Title: Bio-Networking Architecture
1Bio-Networking Architecture
- Michael Wang
- Larry Chen
- Professor Tatsuya Suda
- UCI Network Research Group
- February 26, 1999
2Outline of Talk
- Motivation and Overview of Bio-Networking
Architecture - Example application of Bio-Networking
Architecture and its features - adaptation and evolution
- security and survivability
- scalability and simplicity
- Bio-Networking Architecture implementation
3Motivation
- Requirements for Future Network Services and
Applications - Adapt to heterogeneous and dynamic conditions
- Secure
- Survivable
- Scalable
- Easy to design and manage
4Large Scale Biological Systems
- Example bee colony
- Key features
- Adapt to environment by direction action
- Evolve to more optimal forms
- Secure
- Survivable
- Scalable
- Relatively simple components (individual bees)
- Unifying concept emergent behavior
5Bio-Networking Architecture
- Model the construction of network services and
applications after biological systems - A service or application is implemented by a
distributed, collective entity called a
super-entity - A super-entity consists of multiple, autonomous
entities called cyber-entities - Cyber-entities have behaviors similar to
biological entities - migration, reproduction, mutation, protection,
interaction
6Bio-Networking Architecture
super-entity
Individual cyber-entities
7Bio-Networking Architecture
- Two additional concepts
- Exchange of services for energy (food, money)
- Cyber-entities try to gain as much energy as
possible, while expending as little as possible - Cyber-entities can store energy
- Directory or service discovery capability called
Social Networking - Well suited for Bio-Networking Architecture
because it can find mobile, replicated
cyber-entities
8Bio-Networking Architecture
- Social Networking
- Cyber-entities establish relationships with
family and friend cyber-entities - Cyber-entities with relationships inform each
other of their super-entity membership and
current network location - When a cyber-entity wants to find a cyber-entity
belonging to another super-entity (a web page or
a service), it queries its family and friends - If they dont have a relationship with a
cyber-entity in the target super-entity, then
they query their family and friends - A set of desired cyber-entities may be returned
from this procedure. The querying cyber-entity
can determine which of the desired cyber-entities
has the best response time.
9Bio-Networking Architecture
- Social Networking Issues
- Overhead recursive querying of friends and
family may cause large amounts of network traffic - queries may be qualified by a maximum propagation
count - initial queries have low propagation count which
is increased if previous queries did not find any
desired cyber-entities - Determinism social network may be partitioned so
that one set of cyber-entities cannot find
another set - need simulation and analysis to determine
probability of this scenario - Security a cyber-entity can lie about who it has
a relationship with - the cyber-entity can compare multiple responses
10Outline of Talk
- Motivation and Overview of the Bio-Networking
Architecture - Example application of the Bio-Networking
Architecture and its features - adaptation and evolution
- security and survivability
- scalability and simplicity
- Bio-Networking Architecture implementation
11Example Super-Presence
- Super-Presence represents a person, organization,
or concept in the network - Super-Presence holds, disseminates, protects
- email address, web pages, files, preferences, etc
- Super-Presence is made up of cyber-entities which
can - migrate
- reproduce with mutations
- protect their bodies
- communicate and interact with other cyber-entities
12Super-Presence
Distributor B
Factory
Corporate and RD Headquarters
Distributor A
Factory
13Super-Presence
Distributor B
Factory
Corporate and RD Headquarters
Distributor A
Factory
14Super-Presence
Distributor B
Factory
Corporate and RD Headquarters
Distributor A
Factory
15Super-Presence
Distributor B
Factory
Corporate and RD Headquarters
Distributor A
Factory
16Super-Presence Adaptation
- Super-Presence adapts its configuration in
response to amount of user demand, location of
user demand, and network resource constraints - This is the emergent result of the
cyber-entities migration and reproduction
behaviors - Administrators do not need to manually replicate
or tune the Super-Presence
17Super-Presence Evolution
- Algorithmic diversity can be automatically
generated and introduced by the designers - Cyber-entities with the most optimal algorithms
live longer and reproduce more - Evolution may produce localized results
- Designer is freed from having to set the optimal
parameters in the new cyber-entities
18Super-Presence Security
- Bio-Architecture based security is used in
conjunction with traditional security techniques
such as encryption and authentication - Replication inherent in Bio-Networking
Architecture gives us additional mechanisms - Variable rate of state update
- Cyber-entities change state after random timeout
period - Consistency checking
- Cyber-entities periodically check each other for
consistency
19Security idea 1 Variable Rate of State Change
- Cyber-entities change their state only after
random timeout
Authenticated (signed) Update Request
20Security idea 2 Consistency Checking
- Cyber-entities periodically check each others
state for consistency
Successful attack
21Super-Presence Survivability
- Replication
- Emergence
- Wide distribution in the network
- Algorithmic diversity
22Super-Presence Scalability
- In the Bio-Networking Architecture, there is no
master entity, so control bottlenecks do not form - Cyber-entities act autonomously, on a local
basis, and using local information. This local
interaction can be repeated as the cyber-entity
population grows
23Super-Presence Reduced Complexity
- Cyber-entity behaviors are relatively simple to
design and implement. The complex behaviors
emerge from collective behaviors and interactions
of the cyber-entities - Administration of Super-Presence is simplified
because it adapts and evolves to heterogeneous
and dynamic network conditions
24Outline of Talk
- Motivation and Overview of the Bio-Networking
Architecture - Example application of the Bio-Networking
Architecture and its features - adaptation and evolution
- security and survivability
- scalability and simplicity
- Bio-Networking Architecture implementation
25Implementation Node Architecture
Resource Cyber-entity
Other Cyber-entities
Resource negotiation Energy exchange
Resource configuration control
Resource access
Bio-Networking Platform Software
Virtual Machine (e.g. Java Virtual Machine)
Unmodified, commercial software
Heterogeneous OS Hardware
26Implementation Platform Software
- Platform software provides
- Cyber-entity execution environment
- protects platform itself and cyber-entities from
each other - Strict resource control
- prevent denial of service through resource
exhaustion - promote natural selection process in
Bio-Networking environment - Migration and lifecycle facilities
- Communications facilities
- Energy management (prevent cheating by
cyber-entities)
27Implementation Platform Software
- Platform software can send a list of all
cyber-entities residing on it to its neighbors.
