Heterogeneity prevails not only among physical machines but also among workloads in real IaaS Cloud data centers (CDCs). The heterogeneity makes performance modeling of large and complex IaaS CDCs even more challenging. This paper considers the scenario where the number of virtual CPUs requested by each customer job may be different. We propose a hierarchical stochastic modeling approach applicable to IaaS CDC performance analysis under such a heterogeneous workload. Numerical results obtained from the proposed analytic model are verified through discrete-event simulations under various system parameter settings.

Today's data center networks are usually over-provisioned for peak workloads. This leads to a great waste of energy since in practice traffic rarely ever hits peak capacity resulting in the links being under-utilized most of the time. Furthermore, the traditional non-traffic-aware routing mechanisms worsen the situation. From the perspective of resource allocation and routing, this paper aims to implement a green data center network and save as much energy as possible. With the benefit of blocking island paradigm, we present a general framework trying to maximize the network power conservation and minimize sacrifices of network performance and reliability. The bandwidth allocation mechanism together with power-aware routing algorithm achieve a bandwidth guaranteed green tighter network. Moreover, our fast efficient heuristics for allocating bandwidth enable the system to scale to large sized data centers. The evaluation result shows that achieving up to more than 50 percent power savings are feasible while guaranteeing network performance and reliability.

The performance, dependability, and security of cloud service systems are vital for the ongoing operation, control, and support. Thus, controlled improvement in service requires a comprehensive analysis and systematic identification of the fundamental underlying constituents of cloud using a rigorous discipline. In this paper, we introduce a framework which helps identifying areas for potential cloud service enhancements. A cloud service cannot be completed if there is a failure in any of its underlying resources. In addition, resources are kept offline for scheduled maintenance. We use redundant resources to mitigate the impact of failures/maintenance for ensuring performance and dependability; which helps enhancing security as well. For example, at least 4 replicas are required to defend the intrusion of a single instance or a single malicious attack/fault as defined by Byzantine Fault Tolerance (BFT). Data centers with high performance, dependability, and security are outsourced to the cloud computing environment with greater flexibility of cost of owing the computing infrastructure. In this paper, we analyze the effectiveness of redundant resource usage in terms of dependability metric and cost of service deployment based on the priority of service requests. The trade-off among dependability, cost, and security under different redundancy schemes are characterized through the comprehensive analytical models.

Virtual machines (VMs) are used in cloud computing systems to handle user requests for service. Failure of VMs cause that the user's request not being completed. Replication mechanisms can be used to mitigate the impact of VM failures. In this paper, we are primarily interested in characterizing the failure-recovery behavior of a VM in cloud with different replication schemes. We use a service-oriented dependability metric called Defects Per Million (DPM) defined as the number of user requests dropped out of a million, due to VM failures. We present an analytical modeling approach for computing the DPM metric in different replication schemes on the basis of structure-state process and checkpointing method. The effectiveness of replication schemes are demonstrated through experimental results. To verify the validity of the proposed analytical models, we extend the widely used cloud simulator CloudSim and compare the simulation results with analytical solutions.

This paper develops a novel approximate analytical model to evaluate the performance of active virtual machines in IaaS clouds using an M/G/m/m+K queue. The proposed model, combined with the transform-based analytical approach, enables the computation of the probability distribution of the number of jobs in the system and subsequently a set of performance measures, including the mean number of jobs in the system, the mean response time, the probability of immediate service, and the blocking probability. Compared to the existing Markov models of cloud data centers, our approach can reflect the system behavior more accurately even when the service-time distribution has a large coefficient of variation (> 1.5) in a medium-sized IaaS cloud. Numerical results obtained from the proposed analytical model are verified through extensive simulations under various system parameter settings, and compared with the results from existing models.

