Showing papers presented at "International Workshop on Quality of Service in 2011"

DynaQoS: model-free self-tuning fuzzy control of virtualized resources for QoS provisioning

Abstract: Cloud elasticity allows dynamic resource provisioning in concert with actual application demands. Feedback control approaches have been applied with success to resource allocation in physical servers. However, cloud dynamics make the design of an accurate and stable resource controller more challenging, especially when response time is considered as the measured output. Response time is highly dependent on the characteristics of workload and sensitive to cloud dynamics. To address the challenges, we extend a self-tuning fuzzy control (STFC) approach, originally developed for response time assurance in web servers to resource allocation in virtualized environments. We introduce mechanisms for adaptive output amplification and flexible rule selection in the STFC approach for better adaptability and stability. Based on the STFC, we further design a two-layer QoS provisioning framework, DynaQoS, that supports adaptive multi-objective resource allocation and service differentiation. We implement a prototype of DynaQoS on a Xen-based cloud testbed. Experimental results on an E-Commerce benchmark show that STFC outperforms popular controllers such as Kalman filter, ARMA and adaptive PI by at least 16% and 37% under both static and dynamic workloads, respectively. Further results with multiple control objectives and service classes demonstrate the effectiveness of DynaQoS in performance-power control and service differentiation.

PERFUME: power and performance guarantee with fuzzy MIMO control in virtualized servers

Abstract: It is important but challenging to assure the performance of multi-tier Internet applications with the power consumption cap of virtualized server clusters mainly due to system complexity of shared infrastructure and dynamic and bursty nature of workloads. This paper presents PERFUME, a system that simultaneously guarantees power and performance targets with flexible tradeoffs while assuring control accuracy and system stability. Based on the proposed fuzzy MIMO control technique, it accurately controls both the throughput and percentile-based response time of multi-tier applications due to its novel fuzzy modeling that integrates strengths of fuzzy logic, MIMO control and artificial neural network. It is self-adaptive to highly dynamic and bursty workloads due to online learning of control model parameters using a computationally efficient weighted recursive least-squares method. We implement PERFUME in a testbed of virtualized blade servers hosting two multi-tier RUBiS applications. Experimental results demonstrate its control accuracy, system stability, flexibility in selecting tradeoffs between conflicting targets and robustness against highly dynamic variation and burstiness in workloads. It outperforms a representative utility based approach in providing guarantee of the system throughput, percentile-based response time and power budget in the face of highly dynamic and bursty workloads.

Adaptive data-driven service integrity attestation for multi-tenant cloud systems

Abstract: Cloud systems provide a cost-effective service hosting infrastructure for application service providers (ASPs). However, cloud systems are often shared by multiple tenants from different security domains, which makes them vulnerable to various malicious attacks. Moreover, cloud systems often host long-running applications such as massive data processing, which provides more opportunities for attackers to exploit the system vulnerability and perform strategic attacks. In this paper, we present AdapTest, a novel adaptive data-driven runtime service integrity attestation framework for multi-tenant cloud systems. AdapTest can significantly reduce attestation overhead and shorten detection delay by adaptively selecting attested nodes based on dynamically derived trust scores. Our scheme treats attested services as black-boxes and does not impose any special hardware or software requirements on the cloud system or ASPs. We have implemented AdapTest on top of the IBM System S stream processing system and tested it within a virtualized computing cluster. Our experimental results show that AdapTest can reduce attestation overhead by up to 60% and shorten the detection delay by up to 40% compared to previous approaches.

An application-level solution for the TCP-incast problem in data center networks

Abstract: Data centers have become very popular for storing large volumes of data. In particular, companies like Amazon, Google, and Yahoo! routinely use data centers for storage, Web search, and large-scale computations. The main characteristics of a data center network are high-speed links, low propagation delays, and limited-size switch buffers. In addition, the data for a given client application are usually striped (spread) over many servers, for increased reliability and performance (i.e., parallelism). Recent research efforts have resulted in several architectures of data centers [1].

