Loss Differentiation Algorithms Vs Congestion Control Schemes: Dynamics and Performance
Author(s) -
Aleksandar Milenkoski,
Biljana Stojčevska
Publication year - 2010
Publication title -
international journal of distributed and parallel systems
Language(s) - English
Resource type - Journals
eISSN - 2229-3957
pISSN - 0976-9757
DOI - 10.5121/ijdps.2010.1102
Subject(s) - computer science , network congestion , algorithm , dynamics (music) , control (management) , artificial intelligence , computer network , network packet , physics , acoustics
This paper carefully analyzes the behaviour of different congestion control schemes when used in combination with Loss Differentiation Algorithm. Three types of congestion schemes are discussed: delay-based, bandwidth estimation and AIMD, with one TCP variant representing each congestion scheme. We simulated two network scenarios with diverse link and traffic properties and evaluated the congestion schemes with integrated Loss Differentiation Algorithm in each of them. The integrated Loss Differentiation Algorithm is ideal, i.e. it makes no errors in its judgement. The behaviour of the schemes is analyzed from aspect of: the properties of the employed mathematical functions, the effect of presence or absence of additional network load (reverse and background traffic), and the achieved throughput. The results show very diverse scene and pinpoint the importance of the careful and delicate design of the congestion avoidance action when a non-congestion loss is detected.
Accelerating Research
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom
Address
John Eccles HouseRobert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom