Sharing Memory Optimally

Efficient design of service facilities, such as data or computer networks that meet random demands, often leads to the sharing of resources among users. Contention for the use of a resource results in queueing. The waiting room is a part of any such service facility. The number of accepted service requests per unit of time (throughput), or the fraction of the time the servers are busy (utilization), are often used as performance measures to compare designs. Most common models in queueing theory consider the design of the waiting rooms with the assumption that, although individual requests may differ from one another, they are statistically indistinguishable. However, there are several instances where available information allows us to classify the requests for service into different types. In such cases the design of the service facility not only involves the determination of an optimum size for the waiting room but also the rules of sharing it among the different types. Even with a fixed set of resources, the rules of sharing them can influence performance. In data networks (or computer networks) the "waiting room" consists of memory of one kind or another. Messages (jobs) destined for different locations (processors) sharing common storage is an important example of shared use of memory. Recently, Kleinrock and Kamoun have modeled such use of memory and computed the performance of various policies for managing the allocation of memory to several types of users. Decisions to accept or reject a demand for service were based on the number of waiting requests of each type. However, the optimal policy was not determined even in the case where there were only two types of users. We determine the structure of optimal policies for the model considered with three types of users. The optimal policy consists of limiting the number of waiting requests of each type, and reserving a part of the memory to each type.     

A fine-grained Ethernet performance model

We have developed a novel analytical Ethernet performance model that contains a complete formulation of system response time from the initiation of user requests for service to the delivery of the service to the user. The estimated response time includes simultaneous input bus delay (requests for service from multiple users), simultaneous output delay (responses to requests from multiple servers), considering time lost to collisions and subsequent backoff time, time lost to deferring transmission due to a busy channel, and transmission time. In addition, we include server disk input and output queueing times and server and user disk access times. By modeling all epochs of collision, backoff, and transmission events, we have developed a way of providing a level of detail and accuracy that other analytic models of Ethernet performance do not provide. In addition to using the model for estimating response time as a function of number of users and servers, the model is used to determine the number of servers that would be appropriate for a given range of number of users.     

Queueing analyses of traffic access control strategies withpreemptive and nonpreemptive disciplines in wideband integrated networks

A multiserver queueing model of access control strategies for a wideband integrated services digital network (ISDN) is considered. There are two types of service requests (SRs): Type 1 SR requires b servers (basic bandwidth units, BBUs) of the c available servers, while Type 2 SR requires a single server. Both types of SRs are queuable in two separate infinite-size buffers. A Type 1 SR is allowed to seize d(⩽b) servers at first and then acquire additional servers as they become available to meet the original requirement of b servers. The arrival processes of both types of SRs are Poisson and the service times are exponentially distributed with mean b/dμ₁ and 1/μ₂ for Type 1 SR and Type 2 SR, respectively. Preemptive and nonpreemptive priority disciplines with movable boundary are analyzed using the Neuts' matrix-analytic approach. Numerical examples of the queue length distribution and the mean waiting time of the SRs are presented     

An on-line, business-oriented optimization of performance and availability for utility computing

Utility computing provides a pay-as-you-go approach to information systems in which application providers (e.g., web sites) can better manage their costs by adding capacity in response to increased demands and shedding capacity when it is no longer needed. This paper addresses application providers who use clusters of servers. Our work develops a framework to determine the number of servers that minimizes the sum of quality-of-service (QoS) costs resulting from service level penalties and server holding costs for the server cluster. The server characteristics considered are service rate, failure rates, repair rates, and costs. The contributions of this paper are: 1) a model for the performance and availability of an e-Commerce system that is consistent with data from a multisystem testbed with an e-Commerce workload; 2) a business-oriented cost model for resource allocation for application providers; 3) a closed form approximation for the optimal allocation of servers for an application provider based on the performance model in 1) and the cost model in 2); and 4) a simple criteria for utility owners and server manufacturers to make tradeoffs between server characteristics.     

Performance model of interactive video-on-demand systems

An interactive video-on-demand (VoD) system allows users to access video services, such as movies, electronic encyclopedia, interactive games, and educational videos from video servers on a broadband network. This paper develops a performance evaluation tool for the system design. In particular, a user activity model is developed to describe the usage of system resources, i.e., network bandwidth and video server usage, by a user as it interacts with the service. In addition, we allow batching of user requests, and the effect of such batching is captured in a batching model. Our proposed queueing model integrates both the user activity and the batching model. This model can be used to determine the requirements of network bandwidth and video server and, hence, the trade-off in communication and storage costs for different system resource configurations     

Dynamic parallel access to replicated content in the Internet

Popular content is frequently replicated in multiple servers or caches in the Internet to offload origin servers and improve end-user experience. However, choosing the best server is a nontrivial task and a bad choice may provide poor end user experience. In contrast to retrieving a file from a single server, we propose a parallel-access scheme where end users access multiple servers at the same time, fetching different portions of that file from different servers and reassembling them locally. The amount of data retrieved from a particular server depends on the resources available at that server or along the path from the user to the server. Faster servers deliver bigger portions of a file while slower servers deliver smaller portions. If the available resources at a server or along the path change during the download of a file, a dynamic parallel access automatically shifts the load from congested locations to less loaded parts (server and links) of the Internet. The end result is that users experience significant speedups and very consistent response times. Moreover, there is no need for complicated server selection algorithms and load is dynamically shared among all servers. The dynamic parallel-access scheme presented does not require any modifications to servers or content and can be easily included in browsers, peer-to-peer applications or content distribution networks to speed up delivery of popular content.     