Thwarting spam over internet telephony (SPIT) attacks on VoIP networks

Abstract: The threat of voice spam, commonly known as Spam over Internet Telephony (SPIT) is a real and contemporary problem. We present two approaches based on the anomaly detection of the distributions of selected call features (i.e., day and time of calling, call durations etc.) to detect and prevent SPITting over the Internet. The first approach uses Mahalanobis Distance as a summarization tool and it is able to reliably detect individual spam VoIP calls at a microscopic level. The second approach is designed to detect groups of (potentially collaborating) VoIP spam calls at a macroscopic level. By computing entropy of call durations of groups of calls, we are able to build profile of normal calls and reliably detect the deviation from normal human call behavior that are caused by bulk spam calls.

Multi-channel reliability and spectrum usage in real homes: empirical studies for home-area sensor networks

Abstract: Home area networks (HANs) consisting of wireless sensors have emerged as the enabling technology for important applications such as smart energy. These applications impose unique QoS constraints, requiring low data rates but high network reliability in the face of unpredictable wireless environments. This paper presents two in-depth empirical studies on wireless channels in real homes, providing key design guidelines for meeting the QoS constraints of HAN applications. The spectrum study analyzes spectrum usage in the 2.4 GHz band where HANs based on the IEEE 802.15.4 standard must coexist with existing wireless devices. We characterize the ambient wireless environment in six apartments through passive spectrum analysis across the entire 2.4 GHz band over seven days in each apartment. We find that the wireless conditions in these residential environments are much more complex and varied than in a typical office environment. Moreover, while 802.11 signals play a significant role in spectrum usage, there also exists non-negligible noise from non-802.11 devices. The multichannel link study measures the reliability of different 802.15.4 channels through active probing with motes in ten apartments. We find that there is not always a persistently reliable channel over 24 hours, and that link reliability does not exhibit cyclic behavior at daily or weekly timescales. Nevertheless, reliability can be maintained through infrequent channel hopping, suggesting dynamic channel hopping as a key tool for meeting the QoS requirements of HAN applications. Our empirical studies provide important guidelines and insights in designing HANs for residential environments.

Auto-scaling emergency call centres using cloud resources to handle disasters

Abstract: The fixed-line and mobile telephony network is one of the crucial elements of an emergency response to a disaster event. However, frequently the phone network is overwhelmed in such situations and is disrupted. It is not cost-effective to maintain an over-provisioned IT infrastructure for such rare events. Cloud computing allows users to create resources on-demand and can enable an IT infrastructure that scales in response to the demands of disaster management. In this paper, we introduce a system that uses the Amazon EC2 service to automatically scale up a software telephony network in response to a large volume of calls and scale down in normal times. We demonstrate the efficacy of this system through experiments based on real-world data.

Future internet video multicasting with essentially perfect resource utilization and QoS guarantees

Abstract: The multicasting of aggregated digital video over a proposed Future Internet backbone network with essentially perfect throughput, resource-utilization and QoS guarantees is summarized. The Future Internet routers require only minor modifications to the existing router designs. Buffers in existing internet routers are partitioned into 2 traffic classes which can co-exist, the Essentially-Perfect QoS class and the Best-Effort class, i.e., no new buffers are required. Each router includes an FPGA-based Scheduler Lookup Table for the essentially perfect QoS class. RSVP-TE is used to provision the multicast trees in an MPLS-TE network. Each router computes an essentially-perfect transmission schedule for all its QoS-enabled traffic flows, which never experience interference or congestion. (This integer-programming scheduling problem is a long-standing unsolved problem.) The Best-Effort traffic is scheduled using the usual Best-Effort schedulers. It is shown that thousands of bursty self-similar video streams can be multicast across the proposed Future Internet with essentially-perfect link efficiencies and QoS guarantees. The technology (i) can be added into new Internet routers with minimal cost (i.e., a few FPGAs); (ii) it allows for the co-existence of the Essentially-Perfect QoS and the usual Best-Effort traffic classes; (iii) it is compatible with the existing IEFT DiffServ and MPLS-TE service models; (iv) it allows for Internet link efficiencies as high as 100%, and (v) it can reduce Internet router buffer and power requirements significantly.

