QoS provisioning performance of IntServ, DiffServ and DQS with multiclass self‐similar traffic

Abstract

Quality of service (QoS) provisioning, especially delay guarantee, is particularly important to the explosive growth of Internet applications. Traditional QoS schemes include the per‐flow integrated services (IntServ) and per‐class differentiated services (DiffServ) as well as their variants. To address the scalability problem of IntServ and the coarse QoS granularity of DiffServ, we proposed the per‐packet differentiated queueing service. Although the three schemes all aim to provide delay guarantee, the network resource in terms of bandwidth and buffer required by them will vary because of their distinct service disciplines. Indeed, network resources are so precious especially in wireless networks, whereas QoS provisioning is so important that it is necessary to investigate the resource consumption of these schemes to find better QoS solutions for different kinds of networks. Therefore, QoS provisioning performance of the previous schemes should be modelled analytically. To this end, this paper derives the delay bound violation probability for these QoS schemes with the self‐similar and long‐range dependent traffic model because this traffic model can well approximate the Internet traffic, especially delay‐sensitive traffic. The minimum bandwidth and the buffer size suitable for delay guarantee are further derived. Numerical results show that differentiated queueing service consumes the least bandwidth resources and DiffServ can minimise the buffer size whereas IntServ can provide exactly QoS in per‐flow granularity. In this sense, there is a trade‐off between QoS granularity and bandwidth as well as buffer requirements for QoS provisioning. The analytical results are also studied through simulations. Copyright © 2013 John Wiley & Sons, Ltd.

A Differentiated Queueing Service based admission control policy for wireless multimedia

“…In IBG policy, the requested bandwidth of a call is calculated considering the delay requirements similarly to per-flow service rate calculation in literature [3]. Differently, the requested bandwidth in MBG policy is the mean arrival rate of a call.…”

“…m j , a j and H j are traffic parameters of class j packets. Note that, we have illustrated in literature [3] that, a resource shared system gives fair delay bound violation probabilities to arriving flows with DQS service discipline if one DQS queue is used. Hence herein we assume all the incoming calls tolerate identical delay bound violation probability but distinct delay bounds.…”

“…With the given packet granular traffic model as described in subsection II-B, we could approximate the delay bound violation probability based on DQS with the formulas (51) to (53) in [3] as follows…”

“…DQS provides QoS support for admitted traffic via per packet buffering and scheduling based on the actual delay experience of each packet, thus is feasible for mixed streams with various QoS requirements in multimedia applications. The literature [3] has illustrated that, the perpacket DQS scheduling discipline requires the least bandwidth in provisioning the same packet-level QoSs for mixed streams with various delay bounds in comparison with per-flow and per-class QoS schemes. However, it still lacks a resourceprovisioning scheme at call level.…”

Dynamic frame aggregation scheduler for multimedia applications in IEEE 802.11n networks

“…In the literature, there are numerous schedulers that exist such as first in first out, priority queuing, earliest due date, fair queuing, round robin, random access, weighted fair queuing, and weighted round robin in addition to other well‐known opportunistic scheduling algorithms such as proportional fair, maximum signal‐to‐noise ratio and weighted fair opportunistic .…”