Proportional Service Differentiation Research Articles

Multi-tier service differentiation by coordinated learning-based resource provisioning and admission control

Multiple Internet applications are often hosted in one datacenter, sharing underlying virtualized server resources. It is important to provide differentiated treatment to co-hosted applications and to improve overall system performance by efficient use of shared resources. Challenges arise due to multi-tier service architecture, virtualized server infrastructure, and highly dynamic and bursty workloads. We propose a coordinated admission control and adaptive resource provisioning approach for multi-tier service differentiation and performance improvement in a shared virtualized platform. We develop new model-independent reinforcement learning based techniques for virtual machine (VM) auto-configuration and session based admission control. Adaptive VM auto-configuration provides proportional service differentiation between co-located applications and improves application response time simultaneously. Admission control improves session throughput of the applications and minimizes resource wastage due to aborted sessions. A shared reward actualizes coordination between the two learning modules. For system agility and scalability, we integrate the reinforcement learning approach with cascade neural networks. We have implemented the integrated approach in a virtualized blade server system hosting RUBiS benchmark applications. Experimental results demonstrate that the new approach meets differentiation targets accurately and achieves performance improvement of applications at the same time. It reacts to dynamic and bursty workloads in an agile and scalable manner.

Implementing assuring forward per hop behaviour class in web servers with dynamic scheduling and selective discards

This paper proposes an approach for services differentiation in web servers that preserves the network end-to-end quality of service (QoS) mapping in conformance to per hop behaviour assuring forward group. This novel approach uses dynamic scheduling techniques and selective discard of requests to provide a proportional service differentiation to web clients. A prototype was built over the Apache web server and measurements were carried out aiming to evaluate the approach, contrasting it with other traditional techniques of services differentiation.

P-RWBO: a novel low-collision and QoS-supported MAC for wireless ad hoc networks

QoS supported MAC mechanism is a key issue for supporting QoS in wireless ad hoc networks. A new backoff algorithm, named RWBO+BEB, was proposed previously to decrease the packet collision probability significantly. In this paper, it is explored how to make RWBO+BEB support service differentiation in wireless ad hoc networks, and a novel proportional service differentiation algorithm, named p-RWBO, is proposed to allocate the wireless bandwidth according to the bandwidth ratio of each station. In p-RWBO, station n’s walking probability (pw,n) is selected according to its allocated bandwidth ratio. An analytical model is proposed to analyze how to choose pw,n according to the bandwidth ratios of station n. The simulation results indicate that p-RWBO can differentiate services in terms of both bandwidth and delay.

Resource allocation optimization for quantitative service differentiation on server clusters

The goal of service differentiation is to provide different service quality levels to meet changing system configuration and resource availability and to satisfy different requirements and expectations of applications and users. In this paper, we investigate the problem of quantitative service differentiation on cluster-based delay-sensitive servers. The goal is to support a system-wide service quality optimization with respect to resource allocation on a computer system while provisioning proportionality fairness to clients. We first propose and promote a square-root proportional differentiation model. Interestingly, both popular delay factors, queueing delay and slowdown, are reciprocally proportional to the allocated resource usage. We formulate the problem of quantitative service differentiation as a generalized resource allocation optimization towards the minimization of system delay, defined as the sum of weighted delay of client requests. We prove that the optimization-based resource allocation scheme essentially provides square-root proportional service differentiation to clients. We then study the problem of service differentiation provisioning from an important relative performance metric, slowdown. We give a closed-form expression of the expected slowdown of a popular heavy-tailed workload model with respect to resource allocation on a server cluster. We design a two-tier resource management framework, which integrates a dispatcher-based node partitioning scheme and a server-based adaptive process allocation scheme. We evaluate the resource allocation framework with different models via extensive simulations. Results show that the square-root proportional model provides service differentiation at a minimum cost of system delay. The two-tier resource allocation framework can provide fine-grained and predictable service differentiation on cluster-based servers.

Simultaneous and Proportional Bandwidth, Delay, and Loss Differentiation

A robust adaptive scheduler for proportional delay differentiation services is presented. Proportional services are further policed by a class-based packet dropper. Our proposed combination of the adaptive scheduler and the packet dropper treats the traffic classes proportionally in terms of three quality of service (QoS) metrics: bandwidth, delay, and packet loss, simultaneously. Traffic types used to study the performance of the scheme ranges from ordinary FTP data to bursty Pareto traffic. Numerical results validate our claim that regardless of the network traffic characteristics, especially burstiness, the adaptive delay scheduler combined with the packet dropper can effectively differentiate services in terms of delay, bandwidth, and loss simultaneously. In addition, our proposed scheme showed on an average 11% improvement in performance when compared to the scheme proposed in C. Dovrolis, D. Stiliadis, & P. Ramanathan, Proportional differentiated services: Delay differentiation and packet scheduling, IEEE/ACM Transactions on Networking, 10(1), 2002, 12–26.

