Modeling, simulation and measurements of queuing delay under long-tail internet traffic (original) (raw)

End-to-end queuing delay assessment in multi-service ip networks

Journal of Statistical Computation and Simulation, 2002

Packet-based networks are more and more used to transport interactive streaming services like telephony and videophony. To guarantee a good quality for these services, the queuing delay and delay jitter introduced in the transport of voice or video flows over the packet-based network should be kept under control. Because data sources tend to increase their sending rate until (a part of) the network is congested, mixing real-time traffic and data traffic in one queue would lead to unacceptable high delays for real-time services. Therefore, voice and video packets need to get preferential treatment (e.g. head-of-line priority) over data packets in the network nodes. Therefore, the queuing behavior of the voice and video packets can be studied more or less independently from the traffic generated by data services. Simple methods to assess the end-to-end delay are primordial. Since it is well known that an aggregate of voice (and CBR video) sources is accurately modeled by a Poisson arrival process and that delays in consecutive nodes are more or less statistically independent, this boils down to developing methods to calculate quantiles of the total queuing delay through a system of N statistically independent M/G/1 nodes. This paper develops four methods to calculate quantiles of the total queuing delay: a Gaussian method, a method based on the numerical inversion of the moment generating function of the total queuing delay developed by Abate & Whitt and two methods based on the assumption that the tail distribution of the individual queuing delay of one node is approximately exponential. The Gaussian method is the simplest, but only gives crude results. The method of Abate & Whitt is the most complex and breaks down for large quantiles. The methods based on the assumption of an exponential tail produce results that are more or less equally accurate as long as there is a node where the load is high enough.

On the Use of Queueing Network Models to Predict the Performance of TCP Connections

Lecture Notes in Computer Science, 2001

In this paper we describe an analytical approach to estimate the performance of greedy and short-lived TCP connections, assuming that only the primitive network parameters are known, and deriving from them round trip time, loss probability and throughput of TCP connections, as well as average completion times in the case of short-lived TCP flows. It exploits the queuing network paradigm to develop one or more 'TCP sub-models' and a 'network sub-model,' that are iteratively solved until convergence. Our modeling approach allows taking into consideration different TCP versions and multi-bottleneck networks, producing solutions at small computational cost. Numerical results for some simple single and multi-bottleneck network topologies are used to prove the accuracy of the analytical performance predictions, and to discuss the common practice of applying to short-lived TCP flows the performance predictions computed in the case of greedy TCP connections.

Long-run performance analysis of a multi-scale TCP traffic model

IEE Proceedings - Communications, 2004

The long-run queueing performance of a multi-scale TCP traffic model, the HOO model, is analysed. Since the model links the multi-scale behaviour with practical traffic elements and approximates TCP traffic very well, the analysis is expected to provide insights into the physical interpretation of multi-scale traffic and to give useful results for performance prediction. To derive a meaningful solution and avoid the extreme difficulty of an exact analysis, the authors adopt several techniques to track the problem, among which are techniques to establish equivalent processes to the traffic process or the queue content process in two cases, the case of fast flows and the case of slow flows. Quantitative results for the queue tails are obtained in both cases, and a unified form is derived. It indicates that three levels of traffic elements in different time scales, i.e. the connection, the burst and the packet, all affect the asymptotic queueing performance. It shows quantitatively how the connection determines the index of the queue tail, and the burst and the packet contribute to the tail with their averages. Used with simple statistical inferences, the analytical result is shown to predict the queueing performance of real traffic well.

Analytical computation of completion time distributions of short-lived TCP connections

Performance Evaluation, 2005

A new technique for the analytical evaluation of distributions (and quantiles) of the completion time of shortlived TCP connections is presented and discussed. The proposed technique derives from known open multiclass queuing network (OMQN) models of the TCP protocol and computes a discrete approximation, with arbitrary accuracy, of the distribution of sojourn times of customers in the OMQN, which corresponds to the distribution of completion times of the modeled TCP connections. The proposed technique is computationally efficient, and its asymptotic complexity is independent of the network topology, of the number of concurrent flows, and of other network parameters.

Large Scale Internet Queueing Delay Tomography

2006

Queuing delay tomography of the Internet is mostly a theoretical research topic, and measurements were mainly performed to prove the validity of a certain measurement methods. We propose a large scale Internet tomography survey to map the queueing delay in the European networks in great details, and the rest of the world to a lesser extent. The measurements will be based on the ETOMIC high accuracy packet capturing infrastructure and on DIMES vast distributed agent community. We present the rational behind the effort, the new technical tools developed to enable it, and some results from initial trials.