On traffic characteristics and user experience of Skype video call

Abstract: In this paper, we study Skype video call in LAN, WAN, and WiMAX. Our main interest is on traffic characteristics and user experience. Using the standard Foreman video sequence (320×240, 30fps, 8 second), we show that a Skype video call can adapt its source rate from < 5kBps up to ∼60kBps. The maximum achievable MOS (Mean Opinion Score) and EFR (Effective Frame Rate) is about 3.5 and 20 fps, respectively. Our experimental results also show that source rate is the dominating factor in determining both traffic characteristics and user experience of a Skype video call; while end-to-end delay or transport layer protocol, e.g. TCP or UDP, plays very little role. We also study the minimum RSSI requirement in a typical WiMAX network for achieving various levels of user experience: good, fair, and low, and derive the explicit logarithmic functions using the curve fitting technique to predict MOS and mean packet inter-arrival time (PIT) based on source rate.

Auto-scaling emergency call centres using cloud resources to handle disasters

Abstract: The fixed-line and mobile telephony network is one of the crucial elements of an emergency response to a disaster event. However, frequently the phone network is overwhelmed in such situations and is disrupted. It is not cost-effective to maintain an over-provisioned IT infrastructure for such rare events. Cloud computing allows users to create resources on-demand and can enable an IT infrastructure that scales in response to the demands of disaster management. In this paper, we introduce a system that uses the Amazon EC2 service to automatically scale up a software telephony network in response to a large volume of calls and scale down in normal times. We demonstrate the efficacy of this system through experiments based on real-world data.

Measuring and enhancing the social connectivity of UGC video systems: a case study of YouKu

Abstract: The social connections among users have significant impacts on UGC video systems The goal of this paper is to study the social connectivity of such systems by measuring YouKu, the most popular UGC video system in China We have collected 627 thousand user profiles, 3 million social connections and 136 million videos' information The analysis results have shown that the social connectivity is extremely weak and there are a considerable proportion of friend pairs sharing common semantic interests These facts motivate us to enhance the connectivity by recommending semantically relevant users as friends We thus propose a friend recommendation algorithm which locates potential friends quickly and accurately through the links to related videos, a unique feature of YouKu and similar sites We apply the algorithm on our dataset of YouKu and evaluate it through one-hop video search The social connectivity is greatly enhanced and the number of matched videos on friends is greatly increased To the best of our knowledge, this work is the first to identify the semantic relevance of friend pairs in UGC video systems and to study the friend recommendation

Leveraging statistical multiplexing gains in single- and multi-hop networks

Abstract: Packet switched networks achieve significant resource savings due to statistical multiplexing In this work we explore statistical multiplexing gains in single and multi-hop networks To this end, we analyze performance metrics such as delay bounds for a through flow comparing different results from the stochastic network calculus We distinguish different multiplexing gains that stem from independence assumptions between flows at a single hop as well as flows at consecutive hops of a network path Further, we show corresponding numerical results In addition to deriving the benefits of various statistical multiplexing models on performance bounds, we contribute insights into the scaling of end-to-end delay bounds in the number of hops n of a network path under statistical independence

Architecture and algorithms for virtual routers as a service

Abstract: The deployment of virtualized network resources has the potential to spur new business models and increase flexibility for network customers as well as infrastructure operators. It is worthwhile to re-evaluate how to effectively express traditional network elements in the virtualization domain. In this paper we consider network routers and argue that the representation of routing functionality as a service, rather than an isolated virtual resource is better suited in the virtualization context. We present an architecture enabling physical infrastructure operators to provide routing as a service. To this end, distributed forwarding elements are combined to appear a single virtual router instance which routes traffic between a set of customer points of presence. We provide embedding algorithms for virtual router topologies with minimum allocation cost. We consider the customer's geographical attachment to the network, bandwidth demands as well as capacity constraints in the core substrate. Moreover, we present a live-migration approach for the virtual router data plane which allows network operators to quickly adapt resources to changing network demands.