Analysis of average burst-assembly delay and applications in proportional service differentiation

In Optical Burst-Switched (OBS) networks, the limitation of optical buffering devices make it impractical to deploy conventional delay-based differentiation algorithms such as Active Queue Management, Weighted Fair Queuing, etc. Furthermore, only the delay that appears due to the burst-assembly process constitutes a variable quantity (all the other sources of delay are mostly fixed), it is then reasonable to make use of the burst-assembly algorithm to provide class-based delay differentiation. The aim of the following study is twofold: first it defines an average assembly delay metric, which represents the assembly delay experienced by a random arrival at the burst assembler of an edge OBS node; and second, this metric is used to define and configure a two-class burst-assembly policy, which gives preference to high-priority traffic over low-priority packet arrivals. The results show that, (1) tuning the parameters of the two-class assembly algorithm, the two classes of traffic exhibit different burst-assembly delay; and, (2) such parameters can be adjusted to provide a given differentiation ratio in the light of the proportional QoS differentiation approach proposed in the literature. A detailed analysis of the two-class assembly algorithm is given, along with an exhaustive set of experiments and numerical examples that validate the equations derived.

Feedback-based offset time selection for end-to-end proportional QoS provisioning in WDM optical burst switching networks

Wavelength division multiplexing (WDM) optical burst switching (OBS) is a promising technology for the next generation Terabit networks. The capability of this technology to support service differentiation has become a key research issue due to the increasing need for traffic differentiation in Internet backbone networks. We develop a feedback-based offset time selection method to provide proportional service differentiation in terms of burst dropping probability. Compared to other relative Quality-of-Service (QoS) methods, proportional QoS allows a service provider to adjust the performance spacing of a class of traffic relative to others. Unlike other proportional QoS methods which guarantee only per-hop proportional QoS, our proposed method is able to achieve end-to-end proportional QoS. By making use of link states collected by probe packets, we dynamically adjust the offset times needed to achieve the predefined proportional QoS among different classes of traffic for various ingress–egress node pairs. As the offset time selection is done for the node pairs, fairness among node pairs with various hop lengths in terms of achieving the proportional QoS is maintained. The effectiveness of the proposed method is evaluated through simulation experiments.

Provisioning link layer proportional service differentiation in wireless networks with smart antennas

In this paper, we present a collision free MAC protocol for wireless networks with smart antennas that provides proportional service differentiation to various classes of traffic based on their respective bandwidth demand. The proposed protocol works for diverse physical parameters such as number of interfaces at each node, number of communication frequencies, and antenna beamwidth. To the best of our knowledge, this is the first work that provides link layer differentiated services for wireless networks with smart antennas and explores the influence of the physical parameters and network topology on the performance of the MAC layer.

Proportional Service Differentiation in Wireless LANs Using Spacing-based Channel Occupancy Regulation

In this paper, we propose Spacing-based Channel Occupancy Regulation (SCORE) MAC protocol for wireless LANs that provides proportional service differentiation in terms of normalized throughput. As shown by our system model and simulation study, SCORE provides consistent, scalable and adjustable proportional differentiation for any network size, any service class distribution, any node data rate and any packet size. Compared to state-of-the-art prioritized service differentiation schemes like Enhanced Distributed Coordination Function (EDCF), SCORE can quantitatively control the channel sharing between different service classes. Moreover, SCORE obtains significant performance improvements in terms of higher network throughput, higher transmission efficiency, lower medium access delay and lower delay jitter.

The QoSbox: Quantitative service differentiation in BSD routers

We describe the design and implementation of the QoSbox, a configurable IP router that provides per-hop service differentiation on loss, delays and throughput to classes of traffic. The novel aspects of the QoSbox are that (1) the QoSbox does not rely on any external component (e.g., no traffic shaping and no admission control) to provide the desired service differentiation, but instead, (2) dynamically adapts packet forwarding and dropping decisions as a function of the instantaneous traffic arrivals and allows for temporary relaxation of some service objectives; also, (3) the QoSbox can enforce both absolute and proportional service differentiation on queuing delays, loss rates, and throughput at the same time. We focus on a publicly available implementation of the QoSbox in BSD-based PC-routers. We evaluate our implementation in a testbed of BSD routers over a FastEthernet network, and we sketch how the QoSbox can be implemented in high speed architectures.

Computing the loss differentiation parameters of the proportional differentiation service model

The proportional differentiation service model has emerged as a refined version of the DiffServ qualify of service (QoS) architecture. It relies on a series of parameters to enforce proportionally differentiated QoS criteria, such as queueing delay and packet loss. From the perspective of proportional loss differentiation, a large amount of work has been done on carrying out loss differentiation based on given parameters. Under certain network conditions, however, the loss differentiation cannot be met based on these hand picked parameters. While existing work focuses on enhancing dropping mechanisms themselves to honour the loss differentiation, the paper looks into calculating feasible differentiation parameters. By forming an optimisation problem based on multiple class blocking thresholds, the paper introduces a simple quantitative guideline to compute loss differentiation parameters. Derived closely related to network statuses and packet dropping mechanisms, these parameters ease the difficulty that dropping mechanisms may encounter when enforcing packet loss differentiation. Its finite computation time, moreover, makes practical implementation possible. Analytical and numerical results are given to substantiate the new approach and its merits.