服务器个数随机可变的轮询系统分析

前人关于轮询系统的大量研究均以服务器无故障为前提。本文提出了一种容许服务器个数随机可变的多服务器轮询模型。同时考虑排队延迟与服务器故障／修复的影响，获得了顾客等待时间的平均值公式。     

Queues with two units connected in series with a multiserver general bulk service in unit II

This paper analyses a queueing model consisting of two units I and II connected in series, separated by a finite buffer of size N. Unit I has only one exponential server capable of serving customers one at a time. Unit II consists of c parallel exponential servers and they serve customers in groups according to the general bulk service rule. The steady state probability vector of the number of customers waiting and receiving service in unit I and waiting in the buffer is obtained using the modified matrix-geometric method. Numerical results are also presented.     

基于NOMA的超密集MEC网络任务卸载和资源分配方案

为了提高移动边缘计算(mobile edge computing,MEC)网络的频谱效率,满足大量用户的服务需求,建立了基于非正交多址接入(non-orthogonal multiple access,NOMA)的超密集MEC系统模型。为了解决多个用户同时卸载带来的严重通信干扰等问题,以高效利用边缘服务器资源,提出了一种联合任务卸载和资源分配的优化方案,在满足用户服务质量的前提下最小化系统总能耗。该方案联合考虑了卸载决策、功率控制、计算资源和子信道资源分配。仿真结果表明,与其他卸载方案相比,所提方案可以在满足用户服务质量的前提下有效降低系统能耗。     

Trafic d’un nombre limité de sources mélangeant attente et perte

We consider a multiserver queueing system carrying traffic from a finite-size population (Engset model). When faced with congestion, some of the requests vanish (lost calls), while the others are allowed to wait for a free server, resulting in a mixed (waiting + loss) system. Grade of service requirements lead to calculate loss probability and waiting time distribution — since standards usually upperbound waiting times.     

远程故障诊断系统的排队论分析

引入随机服务系统,分析了远程故障诊断系统,建立了基于等待制的M/M/C/∞排队系统模型,并由此分析得到排队队长和等待时间等系统运行的指标。为平衡客户对故障诊断及时性的要求与系统维护费用之间的关系,根据排队等待时间和诊断服务器数量的关系,结合服务器的有效利用率,提出确定服务器最佳配置数量的方法。并以变频器的远程故障诊断系统为例,根据客户要求的等待时间数量级确定了服务器的合理配置。     

Queue Model Analysis for Spectrum-Sharing Cognitive Systems Under Outage Probability Constraint

This article investigates the behavior of the cognitive queueing system in a spectrum-sharing environment under the primary outage probability constraint. The queueing model is investigated under two cognitive transmit modes. A variable transmit rate that is equal to the cognitive channel’s capacity is first revisited, then the case of constant transmit rate is considered. Both transmit modes are found to lead to an $M/G/1$ queueing model. The performance measures of the cognitive queue under these two modes are investigated and numerically compared. The performance measures include the mean transmit rate, mean service time, server utilization, mean waiting and transit times of the packets in the queue, mean number of waiting and transit packets in the queue, mean duration of the server’s busy period, and mean number of packets served during the server’s busy period. The effects of changing the primary outage probability constraint and the primary transmit rate on the performance measures are investigated as well.     

DataWare在网络管理中的应用

NT服务器是网络中至关重要的设备，要求必须严格监控他以便为企业提供可靠、及时、不间断的服务。DataWare将两部NT服务器构成相互备份的服务器簇，当工作主机当机时，备份主机可迅速侦测出并自动探取接替(Takeover)，用户可在最短时间内恢复作业，避免服务器修机造成损失。     

Performance evaluation of a Batcher-banyan interconnection networkwith output pooling

A Batcher-banyan interconnection network that overcomes the limitation of switch throughput and unfair allocation of bandwidth is proposed, and its performance is evaluated. The throughput limitations of the interconnection network are substantially reduced by defining pools of outputs, in which groups of outputs that offer the same service behave each as a set of servers sharing a single waiting list. Through a careful system design and a suitable priority scheme, the servers in a set can be evenly allocated to the users requesting service from the inputs of the interconnection network. Three different solutions are proposed for the allocation of the output channels within a pool to those users requesting the same pool. Extensive computer simulation was used to evaluate the overall packet delay performance and the degree of fairness provided by each of these solutions     

