocate resources to
benefit us optimally. - In Bionet, CE population and distribution must be
set to service users optimally. - Market economies use prices- elegant
decentralized way of gauging supply and demand.
11Using Economics Prices as indicators
- Suppose I am a trader of Llama milk in Peru.
- I look at prices in Peruvian cities.
- I distribute milk where I get can charge highest
price.
12Using Economics Price equilibriums
- Ideally, supply and demand ironed out using
prices. - We have an equilibrium
- Efficient distribution of Llama milk.
- Traders and consumers happy.
- Why not use this for Bionet?
13Using Economics Prices and Bionet
- Users provide CEs with energy, i.e. they supply
it. - CEs look for energy, i.e. they demand it.
- Associate a price with energy, i.e. Energy
Supply/demand
14Using Economics Price propagation
- Price Propagation (PP) is a mechanism for
generating and distributing energy prices over a
network. - Prices can be used by CEs to decide whether to
reproduce and where to migrate.
15Calculating Prices Getting local prices
- Performed by CEs on every node.
- Each CEs local price (LP) Energy supply on
current node/ No. of CEs on current node
16Calculating Prices Getting Global Prices
- CEs calculate Global Price (GP) by averaging
their LP with incoming prices. - E.g. GP2 (3.3 5 20)/3 9.4
17Calculating Prices Propagating Prices
- GPs on each node are broadcast to connected
nodes. - E.g. Node2 sends its LP to Node 1 and Node 3
18Using Prices Migration
- If an incoming price is greater than the Local
Price, then a CE probably needs to migrate. - Each CE compares its Global Price to the Maximum
Incoming Price (MIP).
19Using Prices Migration 2
- CEs calculate migration need (MN) MN
(MIP-GP)/GP - Each CE has a migration factor (?m).
- If MNgt ?m, CEs migrate in the direction of the
MIP.
20Using Prices Migration 3
- CE marked on right migrates from Node 1 to Node
2.
21Using Prices Reproduction
- If a CE notices that the Global Price is
significantly more than a preset ideal GP, it
will reproduce if it can. - This increases energy demand and lowers the GP.
22Using Prices Death
- If a CE notices that the Global Price is much
lower than the ideal preset GP, it can
sacrifice itself. - This brings down the GP and decreases demand for
energy.
23Using prices The Big Picture
- Migration ensures that CEs always close to users.
- Reproduction ensures CEs are present in just the
right numbers.
24Using prices The Big Picture 2
25Current Work
- PP is simple and elegant, yet many wrinkles to be
ironed out. - Refine current algorithms
- Comparing behavior with other algorithms.
26Future Work
- Formal verification of PPs efficiency.
- Integrating more advanced ideas from Economics
and Biology. - More fine-tuning!
27Conclusions
- The internet is changing .
- Bionet is a dynamic, adaptable, evolvable and
scalable alternative to failing centralized
methods. - Price Propagation promising for ensuring proper
agent populations and distributions.
28Acknowledgements
- Dr. Tatsuya Suda
- Dr. Tadashi Nakano
29Links
- The Bio-Networking architecture-
http//netresearch.ics.uci.edu/bionet/ - My email address- vishakh_at_uci.edu
- More information about my research-
http//www.ics.uci.edu/vishakh/research/bionet
(Coming Soon!)