Applying stochastic network calculus to 802.11 backlog and delay analysis

Abstract: Stochastic network calculus provides an elegant way to characterize traffic and service processes. However, little effort has been made on applying it to multi-access communication systems such as 802.11. In this paper, we take the first step to apply it to the backlog and delay analysis of an 802.11 wireless local network. We found the derived bounds are very loose in comparison to ns-2 simulation results especially for heavy arriving traffic, indicating that improvements are needed for the current version of stochastic network calculus.

Minimizing energy cost for internet-scale datacenters with dynamic traffic

Abstract: In this paper, our goal is to achieve an optimal tradeoff between energy efficiency and service performance over a set of distributed IDCs with dynamic demand. In particular, we consider the outage probability as the QoS metric, where outage is defined as service demand exceeding the capacity of an IDC. Our goal is thus to minimize total energy cost over all IDCs, subject to the outage probability constraint. We achieve the goal by dynamically adjusting server capacity and performing load shifting in different time scales. We propose three different load-shifting and joint capacity allocation schemes with different complexity and performance. Our schemes leverage both stochastic multiplexing gain and electricity-price diversity.

QoS routing algorithms using fully polynomial time approximation scheme

Abstract: Routing with end-to-end Quality-of-Service (QoS) guarantees is a key to the widespread deployment of recent emerged services. Developing QoS routing algorithm in the network is an important open topic. This paper investigates the QoS routing related problems and proposes a Fully Polynomial Time Approximation Scheme (FPTAS) for QoS routing. In the proposed FPTAS, a graph-extending based dynamic programming approach is developed, and an extended version of the proposed algorithm is studied. The theoretical analyses show that the proposed algorithms outperform the previous best-known studies.

TCAM-based DFA deflation: a novel approach to fast and scalable regular expression matching

Abstract: Regular expression matching is the foundation of many network functions including intrusion detection, worm detection, traffic analysis and so on, where known patterns such as worm fingerprints are characterized using regular expressions and searched in network traffic for pattern match. As the quantity and diversity of known patterns keep increasing, regular expression pattern sets have rapidly grown in both size and complexity, while having to be matched in network traffic at accelerating wire speeds. Fast and scalable regular expression matching, therefore, is fundamental to the development of practical network systems.

An MDP-based admission control for a QoS-aware service-oriented system

Abstract: In this paper, we address the problem of providing a service broker, which offers to prospective users a composite service with a range of different Quality of Service (QoS) classes, with a forward-looking admission control policy based on Markov Decision Processes (MDPs).

On the performance of TCP over throughput-optimal CSMA

Abstract: An interesting distributed throughput-optimal CSMA MAC protocol, called adaptive CSMA, was proposed recently to schedule any strictly feasible rates inside the capacity region. Of particular interest is the fact that the adaptive CSMA can achieve a system utility arbitrarily close to that is achievable under a central scheduler. However, a specially designed transport-layer rate controller is needed for this result. An outstanding question is whether TCP Reno (one of the most mature versions of TCP) is compatible with adaptive CSMA and can achieve the same result. The answer to this question will determine how close to practical deployment adaptive CSMA is. Our answer is yes and no. First, we observe that running TCP Reno directly over adaptive CSMA results in severe starvation problems. Effectively, its performance is no better than that of TCP Reno over legacy CSMA (IEEE 802.11), and the potentials of adaptive CSMA cannot be realized. We then propose a multi-connection TCP solution with active queue management and prove that it can work with adaptive CSMA to achieve optimal utility. NS-2 simulations demonstrate that our solution can alleviate starvation and achieve fair and efficient rate allocation. We remark that multi-connection TCP can be implemented at either application or transport layer. Application-layer implementation requires no kernel modification, making the solution readily deployable in networks running adaptive CSMA. Our results show that adaptive CSMA can work well with only light-weight TCP modifications, bringing it a step closer to practicality.

