Thomas Begin

Evaluation and comparison of MBAC solutions

Admission control is a mechanism used to restrict access to a computer network to some flows based on the current utilization level of the network resource. By regulating the number of on-going flows, admission control aims at preventing overloading, congestion and performance collapses, so that, accepted flows receive a sufficient level of Quality of Service (QoS). In this paper, we evaluate three existing measurement-based admission control (MBAC) solutions, and we compare their efficiency in the context of semantic networks. Semantic networks refer to networks that autonomously acquire a knowledge on the on-going traffic as well as on any new incoming flow requesting admission. In this framework, we configure the three MBAC solutions in a way they have an identical target in terms of maximum tolerable packet loss rate or maximum tolerable packet queueing delay. We evaluate the solutions performance analytically or by simulation, and compare them to the “ideal” admission control. The results show that one solution, outperforms the others in meeting the target performance.1

Promoting quality of service in substitution networks with controlled mobility

A substitution network is a rapidly deployable backup wireless solution to quickly react to network topology changes due to failures or to flash crowd effects on the base network. Unlike other ad hoc and mesh solutions, a substitution network does not attempt to provide new services to customers but rather to restore and maintain at least some of the services available before the failure. Furthermore, a substitution network is not deployed directly for customers but to help the base network to provide services to customers. Therefore, a substitution network is not, by definition, a stand-alone network. In this paper, we describe a quality of service architecture for substitution networks and discuss provisioning, maintenance, as well as adaptation of QoS inside and between the base and the substitution networks.

Towards an automatic modeling tool for observed system behavior

Current computer systems and communication networks tend to be highly complex, and they typically hide their internal structure from their users. Thus, for selected aspects of capacity planning, overload control and related applications, it is useful to have a method allowing one to find good and relatively simple approximations for the observed system behavior. This paper investigates one such approach where we attempt to represent the latter by adequately selecting the parameters of a set of queueing models. We identify a limited number of queueing models that we use as Building Blocks in our procedure. The selected Building Blocks allow us to accurately approximate the measured behavior of a range of different systems. We propose an approach for selecting and combining suitable Building Blocks, as well as for their calibration. We are able to successfully validate our methodology for a number of case studies. Finally, we discuss the potential and the limitations of the proposed approach.

Performance Evaluation of Cloud Computing Centers with General Arrivals and Service

Cloud providers need to size their systems to determine the right amount of resources to allocate as a function of customers needs so as to meet their SLAs (Service Level Agreement), while at the same time minimizing their costs and energy use. Queueing theory based tools are a natural choice when dealing with performance aspects of the QoS (Quality of Service) part of the SLA and forecasting resource utilization. The characteristics of a cloud center lead to a queueing system with multiple servers (nodes) in which there is potentially a very large number of servers and both the arrival and service process can exhibit high variability. We propose to use a G/G/c-like model to represent a cloud system and assess expected performance indices. Given the potentially high number of servers in a cloud system, we present an efficient, fast and easy-to-implement approximate solution. We have extensively validated our approximation against discrete-event simulation for several QoS performance metrics such as task response time and blocking probability with excellent results. We apply our approach to examples of system sizing and our examples clearly demonstrate the importance of taking into account the variability of the tasks arrivals and thus expose the risk of under- or over-provisioning if one relies on a model with Poisson assumptions.

KBAC: Knowledge-Based Admission Control

Many methods have been proposed in the literature to perform admission control in order to provide a sufficient level of Quality of Service (QoS) to accepted flows. In this paper, we introduce a novel data-driven method based on a time-varying model that we refer to as Knowledge-Based Admission Control solution (KBAC). Our KBAC solution consists of three main stages: (i) collect measurements on the on-going traffic over the communication link; (ii) maintain an up-to-date broad view of the link behavior, and feed it to a Knowledge Plane; (iii) model the observed link behavior by a mono-server queue whose parameters are set automatically and which predicts the expected QoS if a flow requesting admission were to be accepted. Our KBAC solution provides a probabilistic guarantee whose admission threshold is either expressed, as a bounded delay or as a bounded loss rate. We run extensive simulations to assess the behavior of our KBAC solution in the case of a delay threshold. The results show that our KBAC solution leads to a good trade-off between flow performance and resource utilization. This ability stems from the quick and automatic adjustment of its admission policy according to the actual variations on the traffic conditions.

Hierarchical modeling of IEEE 802.11 multi-hop wireless networks

IEEE 802.11 is implemented in many wireless networks, including multi-hop networks where communications between nodes are conveyed along a chain. We present a modeling framework to evaluate the performance of flows conveyed through such a chain. Our framework is based on a hierarchical modeling composed of two levels. The lower level is dedicated to the modeling of each node, while the upper level matches the actual topology of the chain. Our approach can handle different topologies, takes into account Bit Error Rate and can be applied to multi-hop flows with rates ranging from light to heavy workloads. We assess the ability of our model to evaluate loss rate, throughput, and end-to-end delay experienced by flows on a simple scenario, where the number of nodes is limited to three. Numerical results show that our model accurately approximates the performance of flows with a relative error typically less than 10%.