Traditional system-oriented dependability metrics like reliability and availability do not fully reflect the impact of system failure-repair behavior in service-oriented environments. The telecommunication systems community prefers to use Defects Per Million (DPM), defined as the number of calls dropped out of a million calls due to failures, as a user-perceived dependability metric. In this paper, we provide new formulation for the computation of the DPM metric for a system supporting Voice over IP functionality using the Session Initiation Protocol (SIP). We evaluate different replication schemes that can be used at the SIP application server. They include the effects of software failure, failure detection, recovery mechanisms, and imperfect coverage for recovery mechanisms. We derive closed-form expressions for the DPM taking into account the transient behavior of recovery after a failure. Our approach and underlying models can be readily extended to other types of service-oriented environments.

In cloud computing systems, a user request goes through several cloud service provider specific processing steps from the instant it is submitted until the service is completed. In this paper, we use service-oriented metrics to characterize the dependability of cloud computing systems in order to find the pitfalls and improve the service. We find that it is not possible to fully reflect the impact of a cloud-service's dependability behavior through traditional dependability metrics like availability or reliability. We use a user-perceived dependability metric called Defects Per Million (DPM), defined as the number of user requests dropped out of a million. We demonstrate a new formulation for computing DPM metric in cloud computing systems. We incorporate check pointing scheme for job execution in the cloud to mitigate the impact of virtual machine failures, and compute DPM in order to characterize the improvement in the DPM due to the check pointing scheme compared to no-check pointing scheme.

From the perspective of resource allocation and routing, this paper aims to save as much energy as possible in data center networks. We present a general framework, based on the blocking island paradigm, to try to maximize the network power conservation and minimize sacrifices of network performance and reliability. The bandwidth allocation mechanism together with power-aware routing algorithm achieve a bandwidth guaranteed tighter network. Besides, our fast efficient heuristics for allocating bandwidth enable the system to scale to large sized data centers. The evaluation result shows that up to more than 50% power savings are feasible while guaranteeing network performance and reliability.

This paper presents the design, implementation and evaluation of SprintNet, a novel network architecture for data centers. SprintNet achieves high performance in network capacity, fault tolerance, and network latency. SprintNet is also a scalable, yet low-diameter network architecture where the maximum shortest distance between any pair of servers can be limited by no more than four and is independent of the number of layers. The specially designed routing schemes for SprintNet strengthen its merits. Both theoretical analysis and simulation experiments are conducted to evaluate its overall performance with respect to average path length, aggregate bottleneck throughput, and fault tolerance.

Virtual machines (VM) are used in cloud computing systems to handle user requests for service. A typical user request goes through several cloud service provider specific processing steps from the instant it is submitted until the service is completed. In the process of providing the service, VM failures cause the user's request to be dropped. To mitigate the adverse impact of VM failure, replication mechanisms, either using cold, warm or hot replication, can be used. In this paper, we model the system behavior with a structure-state process to characterize the failure-recovery behavior of a VM in a cloud that uses one of the aforementioned replication schemes. We use a service-oriented dependability metric called Defects Per Million (DPM), defined as the number of user requests dropped out of a million. The structure-state process approach is used to analyze the job completion time distribution and subsequently we compute the DPM by counting the number of requests exceed the specified deadline. The effectiveness of replication schemes are demonstrated through numerical results.

Data center networks (DCNs) are inherently failure-prone owing to the existence of many links, switches and servers. Many times the failures of the components may be correlated resulting a set of connected components failing together. This correlated failure behaviour could be captured through the use of fault regions [1]. This paper explores the effect of such failures in DCNs, using four topologies, viz., Fat Tree, DCell, FlatNet and BCube. We used two categories of metrics for evaluation: connection-oriented metrics, including aggregated bottleneck throughput (ABT), average path length (APL) and routing failure rate (RFR); and network size-oriented metrics, including Component Decomposition Number (CDN) and Smallest/Largest Component Size (SCS/LCS).

The Third International Workshop on Dependability of Clouds, Data Centers, and Virtual Machine Technology (DCDV 2013) features papers covering various aspects of dependability and security in Clouds and Data Centers. Four sessions covering Cloud and Data Center Networking, Dependability Evaluation, Mobile and Cloud Computing, and Virtualization and Cloud include eleven papers.