This is a form of pheromone emission and
propagation. - This allows a cyber-entity to know what other
cyber-entities on neighboring nodes - This information may be used to improve the
performance of the social networking mechanism - Presence of sibling cyber-entities nearby also
affects other cyber-entity behaviors (e.g.
migration, reproduction)
28Implementation Platform Software
- Implementing functionality in platform software
versus implementing functionality in cyber-entity
behaviors - Platform software can implement common
functionality - this reduces size of the cyber-entities
- platform software functionality can be optimized
- Platform software is more secure
- Implementing functionality as a cyber-entity
behavior allows them to be easily changed and
evolved
29Implementation Cyber-entity
ID
Super-entity ID
Attributes
Type
Stored Energy
Age
Non-Executable Data
Body
Cyber-entity
Executable Code
Migration
Replication
Reproduction
Behavior
Protection
Service
Communication
Pheromone Emission
...
30Implementation Behaviors
- A cyber-entity behavior can be implemented by a
number of algorithms - Each algorithm consists of several
- factors - small blocks of code
- parameters - variables
- Factors and parameters are automatically varied
during replication and reproduction - Human designers can introduce new algorithms,
factors, or variables into the population at any
time
31Implementation Migration Behavior
One possible migration algorithm
calculate if more than 80 of requests coming
from a single direction
W2
calculate cost of migration
W1
gt M
If
then migrate
Factors which affect migration
W1, W2, and M are parameters
32Implementation Reproduction
- Factors that affect reproduction behavior
- stored energy level
- current resource cost in the network
- availability of desirable mate
- reproductive aggressiveness
33Implementation Crossover Mutation
cyber-entity A
child cyber-entity
Ownermwang
Behaviors abcde
Info xyz
Ownermwang
Behaviors abcde
Info xyz
child got behaviors a,b from parent A
behaviors c,d from parent B behavior e is a
mutation
Ownermwang
Behaviors abcde
cyber-entity B
Info xyz
34Implementation Pheromone Emission
- Pheromones are emitted by biological entities for
communication (e.g. marking a trail to food,
readiness to mate, signify danger). Pheromones
can be propagated, but decay with distance and
time. Cyber-entities can also use this mechanism
for establishing relationships in social
networking, reproduction, etc. - Factors of the pheromone emission behavior
- information contained in pheromone
- frequency and strength of pheromone emission
35Implementation Increasing Diversity
- Human designers can introduce behavioral
diversity into the cyber-entity population - This will increase rate of adaptation and
evolution - This will also make the population more immune to
specific attacks or failures
36Implementation Deployment
- Bio-Networking Architecture can be deployed
incrementally in todays IP networks - Administrators load Bio-Networking Platform
Software on various computers - Bio-Networking nodes form a virtual network on
top of IP network, similar to mbone - Users can access Bio-Networking services and
applications by download a Bio-Networking enabled
applet into their conventional browsers
37Research Questions
- What are the beneficial concepts and mechanisms
from the biological world? - What is the relationship between individual
behaviors and emergent behaviors? - What is the stability/adaptability tradeoff?
- Can Bio-Networking Architecture evolve fast
enough? - How secure and survivable is the Bio-Networking
Architecture? - What is the performance and overheads of Social
Networking?
