Search

In this paper we attempt to establish patterns for the Normalized Measure of Dispersion in order to compute the total average end-to-end packet delay in IEEE 802.11 wireless local area network. In order to achieve this, we adopt a reverse engineering approach in which the normalized measure of dispersion is used for determining variation in service times which when applied to the traditional M/G/1 queueing system points toward developing a complete model based on this normalized measure of dispersion and offered load dependency study. From this analysis we have driven nonlinear equations through best fit model for computing the average packet delay. Our analysis can further be extended by deriving nonlinear equations with network parameters for accurately modeling IEEE 802.11 DCF.

Next generation communication networks are moving towards autonomous infrastructures that are capable of working unattended under dynamically changing conditions. The new network architecture involves interactions among unsophisticated entities which may be characterized by constrained resources. From this mass of interactions collective unpredictable behavior emerges in terms of traffic load variations and link capacity fluctuations, leading to congestion. Biological processes found in nature exhibit desirable properties e.g. self-adaptability and robustness, thus providing a desirable basis for such computing environments. This study focuses on streaming applications in sensor networks and on how congestion can be prevented by regulating the rate of each traffic flow based on the Lotka-Volterra population model. Our strategy involves minimal exchange of information and computation burden and is simple to implement at the individual node. Performance evaluations reveal that our approach achieves adaptability to changing traffic loads, scalability and fairness among flows, while providing graceful performance degradation as the offered load increases.

In this paper, we analyze packet delay and energy consumption of the sensor-medium access control (S-MAC) protocol under unsaturated conditions. Since the S-MAC protocol always behaves with a fixed frame length, the frame-based architecture should be reflected in the S-MAC analysis. In addition, the node contention for transmitting a packet is also considered for more practical analysis on packet delay than that in the original paper suggested by Ye et al. [2]. Hence, we employ a Markov chain model to express the S-MAC behavior for the consideration of back-off delay, including node contention. Numerical results show the average packet delay and energy consumption of the node according to offered load and duty cycle where a practical mote running S-MAC is used.

In todays Internet, a trend towards distributed content delivery systems can be observed. These systems still have to offer the same functionality as centralized architectures, e.g. content localization. While several advantages like load distribution and cost reduction could be gained by using a decentralized distribution platform, they come with a tradeoff in resource consumption. In this paper, we analyze and evaluate a decentralized Content Distribution Application (CDA) with respect to the time needed to locate specific content. We introduce a node model based on queueing theory and provide methods to compute important characteristics like the mean search time. In this model, we do not only consider the external offered load, but also the internal traffic created as a consequence of decentralization.

More Recently, researchers have explored to provide a queue management scheme with differentiated loss guarantees for the future Internet. Various types of real time and non-real time traffic with varying requirements are transmitted over the Internet. The sides of a packet drop rate, an each class to differential drop probability on achieving a low delay and high traffic intensity. Improved a queue management scheme to be enhanced to offer a drop probability is desired necessarily. This paper considers multiple random early detection with differential drop probability which is a slightly modified version of the MQDDP model, to get the performance of the best suited, we analyzes its main control parameters (maxth, minth, maxp) for achieving the proportional loss differentiation (PLD) model, and gives their setting guidance from the analytic approach. we propose Dynamic-multiple queue management scheme based on differential drop probability, called Dynamic-MQDDP, is proposed to overcome MQDDP's shortcoming as well as supports static maxp parameter setting values for relative and each class proportional loss differentiation. MQDDP is static according to the situation of the network traffic, Network environment is very dynamic situation. Therefore maxp parameter values needs to modify too to the constantly and dynamic. The verification of the guidance is shown with figuring out loss probability using a proposed algorithm under dynamic offered load and is also selection problem of optimal values of parameters for high traffic intensity and show that Dynamic-MQDDP has the better performance in terms of packet drop rate. We also demonstrated using an ns-2 network simulation.