Integrating system-level virtualization technology with Trusted Computing technology can significantly improve system security. The open-source virtual TPM facility, shipped with the open-source hyper visor Xen, aims to provide the illusion of a physical TPM to TPM-based trusted software executed in trusted virtual machines (TVMs) such that TPM-based trusted software works well in a TVM as in a native Operating System (OS). However, it is not true for the trusted software which works in a trusted para-virtual machine (TPVM). The TPM command packets sent in the TPVM OS boot phase may cause the TPVM OS boot failure. This paper design and implement E2VT, an effective and efficient mechanism for preventing the TPVM OS boot failure while (1) being transparent to the TPVM system, (2) maintaining the original system performance, (3) making minimal modifications to the existing architecture. We validate our analysis and design through experiments.

The design of the data center network that interconnects the massive number of servers is very important to ensure the agility and robustness of the data center to meet the requirements of the applications. In response to this challenge, the research community have begun exploring novel interconnect topologies including FatTree, DCell, BCube, HyPaC, etc. However, the solutions proposed so far either scale too slowly, suffer from performance bottlenecks, are server-location dependent, inherit poor availability, or can be too complex/expensive to construct. Motivated by these very important challenges, we propose a new data center interconnect called FlatNet that combines the advantages of previous architectures while avoiding their limitations. FlatNet is a cost-effective, high-performance, reliable and scalable interconnect with almost flat architecture. For example, given an equal-sized data center, the costs of a FlatNet in terms of number of links and switches are roughly 2/3 and 2/5 that of Portland, while still delivering comparable overall performance.

Mega data centers are being built around the world to provide various cloud computing services. As a result, data center networking has recently been a hot research topic in both academia and industry. A fundamental challenge in this research is the design of the data center network that interconnects the massive number of servers, and provides efficient and fault-tolerant routing service to upper-layer applications. In response to this challenge, the research community have begun exploring novel interconnect topologies including Fat-Tree, DCell, and BCube and etc. Understandably, this research is still in its infancy. The proposed solutions either scale too fast (i.e., double exponentially) or too slow, suffer from performance bottlenecks, or can be quite costly in both routing and construction. In this paper, we propose a cost-effective and gracefully scalable data center interconnect termed Hyper-BCube that combines the advantages of both DCell and BCube architectures while avoiding their limitations. We then propose fault-tolerant and routing mechanisms for Hyper-BCube. Finally, we propose a comprehensive benchmarking environment that can be used for accurately and practically evaluating and testing the proposed data center architecture.

High-speed routers rely on well-designed packet buffers that support multiple queues, provide large capacity and short response times. Some researchers suggested combined SRAM/DRAM hierarchical buffer architectures to meet these challenges. However, these architectures suffer from either large SRAM requirement or high time-complexity in the memory management. In this paper, we present scalable, efficient, and novel distributed packet buffer architecture. Two fundamental issues need to be addressed to make this architecture feasible: 1) how to minimize the overhead of an individual packet buffer; and 2) how to design scalable packet buffers using independent buffer subsystems. We address these issues by first designing an efficient compact buffer that reduces the SRAM size requirement by (k-1)/k. Then, we introduce a feasible way of coordinating multiple subsystems with a load-balancing algorithm that maximizes the overall system performance. Both theoretical analysis and experimental results demonstrate that our load-balancing algorithm and the distributed packet buffer architecture can easily scale to meet the buffering needs of high bandwidth links and satisfy the requirements of scale and support for multiple queues.

Cloud computing can be characterized as the culmination of the integration of computing and data infrastructures to provide a scalable, agile and cost-effective approach to support the ever-growing critical IT needs (in terms of computation and storage) of both enterprises and the general public. Cloud computing introduces a paradigm shift in computing where the ownership of computing resources is no more necessary for businesses and individuals to provide services to their end-users over the Internet. Cloud computing relieves its users from the burdens of provisioning and managing their own data centers and allows them to pay for resources only when they are actually needed and used. However, a shared cloud infrastructure introduces a number of new dependability challenges both for the cloud providers and users. Indeed all the data gets created, stored, shared and manipulated within the cloud.

