Quality of service provision in noncooperative networks

Park, Kihong; Sitharam, Meera; Chen, Shaogang

doi:10.1145/288994.289022

Cited by 33 publications

(10 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This implies that multi-media applications|a tra c source with voluminous data rate requirements which promotes a structural environment conducive t o n e t work tra c self-similarity|are also the tra c sources that su er most from their sensitive quality of service (QoS) requirements unless resource reservations or strati cation of best-e ort services is employed. 20 Increasing link bandwidth, given a xed large bu er capacity, has the e ect of decreasing queueing delay m uch more drastically under highly self-similar tra c conditions than when tra c is less self-similar. This suggests that high-bandwidth communication links be employed to alleviate the exponential trade-o relationship between queueing delay and packet loss/throughput for supporting QoS-sensitive tra c. The e ect of bandwidth on queueing delay i s m uch less pronounced when bu er capacity i s small.…”

Section: Introductionmentioning

confidence: 99%

<title>Effect of traffic self-similarity on network performance</title>

Park

Kim

Crovella

1997

Performance and Control of Network Systems

Self Cite

136

View full text Add to dashboard Cite

Recent measurements of network tra c have s h o wn that self-similarity is an ubiquitous phenomenon present i n b o t h local area and wide area tra c traces. In previous work, 1 we h a ve s h o wn a simple, robust application layer causal mechanism of tra c self-similarity, namely, the transfer of les in a network system where the le size distributions are heavy-tailed. In this paper, we study the e ect of scale-invariant burstiness on network performance when the functionality of the transport layer and the interaction of tra c sources sharing bounded network resources is incorporated.First, we s h o w that transport layer mechanisms are important factors in translating the application layer causality into link tra c self-similarity. Network performance as captured by throughput, packet loss rate, and packet retransmission rate degrades gradually with increased heavy-tailedness while queueing delay, response time, and fairness deteriorate more drastically. The degree to which h e a vy-tailedness a ects self-similarity is determined by h o w well congestion control is able to shape a source tra c into an on-average constant output stream while conserving information.Second, we s h o w that increasing network resources such as link bandwidth and bu er capacity results in a superlinear improvement in performance. When large le transfers occur with nonnegligible probability, the incremental improvement in throughput achieved for large bu er sizes is accompanied by long queueing delays vis-a-vis the case when the le size distribution is not heavy-tailed. Bu er utilization continues to remain at a high level implying that further improvement in throughput is only achieved at the expense of a disproportionate increase in queueing delay. A similar trade-o relationship exists between queueing delay and packet loss rate, the curvature of the performance curve being highly sensitive to the degree of self-similarity.Third, we i n vestigate the e ect of congestion control on network performance when subject to highly self-similar tra c conditions. We implement an open-loop congestion control using unreliable transport on top of UDP where the data stream is throttled at the source to achieve a xed arrival rate. Decreasing the arrival rate results in a decline in packet loss rate whereas link utilization increases. In the context of reliable communication, we compare the performance of three versions of TCP|Reno, Tahoe, and Vegas|and we nd that sophistication of control leads to improved performance that is preserved even under highly self-similar tra c conditions. The performance gain from Tahoe to Reno is relatively minor while the performance jump from TCP Reno to Vegas is more pronounced consistent with quantitative results reported elsewhere.

show abstract

Section: Introductionmentioning

confidence: 99%

<title>Effect of traffic self-similarity on network performance</title>

Park

Kim

Crovella

1997

Performance and Control of Network Systems

Self Cite

136

View full text Add to dashboard Cite

show abstract

“…Several works have also studied the [33], [39], [40] priority queueing systems (a la Diffserv) under game-theoretic frameworks. So for example, Marback [33] analyzes a priority queueing scheme where packets get charged based on their priority, and selfish users compete for bandwidth.…”

Section: Related Workmentioning

confidence: 99%

“…A fundamental distinction in this case is that T&C enables different valuations for different classes of traffic, and uses these valuations to leverage the trading system. Park et al [39] consider a QoS class assignment game where users share a single Generalized Processor Sharing (GPS) queue and they can assign the class for the traffic. Users do so, to meet the QoS requirements of their application at the minimum possible costs (as higher priority also means higher cost).…”

Section: Related Workmentioning

confidence: 99%

Trade & Cap: A customer-managed, market-based system for trading bandwidth allowances at a shared link

2011

View full text Add to dashboard Cite

“…This assumption was challenged by [22], where it was shown that for a natural class of games their realistic utility functions-based on a commonly used network model-are not quasiconcave and result in natural QoS provision games that may not have Nash equilibria.…”

Section: Article In Pressmentioning

confidence: 99%

“…One approach by the networking [1,6,5,12,27,26,21] research community over the past several years is to use a microeconomic model: treat the network as a market and its users and providers as players of a noncooperative game [9,6,5,22]. A number of related, fundamental issues have been isolated-in algorithmic mechanism design, computational aspects of game theory, and complexity of distributed computing and communication-that are of interest to theoreticians [2,16,17,19,20,25] and potentially have other applications as well.…”

Section: Introductionmentioning

confidence: 99%