The Optical Burst Switching (OBS) technology itself is still in the early stage of development and various studies are often performed independently, resulting in difficult comparison between individual data sets (including such controversial studies as burst/packet loss evaluation). To facilitate the future research and evaluation of OBS networks, in this paper we examine relations between standard network parameters and the resulting network offered load, creating a common ground for comparing various simulation results. Means of estimating the resulting network offered load based on parameters describing the network topology and type of traffic are developed and examined for various simulation scenarios (topologies, loads, etc.). It is argued that, given the target offered load value for a given topology, it is always possible to estimate the required idle time which had to be applied in the network nodes, in order to keep the offered network load at the pre-defined level.

Optical Burst Switching (OBS) has intrinsically time-sensitive nature with the separation of the header and the payload. Thus, minute care for offset-time is definitely required, which has not been a highlighted issue in an existing OBS research. In this paper, we focus on the relatively early arrival of the data burst due to the excessive queueing delay of the burst control packet (BCP) under a heavily loaded network. We propose a new scheduling algorithm for the OBS control channel, the Offset-Time Compensation (OTC) algorithm to solve the early arrival problem incurring the data burst drop. In the OTC algorithm, the offset-time is determined by the existing static offset-time scheme and the proposed dynamic offset-time reflecting the offered load of the OBS network. scheme varying with the network condition. By scheduling the BCP, the OTC can reduce the data burst loss rate due to early arrival. With the dynamic offset-time scheme, the data burst loss rate decreases significantly and maintains regular rate regardless of the offered load. We also expend the OTC, QoS-Aware (QA) OTC to provides controllable QoS differentiation in terms of data burst loss rate due to early arrival. The service objective of higher class is satisfied even in a heavily loaded situation.

This paper addresses the requirement for a congestion control mechanism, in order to efficiently handle paging traffic over cellular networks. The standard simultaneous paging approach can easily result in congestion, for relatively small values of the offered load. Instead of differentiating paging capacity based on the incoming load, this approach is more oriented towards a limited-resource system. Considering that a paging channel usually experiences high and low utilization periods, it is proposed to differentiate the paging mechanism according to the processed load. Given medium or even mild overload conditions, several forms of sequential paging substantially decrease blocking probabilities, while presenting good delay behavior. Based on a queueing analysis, we are able to quantitatively estimate potential improvements, and point out the basic points of such an adaptive scheme.

Existing packet-mode schedulers will result in long waiting time for short control packets in IP networks. To overcome this problem, this paper proposes a packet-mode practical scheduling algorithm called short-packets-first (SPF). With uniform Poisson arrival process and low to medium offered load, we prove that SPF can reduce the average packet waiting time for overall packets by greatly lowering the average packet waiting time for short packets. Moreover, simulations under a real traffic model demonstrate that SPF performs very well compared with other packet-mode and cell-mode scheduling algorithms, especially for short packets under heavy offered load..

In this paper, we propose the Optimal Offset-time Decision (OOD) algorithm which can effectively reduce the pre-transmission delay compared to the perfect isolation model, and can also be extended to general n priority classes while the target blocking probability of each class is guaranteed for the variable offered load. In order to drive the OOD algorithm, we first analyze the blocking probability of each priority class through class aggregation and iteration method; the analytic results obtained through the proposed algorithm are then validated with results garnered from extensive simulation tests.

This paper proposes a novel reliable multicast transport protocol for multi-hop, wireless ad hoc networks (or MANETs). To recover from the different types of losses that may occur in MANETs, our Reliable Adaptive Congestion-controlled Transport protocol, or Re-ACT, combines source-based congestion- and error control with receiver-initiated localized recovery. While the latter attempts to recover localized losses (e.g., caused by transmission errors), the former is invoked only for losses and congestion that could not be recovered locally (e.g., caused by global congestion). Loss differentiation is an important component of ReACT and uses medium access control (MAC) layer information to distinguish between different types of losses. Through extensive simulations, we evaluate ReACT’s performance under a variety of MANET scenarios, including different offered load and mobility conditions, and compare it against a strictly end-to-end (i.e., no localized recovery) scheme. Our results show that ReACT is the best performer in terms of reliability. Our results also showcase the effect of ReACT’s local recovery mechanism which quickly corrects error- and path breakage induced losses and thus manages to prevent the source from reducing its rate unnecessarily , thus achieving significant throughput improvement with lower overhead when compared to the strictly end-to-end protocol.