High-speed routers rely on well-designed packet buffers that support multiple queues, large capacity and short response times. Some researchers suggested combined SRAM/DRAM hierarchical buffer architectures to meet these challenges. However, these architectures suffer from either large SRAM requirement or high time-complexity in the memory management. Our analysis indicates that they perform exactly the same in the worst case. In this paper, we present a novel packet buffer architecture which reduces the SRAM size requirement by (k-1)/2k, where k denotes the number of DRAMs working in parallel. We use a fast batch load scheme and per-queue Random Round Robin memory management algorithm. Our mathematical analysis and simulation results indicate that the proposed architecture provides guaranteed performance in terms of low time complexity, short access delay and upper-bounded drop rate, when a little speedup is provided.

High-speed routers rely on well-designed packet buffers that support multiple queuing, large capacity and short response times. Some researchers suggested a combined SRAM/DRAM hierarchical buffer architecture to meet these challenges. However, both the SRAM and DRAM need to maintain a large number of dynamic queues which is a real challenge in practice and limits the scalability of these approaches. In this paper, we present a scalable, efficient and novel distributed packet buffer architecture. Two fundamental issues need to be addressed to make this feasible: (a) how to design scalable packet buffers using independent buffer subsystems; and (b) how to dynamically balance the workload among multiple buffer subsystems without any blocking. We address these issues by first designing a basic framework that allows flows to dynamically switch from one subsystem to another without any blocking. Based on this framework, we further devise a load-balancing algorithm to meet the overall system requirements. Both theoretical analysis and experimental results demonstrate that our load-balancing algorithm and the distributed packet buffer architecture can easily scale to meet the buffering needs of high bandwidth links with large number of active connections.

Free-riding is one of the most serious problems encountered in Peer-to-peer (P2P) systems like Bit Torrent. Incentive mechanisms, including those based on reputation have been proposed to deal with this problem, but are still not effective in preventing free-riders from completing a download. This is because they discover the free-riders' behavior during or after the process of trading, giving free-riders the opportunity to download from others. In this paper, we propose PreDiscover, a novel approach to prevent free-riding behavior in Bit Torrent. In PreDiscover, regular peers and free-riders can be recognized before trading. So free-riders have little opportunity to download blocks from others. Our simulation results indicate that this new mechanism is very effective in discouraging free-riders and foster fairness.

In this paper, we explore the problem of scheduling multiple concurrent Bag-of-Tasks (BoT) applications in unstructured peer-to-peer (P2P) volunteer desktop grids built around a super-peer framework. Resource availability information is exchanged among the super-peers using a light-weight threshold-driven gossip protocol with the aim of minimizing the resource discovery overhead. We present a comprehensive study of the properties of fixed threshold and adaptive threshold based gossiping mechanisms using detailed simulation experiments. Our results indicate that while gossiping with fixed threshold reflects resource usage effectively in a scalable network, an adaptive gossip threshold mechanism based on a simple feedback mechanism considering system state adapts much better to a volatile environment with small latency, reduced overhead and yields better system performance compared to fixed threshold protocols.

The current TCG architecture suffers from the Time-of-Check-To-Time-of-Use (TOCTOU) attacks in commodity PC operating systems (OS), in which kernel rootkits can get unrestricted access to OS resources. VMM-based approaches running at a privilege level higher than that of virtual machine (VM) kernel can effectively detect dynamic or static data attacks occurring in VMs. This paper proposes a lightweight response mechanism (LWRM) for TCG TOCTOU attacks occurring in VMs. LWRM has the following features: (1) compared to the existing response mechanism, LWRM is more effective in defeating the TCG TOCTOU attacks; (2) LWRM imposes less overhead on the system during normal execution; (3) LWRM is transparent to the kernel rootkits; and (4) LWRM can work in the scenarios with more than one run-time trusted virtual machine. We describe the design idea and the implementation by using the Xen virtual machine monitor (VMM) and the virtual TPM facility shipped with the Xen.

