Search

In nonsaturated 802.11 distributed coordination function (DCF) networks, existing studies had observed a salient phenomenon: the maximum stable throughput may be far higher than the saturation throughput and last for a long duration (say, months). Such a maximum stable throughput is called the presaturation throughput peak. The long-duration peak suggests that 802.11 DCF networks may provide a far higher stable throughput than what we generally believe, without worrying about a sudden deterioration of quality of service (QoS). This paper is devoted to developing a general theoretical framework to predict the peak. In the framework, we define a backoff entropy to quantify the uncertainty of the inherent random backoff process of 802.11 DCF networks, and then innovatively introduce a BackoffEntropy-Peak linearity property (i.e., the peak is linearly proportional to the backoff entropy) for the peak prediction. This framework is applicable for a general configuration of contention windows (CWs). It enables us to reveal the most essential dependency of the peak on the initial CW size and the number of network nodes. Further, we design peak-based call admission control (CAC) schemes that can be performed quickly and easily, without requiring network performance measurements and complex calculations. We verify our framework and CAC schemes via widely adopted 802.11 DCF ns2 and ns3 simulators (mostly with typical 802.11 standard settings). Extensive simulations show that our framework and CAC schemes trade within 9% peak prediction errors for making 802.11 DCF networks provide high stable throughput with good QoS (such as short delay and low collision probability). This paper is the first to propose applying the information entropy to study the throughput-stability problem. We believe that this study makes a crucial breakthrough in thoroughly investigating nonsaturated performance and network stability of random access wireless networks. [ABSTRACT FROM AUTHOR]

This article investigates the throughput stability of nonsaturated 802.11 networks with distributed coordination function (DCF). We find that this throughput stability exhibits distinct characteristics in two different regions (or scopes) of parameter settings: the maximum stable throughput is higher than the saturation throughput and lasts for a long duration (say, in order of months) in one region, while the former is bounded by the latter in another region. Our core contribution is an innovative approach (called SatOpt partition) that can identify these two regions quickly and accurately, by a comparison between the saturated and optimal attempt rates. Our partition approach is suitable for a wide range of 802.11 DCF contention window settings, while our partition result is unique. Further, for 802.11 DCF standard settings, we predict the maximum stable throughput successfully, and reveal that it is often higher than the saturation throughput. We verify our approach via the widely adopted 802.11 DCF ns2 simulator (mostly with typical 802.11 DCF standard settings). We speculate that this two-region stability property should be inherent for any random-access networks. This suggests that the research on the throughput-stability problem should differentiate regions of parameter settings and failing to do so might produce inconsistent results. This study is expected to take a critical step toward solving the throughput-stability problem. [ABSTRACT FROM AUTHOR]

Big services, both virtual (e.g., cloud services) and physical (e.g., public transportation), are evolving rapidly to handle and deal with big data. By aggregating services from various domains, big services adopt selection schemes to produce composite service solutions that meet customer requirements. However, unlike traditional service selection, a huge number of big services require some lengthy selection processes to improve the service reliability. In this paper, we propose an efficient big service selection approach based on the coefficient of variation and mixed integer programming that improves the solution in two senses: 1) minimizing the time cost and 2) maximizing the reliability. We tested our approach on real-world datasets, and the experimental results indicated that our approach is superior to others. [ABSTRACT FROM AUTHOR]

With the pervasive use of IEEE 802.11 wireless local area networks (LANs), increasingly more application flows with different quality-of-service (QoS) requirements will be deployed therein, e.g., home multimedia. This paper is concerned with providing utility-optimal throughput guarantees when diverse application flows share the limited bandwidth. Our proposal provides the absolute throughput guarantees for inelastic flows (such as Internet Protocol television) and the proportional throughput guarantees for elastic flows (such as text message) while maximizing the entire system bandwidth utilization. To this end, we first establish a general theoretical framework on the utility-optimal problem. We then present the exact and approximate solutions and prove the existence and uniqueness of the approximate solution. Finally, extensive simulations are performed, which validate that the approximate solution can well achieve the desired objectives. [ABSTRACT FROM PUBLISHER]