ItswTCM: a new aggregate marker to improve fairness in DiffServ

Abstract

Recent demands for real time applications have given rise to a need for Quality of Service (QoS) in the Internet. Differentiated Services (DiffServ) is one of such efforts currently pursued by IETF. Previous researchers found unfairness in the DiffServ network. To solve the unfairness problem, we propose a new TSW based three-color marker (ItswTCM ), which achieves proportional fair share of excess bandwidth among aggregates in a DiffServ network. We have compared the fairness of our proposed ItswTCM marker with srTCM, trTCM, and tswTCM. Results show that our proposed marker performs better than the other three schemes for low to medium network provision level (20 -70%); we believe that all well provisioned network will operate in this region. Results also show that our proposed marker is not as sensitive to the number of flows in an aggregate as the previous marking schemes. We point out that yellow packets play a significant role in achieving proportional fair share of excess bandwidth among aggregates. We conclude that in order to achieve proportional fair sharing of excess bandwidth, it is important to inject right amount of yellow packets into the network. q

Performance modeling of DiffServ meter/markers

“…There are several novel solutions belonging to the category of rate estimator-based meter/markers. They include Improved TSW Three Color Marker (ItswTCM) [14], Enhanced Time Sliding Window (ETSW) [15], Intelligent Traffic Conditioner [16], Lightweight Marker [17], Memory-Based Marker (MBM) [18], the Adaptive CIR Threshold (ACT) mechanism [19], and other. More complete list of meter/marker implementations can be found in [6].…”

An adaptive scheduling scheme for fair bandwidth allocation

“…Different TCP traffic flows in the same service class will be aggregated into TCP aggregates. For ith service in scheduling system in Fig.1, the aggregated throughput B i with N i flows in RIO queue can be given by the throughput formula of TCP aggregate in [12]: (4) where k denotes the average TCP packet size, RT T i,r and p out i,r denote the average round trip time and average loss ratio of OUT packets respectively for the rth flows in ith service queue.…”

“…The former fairness origins in the heterogeneity in traffic, such as the TCP flows with different RTT and UDP flows. The method to solve this fairness problem is to identify the misbehavior flows, such as by marking at the network edge [4,5]. The latter fairness problem comes from the bursty characteristics of Internet traffic, such as the changes of active traffic flows or the subscription ratio of current link.…”

Proportional Bandwidth Distribution in IP Networks Implementing the Assured Forwarding PHB

“…The first goal, assuring the contracted target rate of the final user, has been achieved for many of the published schemes. As regard to the second goal, studies done in [4,5,6] introduce algorithms for achieving proportional fairness in the AF-PHB Service. These proposals have in common the use of three colors for each AF-PHB instance.…”