Section: New Results

Performance analysis and networks protocols

Participants : Paulo Gonçalves Andrade, Thomas Begin, Anthony Busson, Isabelle Guérin Lassous, Laurent Reynaud, Thiago Wanderley Matos de Abreu.

Global computing-network-visualisation.

The PetaFlow application aims to contribute to the use of high performance computational resources for the benefit of society. To this goal the emergence of adequate information and communication technologies with respect to high performance computing-networking-visualisation and their mutual awareness is required. In the work published in [5] , we present the developed technology and the algorithms that we applied to a real global peta-scale data intensive scientific problem with social and medical importance, i.e. human upper airflow modeling.

Performance analysis of multi-hop flows in IEEE 802.11 networks

Multi-hop wireless networks are often regarded as a promising means to extend the limited coverage area offered by WLANs. However, they are usually associated with poor and uncertain performance in terms of available bandwidth and packet losses, which clearly stands as a limitation to their use. In [7] , we consider the performance evaluation of a multi-hop path (also called chain), based on the IEEE 802.11 DCF. The proposed modeling framework is constructive and versatile, so that it can handle various types of multi-hop wireless paths, including scenarios with two flows in opposite directions, and topologies where nodes are exposed to the well-known hidden node problem. The models derived from our framework are conceptually simple, easy to implement and produce generally accurate results for the attained goodput of flows, as well as the datagram loss probability. Typical relative errors for these two quantities are below a few percent. Also, fundamental phenomena occurring in multi-hop wireless networks such as performance collapse and starvation, are well captured by the models.

Passive Measurement-based Estimator for the Standard Deviation of the End-to-End Delay.

Emerging architectures for computer networks such as SDN aim at offering a better handling of flows with stringent requirements of QoS. On the one hand, operators would benefit from a detailed description of common network performance (e.g., end-to-end delay and end-to end loss ratio) including their first two moments, namely mean and standard deviation. Indeed, for many applications, the variability in the end-to-end delay (e.g., jitter) deeply affects the actual QoS experienced by a flow. On the other hand, the cost and nuisance associated with the instrumentation, the measurements, and the computations must be kept as low as possible. This typically prevents the availability of end-to-end measurements. In [30] , we propose an algorithm to estimate the second moment of the end-to-end delay experienced by the packets of a flow based only on delay measurements locally collected by the network nodes. Our solution estimates the standard deviation of the end-to-end delay in an easy and computationally efficient way. Based on thousands of simulations using a real-life trace, our solution is found to be accurate, typically differing by only a few percent from the actual value of the standard deviation of the end-to-end delay.

Design of a force-based controlled mobility on aerial vehicles for pest management.

Vespa velutina, also known as the Asian hornet, is considered as an invasive species out of its native zone. In particular, since it preys on honey bees, its recent progression in Europe could soon pose a significant risk to the local apiculture activity. European beekeepers are therefore investigating adapted control strategies, including V. velutina nest destruction. Unfortunately, nest location pinpointing generally follows a manual process which can prove tedious, time-consuming and inaccurate. In [31] , we propose the use of a network of micro aerial vehicles featuring autonomous and cooperative flight capabilities. We describe an adapted controlled mobility strategy and detail the design of our Virtual Force Protocol (VFP) which allows a swarm of vehicles to track and follow hornets to their nests, while maintaining connectivity through a wireless multi-hop communication route with a remote ground station used to store applicative data such as hornet trajectory and vehicle telemetry. In order to achieve the mission objectives with a minimum of vehicles, we identify through simulations appropriate value for the key parameters of VFP and discuss the obtained network performance.

Channel assignment in IEEE 802.11-based substitution networks.

A substitution network is a rapidly deployable wireless network that provides a backup solution to quickly react to failures on an existing network. We assume that the substitution network uses Wi-Fi technology and that wireless routers are equipped with several Wi-Fi cards. The problem, addressed in this work, deals with the channel assignment to these wireless interfaces. In this particular context, there is only one source-destination pair for which paths are known in advance. It is then possible to derive an objective function, function of the channel assignment, that very precisely reflects the overall throughput that can be achieved in this network. This problemis formulated through a linear optimization problem for which we propose different heuristics. Simulation results, performed with ns-3, consider several scenarios, and compare our heuristics to the optimum. Simulations show that, with only a few wireless cards, the throughput is signifficantly increased. Also, we show that the objective function fits to the throughput measured with ns-3.

Performance evaluation and message dissemination in vehicular networks.

Vehicular Ad-Hoc Network (VANET) is becoming a promising technology for improving the efficiency and the safety of Intelligent Transportation Systems (ITS). Smart vehicles are expected to continuously exchange a huge amount of data either through safety or non-safety messages dedicated for road safety or infotainment and passenger comfort applications, respectively. In this context we proposed two contributions: the estimation of the capacity offered by the wireless network [13] in order to dimension the applications, and the proposal of an efficient message dissemination protocol [25] .

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 customer’s 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 [8] .

Prediction of the System Performance from components models.

In this paper we consider the problem of combining calibrated performance models of system components in order to predict overall system performance. We focus on open workload system models, in which, under certain conditions, obtaining and validating the overall system performance measures can be a simple application of Little's law. We discuss the conditions of applicability of such a simple validation methodology, including examples of successful application, as well as examples where this approach fails. Additionally, we propose to analyze the deviations between the model predictions and system measurements, so as to decide if they correspond to “measurement noise” or if an important system component has not been correctly represented. This approach can be used as an aid in the design of validated system performance models [26] .