This paper addresses the design of high-performance buffers for high-end Internet routers. The buffers are typically implemented using a combination of SRAM and DRAM technologies in order to simultaneously meet the routers' high speed and capacity requirements. The major challenge in designing router buffers is to maintain multiple flow queues in the memory, unlike computer memory buffers (i.e., memory system). The major objective is to minimize the use of expensive but fast SRAM while providing acceptable delay guarantees to packets. In this paper, we first investigate hybrid SRAM/DRAM solutions proposed in the past. We show that one of the architectural limitations of these solutions is that the required SRAM size grows linearly with the number of flows in the system. This prevents the solutions from scaling to support a large number of flows. We then break down this shortcoming by proposing a parallel hybrid SRAM/DRAM (PHSD) architecture. We design a series of memory management algorithms (MMAs) for PHSD, based on tradeoffs between the complexity of the MMAs and the guarantee of in-order delivery of packets (segmentations). We perform a detailed analysis of the proposed algorithms and conduct extensive simulations to show that PHSD can significantly outperform solutions proposed in the past in terms of the SRAM requirements and packet delay.

Interrupt coalescing (IC) technique has been used in general-purpose operating systems to mitigate receive livelock (RL) problem in gigabit Ethernet network hosts. Schemes for dynamically tuning the interrupt coalescing behavior of a communication interface based on traffic load or system state have been proposed. However, all the existing IC schemes are designed using heuristics. In this paper we present an analytical model for the IC technique and carry out a detailed study of existing IC schemes in terms of their performance characteristics including system goodput, CPU consumption and latency. We validate our analysis through measurement-based experiments.

Peer-to-peer networks have long being used for file sharing among the connected peers. Recently, the popular BitTorrent tool has permitted the exchange of large sized files by dividing it into fragments, and enabling the peers to exchange the fragments amongst themselves through an overlay network. Each file requires the establishment of a separate overlay among the peers. In this paper, we explore the use of proximity both in the construction of the overlay network, and the efficient exchange of the file fragments, mainly aimed at reducing the download time for the peers and reducing resource usage (esp. link bandwidth) in the underlying network. We give some analytical and simulation results to show the improvement that can be achieved using proximity.

Recent trends in embedded systems indicate a growing importance of the software component of these systems, and the concomitant increase in the importance of formal design and development of embedded software. In this paper we share our experience with designing and offering courses related to embedded software in the Department of Computer Science and Engineering at the Hong Kong University of Science and Technology. We give a detailed overview of the courses, the structure and the teaching/learning methodology followed in our courses, followed by some discussion and reflections on the courses.

Presents the introductory welcome message from the conference proceedings. May include the conference officers' congratulations to all involved with the conference event and publication of the proceedings record.

Gigabit Ethernet is now finding wider deployment in computer networks. The conventional operating system suffers from the receive livelock problem in Gigabit Ethernet networks. The device hybrid (interrupt + polling) scheme has been widely used to overcome this problem in current operating systems such as GNU/Linux and FreeBSD. However, controlling the polling time without regard to the system state can degrade the ability of a hybrid scheme in some situations. This paper focuses on the system performance of the operating systems that employ the device hybrid scheme in kernel space. A queue-based adaptive polling (QAPolling) scheme is introduced that: (1) significantly improves system goodput and reduces packet loss over a wide range of computer hardware configurations and traffic conditions, (2) is scalable and easily deployed. The key idea behind QAPolling is to adjust the polling time adaptively according to the information of the application receiving queues, which are in kernel space and change with the system state, instead of the packet arrival rate. We validate our design through experimental results in Gigabit Ethernet networks.

Studies of the performance of interrupt-driven operating systems in high-speed networks have brought forth the problem of receive livelock. Device hybrid interrupt-polling and interrupt coalescing are two common techniques used in general-purpose operating systems to mitigate this problem. Adaptive schemes based on local knowledge have been proposed for each technique above. However, all the schemes proposed so far are designed using heuristics. In addition, the capabilities of the proposed schemes have not been systematically compared. In this paper, we first analyze the capabilities of these schemes by investigating the relationship between key system parameters and system goodput in different packet protocol processing modes under heavy traffic load. Then we propose a robust device hybrid interrupt-polling (RHIP) scheme which achieves high system goodput, low packet loss and good latency with low consumption of CPU cycles, compared to other schemes. The key idea of RHIP is to use the recipient's buffer information to adjust the interrupt rate and the protocol processing time. We validate our analysis and design through several experiments.

Abstract The Knowledge Grid built on top of the peer‐to‐peer (P2P) network has been studied to implement scalable, available and sematic‐based querying. In order to improve the efficiency and scalability of querying, this paper studies the problem of multicasting queries in the Knowledge Grid. An m‐dimensional irregular mesh is a popular overlay topology of P2P networks. We present a set of novel distributed algorithms on top of an m‐dimensional irregular mesh overlay for the short delay and low network resource consumption end‐host multicast services. Our end‐host multicast fully utilizes the advantages of an m‐dimensional mesh to construct a two‐layer architecture. Compared to previous approaches, the novelty and contribution here are: (1) cluster formation that partitions the group members into clusters in the lower layer where cluster consists of a small number of members; (2) cluster core selection that searches a core with the minimum sum of overlay hops to all other cluster members for each cluster; (3) weighted shortest path tree construction that guarantees the minimum number of shortest paths to be occupied by the multicast traffic; (4) distributed multicast routing that directs the multicast messages to be efficiently distributed along the two‐layer multicast architecture in parallel, without a global control; the routing scheme enables the packets to be transmitted to the remote end hosts within short delays through some common shortest paths; and (5) multicast path maintenance that restores the normal communication once the membership alteration appears. Simulation results show that our end‐host multicast can distributively achieve a shorter delay and lower network resource consumption multicast services as compared with some well‐known end‐host multicast systems. Copyright © 2006 John Wiley & Sons, Ltd.

Achieving fair bandwidth distribution among uplink and downlink flows in the infrastructure based wireless local area networks (WLAN) which the distributed coordination function (DCF) mode is difficult. In this paper we present a new control theoretic approach to achieve a fair bandwidth distribution among the flows regardless of the transport protocol used by the flows. In addition, we explore methods to improve the channel bandwidth utilization and reduce the delay while improving the fair distribution. Our approach combines an AQM scheme for IFQ queue and the MAC layer design to achieve this goal. The effectiveness of our approach is demonstrated through extensive simulations over a wide range of network scenarios.

Recent research studies in over-provisioned networks have shown that the assured forwarding (AF) service in the current differentiated services (DiffServ) architecture fails to provide bandwidth assurance in some situations. The paper focuses on the situation where adaptive and non-adaptive traffic coexist in the same real queue at routers and the buffer management scheme treats the traffic of the same priority in the same AF class indiscriminately. An enhanced RIO is introduced, which, without excessively penalizing non-adaptive flows, can: (1) significantly improve bandwidth assurance of adaptive AF flows; (2) alleviate the starvation imposed on adaptive best-effort flows. These goals are achieved by doing the following when the failure of bandwidth assurance is detected: (1) mapping adaptive OUT traffic and non-adaptive OUT traffic to different virtual queues; (2) adapting the queue length thresholds according to whether bandwidth assurance is achieved. We validate our design through simulations.

Virtual private networks (VPN) provide a cost-effective means of meeting the communication needs among several sites. The hose model for VPN configuration alleviates the scalability problem of the pipe model by reserving bandwidth for traffic aggregates instead of between every pair of endpoints. Existing studies on quality of service (QoS) guarantees in the hose model deal only with bandwidth requirements. In this paper we enhance the hose model to specify delay requirements between endpoints. Three categories of algorithms, namely the pipe mesh, the multiple source-based trees, and the shared tree approaches, are then proposed for VPN provisioning. We investigate methods of implementing the shared tree approach to meet the delay requirements with low provisioning cost and small computation overhead.

