1 AIT Asian Institute of Technology

A study of performance of a threshold based round-robin matching algorithm in virtual output queue switch under self-similar traffic

AuthorPhirada Sumpunsapawa
Call NumberAIT RSPR no.TC-02-08
Subject(s)Simulation methods
Algorithms

NoteA research study submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Advanced Technologies
PublisherAsian Institute of Technology
Series StatementResearch studies project report ; no. TC-02-08
AbstractReal network traffic is highly correlated from cell to cell, and cells tend to arrive in bursts. Self-similar traffic model has been widely accepted as a potential tool for burst modeling. The arrival process feeding each port of an input queued switch from Selfsimilar traffic can lead to a very bad queuing behavior. This research studies the performance of a virtual output queue switch (VOQ) under Self-similar traffic. Delay performance in a Threshold based Round-Robin Matching (TRRM) and Single Iterative Parallel (SLIP) matching algorithm is studied under both Bernoulli and Self-Similar traffic. The traffic trace is generated by using linear approximate Fractional Gaussian Noise with varied Hurst parameter values (0.6 ... 0.9). The effect of different offered utilization from 50% to 90% and Switch sizes are also investigated in this work. Simulation results expose that the degree of Self-similarity of traffic has an impact on network switch. For the high Hurst parameter values, Average delay of cells and delay distribution are increased due to Self-similar traffic strongly. For example under 90% offered load and 16 x 16 switch size, Hurst value 0.9 achieves mean cell delay of 120 cells, whereas Hurst value 0.6 has delay of 114 cells. Therefore, the delay of VOQ switch under Self-Similar traffic is also higher than for the Bernoulli traffic (H=0.5). When the total offered load and switch size are extended, the delay will increase. The results show that the TRRM algorithm works efficiently and improve performance especially at heavy load when compared with SLIP algorithm. In studied cases, TRRM can reduce average delay at most 50% comparing with SLIP algorithm.
Year2002
Corresponding Series Added EntryAsian Institute of Technology. Research studies project report ; no. TC-02-08
TypeResearch Study Project Report (RSPR)
SchoolSchool of Advanced Technologies (SAT)
DepartmentDepartment of Information and Communications Technologies (DICT)
Academic Program/FoSTelecommunications (TC)
Chairperson(s)Erke, Tapio;
Examination Committee(s)Ahmed, Kazi M.;Teerapat Sanguankotchakorn;
Scholarship Donor(s)-
DegreeResearch Studies Project Report (M.Eng.) - Asian Institute of Technology, 2002


Usage Metrics
View Detail0
Read PDF0
Download PDF0