Your browser does not support JavaScript!

Home    Προσαρμόσιμοι αλγόριθμοι μαρκαρίσματος κίνησης για δίκαιη διαμοίραση πόρων σε δίκτυα διαφοροποιημένων υπηρεσιών  

Results - Details

Add to Basket
[Add to Basket]
Identifier uch.csd.msc//2004marinakis
Title Προσαρμόσιμοι αλγόριθμοι μαρκαρίσματος κίνησης για δίκαιη διαμοίραση πόρων σε δίκτυα διαφοροποιημένων υπηρεσιών
Alternative Title Adaptive markers for achieving fairness in Differentiated Services networks
Creator Marinakis, Theocharis M
Abstract Providing Quality of Service (QoS) in IP networks has drawn the attention of the scientific world over the last years. Among various solutions proposed, the Differentiated Services (DiffServ) architecture suggests the use of a small number of mechanisms, which can be used for providing end-to-end QoS over the Internet or other IP networks, in a simple and scalable way. The DiffServ framework introduces a few additional packet-handling schemes (Per-Hop Behaviors - PHBs), among which is the Assured Forwarding PHB, used to build "elastic" services that impose requirements only on throughput. In the common AF architecture, unfairness among heterogeneous traffic sources is observed, concerning sharing of the excess capacity in under-subscribed networks, and degradation of throughput in over-subscribed networks. DiffServ architecture is based on differential packet marking at the network edge based on traffic contracts and differential handling of the marked packets at the core of the network. This study suggests modifications to current marking techniques at the network edge, proposing the use of "adaptive markers" which adapt the marking rate when the traffic mix changes, in order to improve fairness. Two different source sharing models are studied, max-min fairness and proportional sharing based on reserved throughputs, and different adaptive markers are suggested for each case. Finally this paper examines techniques for implementing adaptive markers, and suggests the use of centralized Bandwidth Brokers to adaptively adjust the parameters of the marking algorithms. Extended simulation experiments for various traffic scenarios indicate that the proposed mechanisms achieve to distribute network capacity in a fair way, dealing effectively with every different factor that cause unfairness phenomena, while at the same time they keep their complexity at low levels.
Issue date 2004-04-01
Date available 2004-05-17
Collection   School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
  Type of Work--Post-graduate theses
Views 402

Digital Documents
No preview available

Download document
View document
Views : 2