Objective

The Bio-Networking Architecture that the PI proposes is inspired by the observation that the biological world has already developed the mechanisms necessary to achieve the key requirements for the Next Generation Internet (NGI), such as scalability, adaptability to heterogeneous and dynamic conditions, security, survivability, and simplicity. In the biological world, each individual entity (e.g., a bee in a bee colony) follows a simple set of behavior rules (e.g., migration, reproduction, energy exchange, mutation, and death), yet a group of entities (e.g. a bee colony) exhibits complex, emergent behavior (e.g., adaptation, evolution, security, survivability). Therefore, if services and applications adopt biological concepts and mechanisms, they too may be able to achieve the key requirements of NGI.

In this project, the PI explores an innovative idea of applying key concepts and mechanisms from the biological world onto network application designs to enable construction of large scale network applications.
 
 

Approach

The PI is currently applying key concepts and mechanisms from the biological world to design and empirically evaluate a new network architecture called the Bio-Networking Architecture. In the Bio-Networking Architecture, services and applications are implemented by super-entity, i.e., a collection of multiple entities called cyber-entities (as a bee colony consists of multiple bees). (See Fig.1). These cyber-entities have functionality related to their service or application and follow simple behavior rules (e.g., migration, reproduction, energy exchange, mutation, death) similar to biological entities. (See Fig.2) The Bio-Networking platform software on each node in the Bio-Networking Architecture provides an execution environment and supporting facilities for cyber-entities. (See Fig.3.) A specialized cyber-entity, the resource cyber-entity, manages and allocates resources to the other cyber-entities on the network node. In the Bio-Networking Architecture, useful emergent behaviors (e.g., adaptation, evolution, security, and survivability) result when individual cyber-entities interact.
 

Figure 1: Super-entity

Figure 2: Cyber-entity Components

 

Figure 3: Bio-Networking Node

The innovative claims of the proposed project are:

• The proposed Bio-Networking Architecture is the first attempt to apply the biological concepts of emergent behavior, autonomous control, and adaptation and evolution to a broad and general class of network services and applications. Previous studies of biological architectures were for the purpose of understanding biological processes, simulating life or emergent behavior, distributed computation, or robotics.

• The Bio-Networking Architecture enables the construction of services and applications which meet the key requirements of NGI. Scalability is achieved because cyber-entities act autonomously, and on a local basis using only local information. Construction of the service or application is simplified because only relatively simple behaviors at the cyber-entity level need to be designed. The other requirements of adaptation, security, and survivability are simply the emergent behavior of the cyber-entities acting collectively.

• Services and applications designed using the Bio-Networking Architecture adapt to heterogeneous and dynamically changing network conditions through the autonomous actions of their cyber-entities, each exhibiting simple behaviors. They also evolve to more desirable behaviors through the mutation and natural selection mechanisms of the Bio-Networking Architecture. Current network services must be manually configured and optimized to network conditions.

• Services and applications can use, in addition to traditional security techniques such as authentication and encryption, features of Bio-Networking Architecture such as intrusion detection, autonomous replication, algorithmic diversity, and dispersion of the cyber-entities, as additional layers. Current network services and applications have only one layer of defense, and thus they are unlikely to survive coordinated attacks.

• The Bio-Networking Architecture provides a new approach to the design of large, complex network services such as name resolution, service discovery, and load balancing.

Recent Accomplishment

Aphid: An Adaptive and Scalable Web Content Distribution Application

To demonstrate the adaptability and scalability, as well as to obtain a better understanding of the Bio-Networking Architecture, we have designed an adaptive and scalable web content distribution application called Aphid.

Aphid cyber-entities are constructed according to the design shown in Figure 2. The attributes section of the Aphid cyber-entity contains information about itself, including a unique ID, which web site (super-entity) the cyber-entity belongs to, what web pages are in its body, and what system resources are required by the cyber-entity. The body section of the Aphid cyber-entity contains the web pages that they disseminate. The behavior section of the Aphid cyber-entity contains behaviors (executable methods or functions) that control the autonomous actions of the cyber-entity. They include but are not limited to:

• A service behavior that accepts HTTP requests for web pages and delivers them to the requester.
• A resource buying behavior that uses the cyber-entity’s energy units to buy resources, such as CPU time, memory space, and network bandwidth, from the Bio-Networking platform software.
• A migration behavior which causes migration towards users.
• A reproduction behavior which causes reproduction during periods of high user demand.
• A death behavior which causes the cyber-entity to die during periods of low user demand.
• A relationship behavior which allows a cyber-entity to be in contact with a subset of the other Aphid cyber-entities.

To receive a web page held by an Aphid cyber-entity, users must first locate a nearby cyber-entity containing the desired web page. This can be done via social networking, a fully distributed directory service for mobile and replicated objects currently under development as part of the Bio-Networking Architecture project, or by some conventional directory service, such as DNS. When a user receives a web page from an Aphid cyber-entity, the user pays the cyber-entity with energy units.

Recent Accomplishment: Aphid Simulations

We have developed a simulator for the Bio-Networking Architecture and Aphid. The simulator can simulate a wide variety of network topologies and user demand workloads. In the following, we will present some preliminary simulation results.

The simulation results are based on the network topology shown in Figure 4 and the user workload shown in Figure 5. In Figure 4, the circles represent Bio-net nodes and the numbers next to them are their ID numbers. Figure 5 shows the number of users requesting web pages over a 24 hour period. During most of the day, the number of users vary randomly between 0 and 5. However, there is a surge in user demand between 7am and 10am. Users are randomly placed in the network.

 Three separate simulation runs were conducted. In the first run, a single Aphid cyber-entity was placed on node 9. In the second run, 4 "static" servers were placed on nodes 1, 4, 9, and 13. Static servers are conventional software web servers that do not migrate, reproduce, or die. In the third run, 6 static servers were placed on nodes 1, 4, 9, 13, 8, and 6. In simulations, each user requests one web page per second from the nearest Aphid cyber-entity or static server and pays that cyber-entity or static server 1 energy unit. Aphid cyber-entities and static servers can process 50 requests per second and must pay the Bio-net nodes they are residing on 3 energy units per second. Figures 6 and 7 show the results of the 3 simulation runs super-imposed on each other for comparison purposes.

Figure 6 shows the average delay experienced by users. Of the three simulation runs, users experience the highest delay, approximately 60 seconds, when 4 static servers are used. This is caused by an imbalance of users relative to the location of the servers. As a result, some servers are overloaded with user requests for a short period of time. When 6 static servers are used, there is enough excess processing capacity to overcome any imbalance in user locations. When Aphid cyber-entities are used, users experience a slight increase in delay at the beginning of the surge in user demand. This is due to the fact that there is only 1 cyber-entity in the network at the beginning of the surge. However, the initial Aphid cyber-entity is able to quickly replicate to accommodate the demand.

Figure 7 shows the resource cost when 4 static servers, 6 static servers, and Aphid cyber-entities are used in the simulation runs. Resource costs are defined as the total number of energy units a static server or cyber-entity pays to the platform. Aphid cyber-entities use approximately one fourth the resources of 6 static servers. The cost saving achieved by Aphid cyber-entities is due to their death behaviors which causes excess cyber-entities to die during periods of low user demand.
 

Figure 6: Average Delay Experienced by Users

Figure 7: Resource Cost for 1 Day

Even though the 6 static servers produced the lowest delay, Aphid cyber-entities were able to produce comparable delay characteristics at a lower resource cost. Due to limited space, we only presented a limited set of simulation results. However, we believe they show that the approach used by Bio-Networking Architecture is promising.