A Note on the Effects of Service Time Distribution in the M/G/1 Queue

The M/G/1 queue is a classical model used to represent a large number of real-life computer and networking applications. In this note, we show that, for coefficients of variation of the service time in excess of one, higher-order properties of the service time distribution may have an important effect on the steady-state probability distribution for the number of customers in the M/G/1 queue. As a result, markedly different state probabilities can be observed even though the mean numbers of customers remain the same. This should be kept in mind when sizing buffers based on the mean number of customers in the queue. Influence of higher-order distributional properties can also be important in the M/G/1/K queue where it extends to the mean number of customers itself. Our results have potential implications for the design of benchmarks, as well as the interpretation of their results.

On finding the right balance between fairness and efficiency in WiMAX scheduling through analytical modeling

In this paper, we explore a way to find the right scheduling policy for WiMAX networks, that achieves the best compromise between an efficient use of the resource and a relative fairness among users. This problem is of primary importance as no scheduling policy has been recommended in the WiMAX standard. To do so, we develop an extension of our previous analytical model for WiMAX networks, that takes into account a more general scheduling policy than those previously studied (i.e., instantaneous throughput fairness, slot sharing fairness and opportunistic scheduling). We show that this general policy covers the two extreme cases, namely the instantaneous throughput fairness policy and the opportunistic policy, and offers intermediate policies that are good candidates for finding the right trade-off. In order to formulate the decision criterion, we introduce a new performance parameter, the mean throughput obtained by a user depending on its efficiency to use the resource. The model has a closed-form solution, and all performance parameters can be obtained instantaneously. This allows us to carry out dimensioning studies that require several thousands of evaluations, which would not be tractable with any simulation tool.

Preliminary Results on a Simple Approach to G/G/c-Like Queues

In this paper we consider a multi-server queue with a near general arrival process (represented as an arbitrary state-dependent Coxian distribution), a near general state-dependent Coxian service time distribution and a possibly finite queueing room. In addition to the dependence on the current number of customers in the system, the rate of arrivals and the progress of the service may depend on each other. We propose a semi-numerical method based on the use of conditional probabilities to compute the steady-state queue length distribution in such a queueing system. Our approach is conceptually simple, easy to implement and can be applied to both infinite and finite C m /C k /c -like queues. The proposed method uses a simple fixed-point iteration. In the case of infinite queues, it avoids the need for arbitrary truncation through the use of asymptotic conditional probabilities. This preliminary study examines the computational behavior of the proposed method with a Cox-2 service distribution. Our results indicate that it is robust and performs well even when the number of servers and the coefficient of variation of the service times are relatively high. The number of iterations to attain convergence varies from low tens to several thousand. For example, we are able to solve queues with 1024 servers and the coefficients of variation of the service time and of the time between arrivals set to 4 within 1100 iterations.

Higher-order distributional properties in closed queueing networks

In many real-life computer and networking applications, the distributions of service times, or times between arrivals of requests, or both, can deviate significantly from the memoryless negative exponential distribution that underpins the product-form solution for queueing networks. Frequently, the coefficient of variation of the distributions encountered is well in excess of one, which would be its value for the exponential. For closed queueing networks with non-exponential servers there is no known general exact solution, and most, if not all, approximation methods attempt to account for the general service time distributions through their first two moments. We consider two simple closed queueing networks which we solve exactly using semi-numerical methods. These networks depart from the structure leading to a product-form solution only to the extent that the service time at a single node is non-exponential. We show that not only the coefficients of variation but also higher-order distributional properties can have an important effect on such customary steady-state performance measures as the mean number of customers at a resource or the resource utilization level in a closed network. Additionally, we examine the state that a request finds upon its arrival at a server, which is directly tied to the resulting quality of service. Although the well-known Arrival Theorem holds exactly only for product-form networks of queues, some approximation methods assume that it can be applied to a reasonable degree also in other closed queueing networks. We investigate the validity of this assumption in the two closed queueing models considered. Our results show that, even in the case when there is a single non-exponential server in the network, the state found upon arrival may be highly sensitive to higher-order properties of the service time distribution, beyond its mean and coefficient of variation. This dependence of mean numbers of customers at a server on higher-order distributional properties is in stark contrast with the situation in the familiar open M/G/1 queue. Thus, our results put into question virtually all traditional approximate solutions, which concentrate on the first two moments of service time distributions.