Admission control schemes for proportional differentiated services enabled internet servers using machine learning techniques

Abstract A widely existing problem in contemporary web servers is the unpredictability of response time. Owing to long response delay, revenues of the enterprises are substantially reduced due to many aborted e-commerce transactions. Recently, researchers have been addressing different admission control schemes of differentiated service for web servers to complement the Internet differentiated services model and thereby provide QoS support to the users of the World Wide Web. However, most of these admission control mechanisms do not guarantee the QoS requirements of all admitted clients under bursty workload. Although an Internet service model called proportional differentiated service is enabled in web servers to improve the QoS guarantee predicament in the literature, it still exists some impracticable assumptions and incompatible problems with the current Internet protocols. In this paper, we propose two algorithms for admission control and traffic scheduler schemes of the web server under proportional differentiated service, wherein a time series predictor is embedded to estimate the traffic load of the client in the next measurement time period. Support vector regression and particle swarm optimization techniques are used to implement the time series predictor based on the reports of successful prediction in the literature. The experimental results reveal that the proposed schemes can realize proportional delay differentiation service in multiclass Web server effectively. Meanwhile, the small computation overhead of particle swarm optimization verifies the feasibility of this machine learning technique in the real-time applications such as the admission control of the Internet server as illustrated in this work.

Robust Processing Rate Allocation for Proportional Slowdown Differentiation on Internet Servers

A desirable behavior of an Internet server is that a request's queuing delay depends on its service time in a linear fashion. Measuring the quality of service in terms of slowdown, the ratio of a request's queuing delay to its service time, provides a simple way to attain the objective. Moreover, it treats client requests equally regardless of their service time, whereas response time favors requests that need more processing resources. In this paper, we propose a proportional slowdown differentiation (PSD) service model on Internet servers. It aims to maintain prespecified slowdown ratios between different classes of client requests. To provide PSD services, we first derive a closed-form expression of the expected slowdown in an M/G/1 FCFS queuing system with a typical heavy-tailed service time distribution, the bounded Pareto distribution. Based on the closed-form expression, we design a queuing-theoretic strategy of processing-rate allocation. The rate allocation is realized by deploying a virtual server for each class. Simulation results show that the strategy can provide controllable PSD services on Internet servers. It, however, comes along with large variance and weak predictability due to the dynamics of Internet traffic. To address these issues, we design an integral feedback controller and integrate it into the queuing-theoretic strategy. Simulation results demonstrate that the integrated strategy is robust and can deliver predictable PSD services at a superior fine-grained level. We modified the Apache Web server with an implementation of the integrated processing-rate allocation strategy. Experimental results further demonstrate its effectiveness and feasibility in practice.

Achieving proportional delay differentiation in wireless LAN via cross‐layer scheduling

AbstractProviding service differentiation in wireless networks has attracted much attention in recent research. Existing studies so far have focused on the design of differentiated media access algorithms. Some quality of service (QoS) metrics, such as queueing delay cannot be completely addressed by these approaches. Moreover, without a formalized service differentiation goal that quantifies the outcome of differentiation, the performance of most of the approaches fluctuates, especially in short time‐scales.This paper addresses above problems by introducing the concept of proportional service differentiation, to the domain of wireless network and focuses on providing proportional delay differentiation in wireless LANs. Due to the unique characteristic of distributed medium sharing, the scheduling algorithm employed in wireline networks can not be applied directly to the context of wireless LANs. We argue that delay differentiation in wireless LAN can only be achieved through a joint packet scheduling at the network layer and distributed coordination at the media access control (MAC) layer. Therefore, we present a cross‐layer waiting time priority (CWTP) scheduling algorithm. CWTP consists of two tiers: an intra‐node waiting time priority (WTP) scheduler at the network layer and an inter‐node distributed coordination function at the MAC layer. These two tiers coordinate via a mapping function, which maps the normalized waiting time at the network layer to the backoff time at the MAC layer. Two mapping schemes, namely linear mapping and piecewise linear mapping, are presented and evaluated in this paper. Extensive simulation results show that the CWTP algorithm can effectively achieve proportional delay differentiation in wireless LANs. Copyright © 2004 John Wiley & Sons, Ltd.

Proportional Differentiated Admission Control

This letter presents a new admission control policy inspired in the framework of proportional differentiated services (PDS). While most of previous PDS has focused on average queueing delays and packet drops to differentiate the performance of adaptive applications, the proportional differentiation admission control (PDAC) differentiates inelastic traffic in terms of blocking probabilities. The PDAC is built up using asymptotic approximation theory, employs a class based approach, and conforms with the PDS requirements of predictability and controllability . Numerical experiments confirm a good performance of the approach.

Utility-based scheduling disciplines for adaptive applications over the Internet

This letter presents packet scheduling disciplines based on application utility functions and network traffic measurements. The disciplines support different classes of adaptive applications over the Internet, providing differentiation, fairness, and dynamic allocation of network resources. They are composed of a decision procedure, where a fairness criterion based on utility functions is used; and a measurement procedure, which considers the statistics involving packet arrivals and departures. The underlying algorithm is then applied to emulate the proportional differentiation services, and is shown-via simulation-that its results outperform the best alternative algorithms published in the literature.