Recent Accomplishment: Aphid Implementation

The PI is currently implementing Aphid to show the feasibility of the Bio-Networking Architecture. The current implementation effort focuses around Aphid cyber-entities to study how biological behaviors, such as replication, migration, energy consumption and death of cyber entities, can be implemented with a current programming language. The platform software will be implemented when the implementation of Aphid cyber-entities is completed. In this subsection, the current implementation status is briefly summarized.

As described in Approach section, the Aphid cyber-entity has three sections: attributes, body and behaviors. In the current implementation, the attributes section stores the name, energy level, and the current location of the cyber entity. The body section implements the web server functionality that processes HTTP "get" command. The HTML pages and images that the Aphid cyber-entity serves are also a part of the cyber entity body. To make the HTTP service more responsive and efficient to user requests, the body section of the Aphid cyber-entity runs in an individual thread created by the Aphid cyber-entity in initialization. The behaviors in the current implementation are the methods implemented in the Aphid cyber-entity. The behaviors implemented so far include replication, migration, energy consumption, and death.

Running Aphid cyber entities requires the support of the Bio-Networking platform software as described in Approach section. In the current implementation, commercially available Voyager ORB, a mobile agent platform developed by ObjectSpace Inc., is used as a platform software. Voyager ORB supports migration, communication, and lookup service of mobile objects. In order to aid cyber-entity migration, we have augmented the Voyager ORB with functionality needed to support hop-by-hop migration of cyber entities toward the users.

Aphid cyber entities receive energy from users for the services they provide. In the current implementation, energy is expressed in a positive integer. When requesting web pages from an Aphid cyber–entity, a user specifies the amount of energy he is willing to pay at the end of the URL string. For instance, www.uci.edu/index.html$40 indicates that a user is paying 40 energy units to the Aphid cyber-entity. Aphid cyber-entities also pay energy to the platform software that it is running on. This is currently simulated by letting the Aphid cyber-entity to decrease its energy level by 1 every second. When an Aphid cyber exhausts all energy, it dies and releases resources.

The Aphid cyber-entity is implemented in Java and tested in a Windows NT 4.0 environment. In a demonstration over a small scale testbed of 3 nodes, we observed that Aphid cyber-entities collect energy from users, replicate when their energy level reaches a threshold, migrate toward a user who provides the largest energy amount for the service, and die when the energy is exhausted.
 
 
 

Current Plan

As described in Research Accomplishment section, the PI has initiated feasibility study of the proposed Bio-Networking Architecture through simulation and implementation of an example application, Aphid. Initial results show that the proposed architecture exhibits such key features as adaptability, survivability and availability. The proposed research will follow the three major phases described below.

Task 1: Extensive Simulation and Analysis

Large-scale simulation of the Bio-Networking Architecture will be conducted. Various biological concepts and mechanisms will be simulated, and their benefits and overheads empirically evaluated. Major tasks in this phase include:

• Design and implement a simulation environment for the Bio-Networking Architecture. The simulator will be capable of simulating a large, heterogeneous network.
• Empirically evaluate the benefits and overheads of various biological mechanisms in the simulation environment
• Demonstrate in the simulation environment that the Bio-Networking Architecture is capable of generating the desired emergent behaviors such as adaptation and survivability.

Components of the Bio-Networking Architecture will be designed. The major tasks include:

• Design Bio-Networking platform software.
• Design cyber-entities and cyber-entity behaviors.

Components of the Bio-Networking Architecture designed in Task 2 will be implemented on a test network. The major tasks in this phase include:

• Develop prototype implementation of various components of the Bio-Networking Architecture, including the Bio-Networking platform software and cyber-entities, in the Java environment.
• Demonstrate basic operation of Bio-Networking components on a test network of PCs running the Bio-Networking platform software.

Technology Transition

The project is relatively young, and the DARPA’s support for the project started in August of this year. Thus, the PI has been focusing on the feasibility study aspects of the Bio-Networking Architecture and has not made significant contributions in the area of technology transfer.