Social Networking Project by Thuan Do and Jiming Liu. Outline * Introduction * Project Scope * Mechanism * Results * Discussions * Conclusion * Future Work.
* IntroductionWe design a network, which consists of up to MAX_SUPER super-entities, and each super entity contains up to MAX_ENTITY cyber-entities. Example: up to 10 families, and each family has up to 10 people.
Each super-entity (or family) is like a complete graph. All cyber-entities inside the same family contain the same kind of data and they are all linked together. They all know about the positions of one another.
We will observe the behavior of the network in two cases: static and dynamic. We will compare them in term of the efficiency for locating services and saving of resources.
Both static and dynamic network have the same design at the beginning. The static network will remain the same over time. The dynamic network will evolve. It has “replication” and “deleting” functions for cyber-entities.
At first, we read in the number of super-entities, then the number of cyber-entities in each super-entity, then we randomly generate the coordinates of each cyber-entity as (x,y) within the range of the area.
SEARCHING: At first, if the seeker finds the target family among his closest friends, we stop searching and begin to calculate the coordinate of the target family member which is closest to the seeker.
If the target is not among his closest friends, we will put all of his family’s friends on a queue, then process each friend at a time. This is the same as “breadth first search”. By doing this way, we make sure that the closest target family member will be found first.
- Performance with Replication versus without Replication:Energy at some cyber-entities in the static system becomes very high, and they cannot serve all the requests.
In the dynamic system, we have replications, and we were able to keep the energy low for all cyber-entities, thus, the service is good for all requests.
At first, if we deleted all cyber-entities that got no request after a while: their friend-links were deleted also, and the network was ruined.
We must be careful to choose a policy for deletion: we delete the cyber-entities that got no request after a certain time AND also have no friend-links.
In our small example, we tried to limit the number of hops around of the searching, and we found that the request message would not need to hop more than 6 times in order to find a target family
We also run our program on a bigger network, having 100 super-entities, and each super-entities have 1,2 or 3 cyber-entities. We randomly generated a few links for each super-entities.
The result is the same. The searching succeeds after only some hops, in most cases, two or three hops, since we do not count hopping to a sibling, or within a super-entity, as a hop.
1- Within the scope of this class, we were not able to implement the “moving” of resources toward “clients”. We can implement that movement into our program in a next simulation.
2 - We would like to investigate bigger networks. It suffice to generate a bigger input-file, with many super-entities and lots of requests. We will observe the evolution of the Bio-network over time, in a more real life situation.