Cost-based program allocation for distributed multimedia-on-demandsystems

Examining multimedia program allocation in distributed multimedia-on-demand systems lets us consider how best to provide programs at will from geographically scattered servers. Each user is associated with a local server but can transparently access any program located at any server. The authors consider user demand for various programs, storage limitations of the multimedia servers, and costs of storing and transporting the programs     

Collaboration IoT-Based RBAC with Trust Evaluation Algorithm Model for Massive IoT Integrated Application

With the recent advances in ubiquitous communications and the growing demand for low-power wireless technology, smart IoT device (SID) to access various Internet of Things (IoT) resources through Internet at any time and place alternately. There are some new requirements for integration IoT servers in which each one is individually gathering its local resources in Internet, which cooperatively supports SID to get some flexibility or temporary contract(s) and privileges in order to access their corresponding desired service(s) in a group of collaboration IoT servers. However, traditional access control schemes designed for a single server are not sufficient to handle such applications across multiple collaboration IoT servers to get rich services in IoT environments. It does not take into account both security and efficiency of IoT servers, which securely share their resources. Therefore, the collaboration IoT-based RBAC (Role-based Access Control) with trust evaluation (TE) algorithm model to reducing internal security threats in intra-server and inter server for the massive IoT integrated application is proposed in this paper. Finally, the three trust evaluation algorithms including a local trust evaluation algorithm, a virtual trust evaluation algorithm and a cooperative trust evaluation algorithm in the proposed collaboration IoT-based RBAC model, which are designed and presented for reducing internal security threats in collaborative IoT servers.

     

Queueing study of a 3-priority policy with distinct servicestrategies

A discrete-time, single server, 3-queue system is presented and analyzed. A distinct service strategy, namely the consistent-gated (c-G), 1-limited (L) and head-of-line (HoL), is applied to each of the queues (c-G/L/HoL policy). It is shown that this queueing system provides for an accurate analytical model for a DQDB station, as well as a means for an approximate evaluation of the correlation associated with key traffic processes in that network. In addition, the developed queueing system could be useful for the modeling of the queueing behavior of an ATM link shared by high-priority, low priority and control traffic. Through an asymptotic analysis under heavy low-priority traffic, the worst case performance for the high priority traffic is determined. Furthermore, it is illustrated that the asymptotic analysis provides for a potentially tight delay bounding technique. Finally, the delay performance of the developed queueing system is compared to that of a similar system in which one of the queues receives 1-limited service and the other two exhaustive (HoL^-/L/HoL⁺ policy)     

Optimal memory management strategies for a mobile user in abroadcast data delivery system

Data broadcasting has been considered as a promising way of disseminating information to a massive number of users in a wireless communication environment. In a broadcast data delivery system, there is a server which is broadcasting data to a user community. Due to the lack of communication from the users to the server, the server cannot know what a user needs. In order to access a certain item, a user has to wait until the item appears in the broadcast. The waiting time will be considerably long if the server's broadcast schedule does not match the user's access needs. If a user has a local memory, it can alleviate its access latency by selectively prefetching the items from the broadcast and storing them in the memory. A good memory management strategy can substantially reduce the user's access latency, which is a major concern in a broadcast data delivery system. An optimal memory management policy is identified that minimizes the expected aggregate latency. We present optimal memory update strategies with limited look ahead as implementable approximations of the optimal policy. Some interesting special cases are given for which the limited look-ahead policies are optimal. We also show that the same formulation can be used to find the optimal memory management policy which minimizes the number of deadline misses when users generate information requests which have to be satisfied within some given deadlines     

Performance of Hybrid Load Balancing Algorithm in Distributed Web Server System

The challenges that load balancing (LB) faces are that there may not always be enough physical servers to be able to handle the demand requested at any one time. If a huge influx of players comes online at the same time, there may just not be enough room for them all. In this case, the server would usually initiate a queue system where players are held before they are allowed to attempt to connect to the world server. If too many requests are made however, the load balancer itself may stop working and deny players access completely. These challenges are specific to the Client/Server model of game content distribution. The proposed hybrid Grid model has a slightly different set of challenges. The load balancer in this model has to constantly check the network for player disconnections so that any outstanding events can be cleaned up by the dedicated hosts. It also has to balance load like normal to these hosts as they will appear as one entity on the network but can be scaled to as many machines as are needed. The popularity of decentralization over recent years has led to issues with the demand on server hardware and the need for LB which generates increased costs for companies and affects the quality of service. The aim of this work was to design and construct a Hybrid load balancing solution for Grid and compare it to the other load balancing algorithms. Grid with a load balancer was written in Java using a modular construction which allowed code reuse between different network models. Four common algorithms, one of which was redeveloped to improve its performance, were compared to determine load balancing efficiency for selecting servers from a pool.