Modeling Heavy Tails in Traffic Sources for Network Performance Evaluation

Advances in Intelligent Systems and Computing, 2013

Heavy tails in work loads (file sizes, flow lengths, service times, etc.) have significant negative impact on the performance of queues and networks. In the context of the famous Internet file size data of Crovella and some very recent data sets from a wireless mobility network, we examine the new class of LogPH distributions introduced by Ramaswami for modeling heavy tailed random variables. The fits obtained are validated using separate training and test data sets, and also in terms of the ability of the model to predict performance measures accurately as compared with a trace driven simulation using NS2 of a bottleneck Internet link running a TCP protocol. Use of the LogPH class is motivated by the fact that these distributions have a power law tail and can approximate any distribution arbitrarily closely not just in the tail but in its entire range. In many practical contexts, although the tail exerts significant effect on performance measures, nevertheless the bulk of the data is in the head of the distribution. Our results based on a comparison of the LogPH fit with other classical model fits like Pareto, Weibull, Lognormal, and Log-t demonstrate the greater accuracy achievable by the use of LogPH distributions and also confirm the importance of modeling the distribution in its entire range and not just in the tail.

Calculating End-to-End Queuing Delay for Real-Time Services on an IP Network

Lecture Notes in Computer Science, 2003

A crucial factor for real-time (interactive) services is the end-to-end delay experienced by the application. The contribution resulting from the queuing delay induced by the network nodes is the most difficult to assess. First, it is a stochastic quantity which should be aggregated over many (possibly different) network nodes. Secondly, the queuing delay in a single node stems from two different mechanisms: one related to interference with other interactive flows and one related to interference with the ubiquitous best-effort data flows. Earlier work assessed these two components separately, leading to a 'worst case' result. This paper models both components and develops formulas to calculate exact results for the end-to-end queuing delay. Results are shown indicating an improvement up to 45% over the worst-case method. The formulae developed in this paper are expected to be useful in network dimensioning, in setting network performance requirements and in admission control mechanisms.

Stochastic Models for Throughput Analysis of Randomly Arriving Elastic Flows in the Internet

2002

This paper is about analytical models for calculating the average bandwidth shares obtained by TCP controlled finite file transfers that arrive randomly and share a single (bottleneck) link. Owing to the complex nature of the TCP congestion control algorithm, a single model does not work well for all combinations of network parameters (i.e., mean file size, link capacity, and propagation delay). We propose two models, develop their analyses, and identify the regions of their applicability.

A Simulation Study of the Measurement of Queueing Delay over End-to-End Paths

IEEE Open Journal of the Computer Society, 2020

Determining the qualitative states of the Internet requires an accurate knowledge of queueing delay over an end-to-end path. However, the measurement of queueing delay in a large network is still considered a complex and open problem. Existing schemes that measure queueing delay compensate for this complexity using a high infrastructural support and administrative access to the path under test even though their feasibility and accuracy on the Internet are low. In this paper, we propose an active scheme, called COMPRESS: COMpound Probe compRESSion, to measure queueing delay on all routers over an end-to-end path. The proposed scheme performs per-hop measurement using UDP-based probing packets. It is both simple and self-sufficient in comparison to the existing schemes. We have implemented the proposed scheme in a simulation environment to present a controlled performance evaluation under different levels (e.g., light, moderate, and heavy) and types (e.g., symmetric and asymmetric) of queueing delays over single-and multiple-hop paths. Our simulation results show that the scheme is sensitive to the induced queueing delays and consistently provides a high measurement accuracy. Overall, the scheme has an average measurement error of around 20% or below over the simulated paths.

Assessment of the Impact of Different Queuing Techniques on Different Network Traffic Services

A few years ago, the internet evolved into a communication environment used not only for human and social contact but also for commercial and educational purposes in colleges and schools. As a result, a tremendous amount of new multimedia content has been created, such as interactive environments, 3D videos network gaming, virtual worlds, and other programs that demand a higher bandwidth to function properly. Multimedia applications via IP networks are becoming increasingly popular. The bandwidth consumed has become a major issue for Internet service providers (ISPs) and online communities. Since bandwidth is a limited network resource, numerous types of traffic are employed across the network, including video conferencing, VoIP, and file transfer. As a result, routers employ various traffic management techniques known as queuing approaches to administer these services by managing how packets are delayed while waiting to be dispatched. This article employs three distinct queuing systems, FIFO, PQ, and WFQ, and we attempted to set up three networks with varying traffic volumes and designed network topologies using OPNET tools. These simulations illustrate that the WFQ algorithm performs better under various traffic loads