موازنة الاحتمال لزبائن يتغيرون بشكل مستمر نهج نظرية اللعبة
Load Balancing for Dynamic Clients, Game Theory Approach
عوني سيف الدين جميل عبيد
AWNI SAIF ALDEEN JAMEEL OBAID
MetadataShow full item record
We address the problem of load balancing which is considered as a technique to spread work between two or more computers in order to get optimal response time and resource utilization between servers by using Nash equilibrium which is the central concern of game theory. We use a normal form table to express the payoff for every client, every client has estimate time for execution and every server has speed, from these facts, we innovate a dynamic payoff matrix to evaluate a Nash equilibrium point and then determine which server can serve an appropriate client to achieve best load balancing which is called "server matching" (Each client is matched to exactly one server, but a server can be matched to multiple clients or none.). We use Netlogo simulation to implement this matching, besides a useful game theory oolset called GAMBIT (Gambit toolset homepage, 2005) to solve the payoff matrix and compute Nash equilibrium point. This paper was done between the years 2007 – 2009, and to know what was done in this paper we can argue that we contribute in accomplishing the load balancing between servers, using new technique depends on game theory perspective, for that reason we can answer the question why this thesis was done, because it is very vital in achieving this goal. We use in our thesis a simulation methodology to prove the results that we have obtained from the simulation program which is a Netlogo V4.0.2, and we compare the results with traditionally techniques in load balancing. Finally, the results show that we improve the performance for the whole system by 4% in achieving load balancing and overweight the possibilities of using this technique in real system around the world.
- Computer Science