Jobs that run on parallel systems that use gang scheduling for multiprogramming may interact with each other in various ways. These interactions are affected by system parameters such as the level of multiprogramming and the scheduling time quantum. A careful evaluation is therefore required in order to find parameter values that lead to optimal performance. We perform a detailed performance evaluation of three factors affecting scheduling systems running dynamic workloads: multiprogramming level, time quantum, and the use of backfilling for queue management — and how they depend on offered load. Our evaluation is based on synthetic MPI applications running on a real cluster that actually implements the various scheduling schemes. Our results demonstrate the importance of both components of the gang-scheduling plus backfilling combination: gang scheduling reduces response time and slowdown, and backfilling allows doing so with a limited multiprogramming level. This is further improved by using flexible coscheduling rather than strict gang scheduling, as this reduces the constraints and allows for a denser packing.

Budget based network admission control (NAC) mechanisms can be categorized into four basically distinct approaches. Since they have different complexity and efficiency, we compare their resource utilization in different networking scenarios. Our results show that the network size, the connectivity, and the internal structure of the network have a significant impact on the resource efficiency. Some NAC approaches can achieve a very high utilization if the offered load is large enough while the performance of others is limited by the network topology. This study does not focus on specific protocols because the presented NAC schemes classify most existing resource management schemes. It is intended to optimize the NAC design for future QoS networks.

In this paper, we analyze the tradeoffs between signaling and resource utilization in DiffServ networks partitioned in areas (hierarchical domains) using flow aggregates that can be dynamically adjusted. These tradeoffs are studied using two analytical models. In the first model, based on multidimensional birth-death processes, the offered load is detailed at the flow level, which allows accurate assessment of the signaling overhead. The second model accommodates time-varying offered loads, which allows studying the tradeoffs between the time-scale of the aggregate demand and the time-scale of signaling. Our results show that structuring a network domain in areas achieves high performance gains, which can contribute to reduce significantly the cost of core routers.

Admission control for elastic traffic has been advocated in order to maintain performance (i.e. ensure a minimum bandwidth) for each admitted flow and to avoid unnecessary traffic in the network due to retransmissions of packets or even whole transfers after a temporary overload situation. This paper aims at indicating problems of admission control for elastic traffic on a per-TCP-connection basis is problematic in the context of Web traffic: (i) A TCP connection is not equivalent to a transfer. (ii) It is the variance in connection volumes rather than the connection arrival rate that causes most overload situations. (iii) From an application point of view, the target of maintaining performance for admitted flows under high offered load is not met.

The MetaRing is a Medium Access Control (MAC) protocol for Gigabit LANs and MANs with cells removed by the destination stations (slot reuse). Slot reuse increases the aggregate throughput beyond the capacity of single links but can cause starvation. In order to prevent this the MetaRing MAC protocol includes a fairness algorithm. Two types of fairness algorithms have been proposed: “global” and “local”. The paper analyses the local fairness algorithm specified in [CHEN93] by focusing on a worst-case network scenario which is particularly critical with respect to slot reuse. This scenario consists of a group of N equally spaced stations located in one half of the ring and which all send data traffic to a gateway only. The analysis can be divided into two parts. In the first part stations are assumed to operate in asymptotic conditions (never empty queues) and closed formulae for the aggregate and station throughputs are derived. In the second part of the paper the performances of the local fairness algorithm are studied in normal conditions, that is when the offered load is lower (underload conditions) or slightly higher (overload conditions) than the aggregate throughput achievable in asymptotic conditions. The analysis is performed (by simulation) on the same worst-case network scenario. The results obtained show that in asymptotic conditions the local fairness algorithm always allows a complete utilization of the medium capacity. Although it is not perfectly fair, the unfairness can be reduced by choosing of the algorithm parameters appropriately. This choice guarantees fairness in overload conditions as well, but causes unfairness in underload conditions.