Loading...

Deterministic and Pribabilistic Bounds on Traffic and Service and their Application on Network Performance Analysis

Amiri, Ebrahim | 2013

473 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 45098 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Golestani, Jamaloddin
  7. Abstract:
  8. In a data network, expressing an appropriate model for traffic and service is necessary for the guarantee of quality of service (QoS) parameters, e.g. delay and backlogged traffic. There are four main approaches for modeling the traffic and service. First approach is classic queuing theory. It is a useful tool for analyzing the networks with memory less input traffic. Second approach is deterministic network calculus. In this approach some deterministic bounds on delay and backlogged traffic can be derived. However, it is not accurate for analyzing the bursty traffic and wireless networks, because only trivial bounds can be derived in these scenarios. Third approach is effective bandwidth model for traffic and effective capacity model for service. With these models, asymptotical behavior of delay and backlogged traffic probability distribution in steady state can be expressed. In other word, under some conditions, probability distribution of backlogged traffic and delay approach to exponential distribution when the queue length and delay is large enough. Fourth approach is stochastic network calculus that considers the (σ(θ),ρ(θ)) model for traffic and (b(θ),r(θ)) for service from the moment generating function of traffic and service. With this approach, some probabilistic bounds on delay and overflow probability can be derived. In this dissertation, after defining these two models and analyzing a single queue, we extend previous results to multihop networks. At last, we maximize throughput of a multihop network under a constraint on overflow probability
  9. Keywords:
  10. Stochastic Analysis ; Effective Bandwidth ; Effective Capacity ; Network Analysis ; Service Quality

 Digital Object List

 Bookmark

No TOC