Simulation of a distributed query processing algorithm on token ring networks

• Authors: William Perrizo;V Vananupong
• Venue: Mathematical and Computer Modelling: An International Journal
• Year: 1990

One network architecture which is being considered for information distribution in a Tactical Battle Data Management setting is the token ring network. In token ring local area networks, an information delivery algorithm is simulated which provides a strategy for answering a request for battle information, and also an estimate of the response time required to answer the request. The algorithm is simple, fast, and produces strategies that are competitive with other known algorithms in terms of actual response time. These facts are established through simulation experimentation. Our algorithm and a well established algorithm are analyzed using a computer to simulate their strategies. The relative performance of two retrieval processing strategies generated by these two algorithms are compared on the basis of actual and estimated response time. In order to be unbiased towards any particular algorithm, four different data distributions are used to produce examples of worse, average and best case data. Data for the simulated algorithms are randomly generated. On each run the same data are used to generate the estimated response times and the actual response times for the two algorithms. The evaluation consists of the following comparisons. The estimated response times of the two algorithms are compared. The actual response times of the two algorithms are compared. The actual response time and the estimated response time for each algorithm is compared. Using our algorithm the estimated response time and the actual response time are shown to be close. The result obtained for the well established algorithm show that the actual response time can be significantly larger than the estimated response time. The need for this type of algorithm arises in the Tactical Battle Data Management environment in which decisions must be based on predicted times at which information arrives. This kind of situation requires an algorithm with guaranteed response times. In C^2 real-time applications, where much time is spent checking for retrieval completion, valuable time can be saved since response times can be assured.