38Related Work
- AI and robotics
- use a collection of simple intelligent components
rather than building a monolithic complex
intelligence - Bio-Networking uses the same approach in the
construction of network services and applications
- Artificial Life and Santa Fe Swarm project
- simulate biological processes and emergent
behavior - Bio-Networking applies lessons learned to network
apps - Network applications modeled after the immune
system - Bio-Networking provides a framework for applying
the ideas developed in this area, e.g. mutation,
diversity during reproduction
39Related Work
- Intrusion detection systems
- detects intrusion based on signature or anomaly
- Bio-Networking detects intrusion based on
inconsistency in state. Both approaches can be
used as multiple layers of defense. - Protection of Critical Information Infrastructure
- Presidential commission study saw need to protect
information infrastructure - Bio-Networking is trying to satisfy this new
challenge - Mobile agent systems
- Bio-Networking also consists of mobile agents,
but they are governed by biological principles
40Related Work
- Active Networks
- allows users to add new protocol behaviors in the
network - Bio-Networking allows users to add new
application layer behaviors that can adapt and
evolve based on usage - Web caching
- web objects replicated to reduce server load
- Bio-Networking adds economic model for
determining which objects are most valuable to
cache, deals with updates better, and has
inherent security features - Jini (Sun Microsystems)
- ubiquitous model of computing based on Java
- Bio-Networking is less centralized and can adapt
more
41- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
Bio-Networking Platform w/ limited resources
BlueHat/prodA
BlueHat/prodB
Bio-Networking Platform with abundant (CPU,
memory, disk) resources (high-end server)
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
42- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
BlueHat/prodB
Bio-Networking Platform w/ limited resources
BlueHat/prodA
BlueHat/prodB
Bio-Networking Platform with abundant (CPU,
memory, disk) resources (high-end server)
BlueHat/prodB
BlueHat/prodB
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
43- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
BlueHat/prodB
Bio-Networking Platform w/ limited resources
BlueHat/prodA
BlueHat/prodB
BlueHat/prodC
Bio-Networking Platform with abundant (CPU,
memory, disk) resources (high-end server)
BlueHat/prodB
BlueHat/prodB
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
44- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
BlueHat/prodB
Bio-Networking Platform w/ limited resources
BlueHat/prodA
BlueHat/prodB
BlueHat/prodC
Bio-Networking Platform with abundant (CPU,
memory, disk) resources (high-end server)
BlueHat/prodB
BlueHat/prodC
BlueHat/prodC
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
Bio-Networking Platform w/ limited resources
45- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
46Web Caching Squid
- client browsers are manually configured with
address of proxy - proxy caches web pages
- proxies can be manually configured to look for
web pages in nearby proxies - Difficult issues remain stale cache pages,
dynamic content, tracking hits - cache replacement algorithm(s) do not consider
the value to a user (work related versus
recreational info are cached equally) - no inherent security or survivability features
47Web Caching AWC
- easy migration path from Squid to Adaptive Web
Caching (AWC), only proxies have to be modified - proxies self-organize in a group and share their
cache space to create a larger cache - clients still manually configured to point at a
proxy - difficult issues of stale pages, dynamic content,
tracking page hits are not resolved - cache replacement algorithm does not consider
value to the user - no inherent security or survivability features
48Web Caching Aphid (Bio-Net based)
- cyber-entities each hold a copy of a web page
- cyber-entities autonomously replicate, migrate,
and die based on user demand - clients find closest or least loaded cyber-entity
using the social networking capability of the
Bio-Networking Architecture - update of the web page is propagated to all
cyber-entities (solves the stale web page
problem) - because each cyber-entities is a thread of
execution, it can generate dynamic content,
e.g. mortgage calculator - cyber-entities can track number of hits
49Web Caching Aphid (Bio-Net based)
- cyber-entities receive energy units depending on
the value of their web page, also expends energy
depending on their size (cache replacement
algorithm based on economic principles) - cyber-entities have inherent security features,
e.g. consistency checking - cyber-entities have survivability features, e.g.
there is no master copy of the web page
50Web Caching Aphid (Bio-Net based)
- Limitations
- clients need to download plug-ins which implement
social networking - users/clients may need to indicate value of a
web page request - overhead of communications among cyber-entities
- cyber-entities do not have an algorithm to ensure
globally synchronized, causal update of their
content (this may not be a big problem for web
pages)
51Summary and Conclusion
- Bio-Networking Architecture represents a new
paradigm in the construction of network services
and applications - Services and applications consists of multiple,
autonomous cyber-entities that exhibit emergent
behavior - Bio-Networking Architecture is scalable,
adaptable, evolvable, secure, survivable, and
simple
52More Information
- Bio-Networking Architecture web site
- http//netresearch.ics.uci.edu/bionet
- mwang_at_ics.uci.edu
- larryc_at_ics.uci.edu
- suda_at_ics.uci.edu
53- Aphid A web caching (replication) service based
on the Bio-Networking Architecture
Bio-Networking Platform
Bio-Networking Platform
A
B
C
Bio-Networking Platform
Bio-Networking Platform
Bio-Networking Platform
Bio-Networking Platform
Bio-Networking Platform
Bio-Networking Platform