Initial CWND determination method for fast startup TCP algorithms

Abstract: Since the TCP transmission rate is low during the startup phase, the QoS of various applications is seriously degraded. To improve the TCP transmission rate in the startup phase, many fast startup algorithms have been proposed. In this paper, we evaluate the fundamental performance of fast startup algorithms. Specifically, we focus on the initial window size and Rate-Based Pacing (RBP) start method, which is employed by many of fast startup algorithms. Through the simulation experiments, we first compare the performance of the RBP start method with that of the normal TCP. We then derive a way to set the initial window size for fast startup TCP algorithms.

Trajectory-assisted delay-bounded routing with moving receivers in vehicular ad-hoc networks

Abstract: Vehicular Ad-hoc Networks(VANETs) can facilitate many applications such as road safety, intelligent transportation and advertising. These applications usually call for multi-hop data delivery from access points to moving vehicles with user specified delay requirements. However, most existing routing protocols for VANETs only focus on message forwarding from vehicles to access points or take no account of the delay constraint. In this paper, we focus on the development of a carry-and-forward scheme that delivers data from access points to vehicles. Utilizing the vehicle's trajectory obtained from the navigation system, we propose TaDB, a Trajectory-assisted Delay-Bounded Message Delivery Algorithm. To choose delivery route within delay constraint while minimizing transmission cost, TaDB uses a Cluster-Aware Link Delay Model to estimate link delay for both the Carry and the Forward strategies on each road segment. TaDB also leverages the vehicle's planned trajectory to estimate its future location. Simulation results show that TaDB can achieve a delivery ratio very close to optimal.

Multi-tiered, burstiness-aware bandwidth estimation and scheduling for VBR video flows

Abstract: The increasing demand for high-quality streaming video delivered to mobile clients necessitates efficient bandwidth utilization and allocation at not only the wireless channel but also the wired backhaul of broadband wireless networks. In this context, we propose techniques for increasing the link utilization and enhancing the quality-of-experience (QoE) for end users while multiplexing video streams over a wired link. For increasing the link utilization, we present a generic multi-tiered bandwidth estimation and scheduling scheme that can guarantee lower bounds on loss for flows at lower tiers. This scheme can be used for supporting heterogeneous loss classes, differentiated losses for different layers of video streams, or per-flow guarantees using lower aggregate bandwidth than schemes proposed in the literature. For enhancing the end-user QoE, we present a scheme for minimizing correlated losses and improving the smoothness of video quality by minimizing the maximum loss suffered by any logical unit of a stream and also the variability in loss across the length of the stream. In simulations performed using MPEG-4 sources, our loss-minimization approach could lower the maximum loss by a factor of five and the loss variance by more than an order of magnitude. Our multi-tiered scheme could lower the estimated bandwidth and improve statistical multiplexing gains by 10-20% with three classes, 5--20% with two classes, and over 30% in the context of providing deterministic per-flow guarantees.

On long-term social relationships in peer-to-peer systems

Abstract: BitTorrent, the most popular file delivery system over the Internet, has attracted attention from network operators and researchers for its wide deployment. Recent studies suggest that long-term relationships among BT peers could be explored for peer cooperation, as to achieve better sharing efficiency. However, whether such long-term relationships exist remain unknown. In this paper, we for the first time examine the feasibility of social network based content delivery through the study of Twitter initialized/shared torrents. We show that the peers in such swarms have stronger temporal locality, thus offering great opportunity for improving their degree of sharing. Based on the Hadamard Transform of peers' online behaviors, we develop a social index to quickly locate peers of common patterns. Preliminary PlanetLab experiments indicate that the incorporation of social relations remarkably accelerates the downloading time. The improvement remains noticeable even in a hybrid system with a small set of socially active peers only.