Several active queue management (AQM) mechanisms have been proposed in the literature to provide better support for congestion. However, their performance is examined mainly in best-effort networks. In this paper, we present an empirical study of the effects of these AQM mechanisms on the performance of Assured Forwarding (AF) Services in the Differentiated Services (DiffServ) framework. In addition, we investigate the interaction of AQMs with intelligent traffic conditioners. The major conclusions from our study are: (1) When the standard traffic conditioner is used, the transient response and the ability of controlling queue length of an AQM scheme not only affect the achievement of bandwidth assurance, but also affect the attainment of excessive bandwidth and link throughput. (2) The behaviors of an AQM affect the ability of intelligent traffic conditioners. (3) Rate-based AQM schemes perform better than queue-based AQM schemes; self-tuning AQM schemes perform better than AQMs with fixed-gains in terms of the above performance metrics.

In this paper, we consider two fairness problems (downlink/uplink fairness and fairness among flows in the same direction) that arise in the infrastructure WLAN. We propose a virtual queue management approach, named VQ-RED to address the fairness problems. We demonstrate the effectiveness of our approach by conducting a series of simulations. The results show that compared with standard DCF, VQRED not only greatly improves the fairness, but also reduces packet delays.

Research studies have uncovered the unfair WLAN bandwidth distribution between uplink and downlink TCP flows in the infrastructure WLAN using the distributed coordination function mode. Different mechanisms at or above the MAC layer have been proposed for handling this problem. This paper presents a simple but effective approach in order to achieve the fair distribution by using an adaptive queue management mechanism to manage the queue between the logic link and MAC layers at the access point. Extensive simulation results show that this approach outperforms other mechanisms discussed in the literature over a wide range of network scenarios in terms of fair distribution, small delay, and high WLAN bandwidth utilization

Significant research effort has been devoted to designing robust active queue management (AQM) mechanisms for the Internet, with mixed results. In this paper, we apply sliding mode control theory plus integral control to design a robust controller for AQM, which stabilizes the instantaneous queue length around the desired value with fast transient response over a wide range of network dynamics. We validate our design through extensive simulations and compare its performance with some AQM mechanisms published in the literature.

In this paper a simple robust proportional-integral (R-PI) controller is proposed for active queue management (AQM). We assume that TCP/AQM dynamics can be described by the linearized TCP/AQM model (C. V. Hollot et al., April 2001). R-PI aims to address the tradeoff between responsiveness and stability and the tradeoff between responsiveness and high link utilization over a large range of structured and unstructured uncertainties. This controller achieves these goals by varying its control parameters according to the system state. We show that the closed-loop system is asymptotically stable as long as the control parameters are time-invariant and varying in a range. Extensive simulation results demonstrate the robust ability of R-PI compared with some other AQM mechanisms in the literature.

The assured forwarding (AF) based service in a differentiated services (Diffserv) network fails to provide bandwidth assurance to aggregates in some circumstances. Several intelligent marking mechanisms have been proposed in the literature to improve bandwidth assurance for aggregates using the knowledge gathered at the ingress nodes. In this paper, we apply a control theoretic approach to this problem. We design a nonlinear proportional-integral (NPI) controller, called NPI-ACT, for adapting the CIR threshold. Performance results using extensive simulations demonstrate significant improvement using NPI-ACT in achieving bandwidth assurance over a wide range of network conditions compared to earlier mechanisms proposed in the literature.

Recent research studies have shown that assured forwarding (AF) service in the current differentiated services (Diffserv) framework does not provide bandwidth assurance in some circumstances. This paper proposes an adaptive marking threshold mechanism, called adaptive CIR threshold (ACT), which aims to improve bandwidth assurance for aggregate flows sharing the same AF class only based on local knowledge. Extensive simulation results demonstrate significant improvement with ACT in bandwidth assurance under various conditions: different round trip times (RTT), different numbers of micro-flows in an aggregate, different target rates, different packet sizes, and the presence of non-adaptive flows, compared with earlier mechanisms proposed in the literature.

Recent research studies have shown that assured forwarding (AF) service in the current differentiated services (Diffserv) framework does not provide bandwidth assurance in some circumstances. This paper proposes a mechanism, called Adaptive CIR+PIR Threshold (ACPT), which improves bandwidth assurance and domain throughput simultaneously. Extensive simulation results demonstrate significant improvement with ACPT in bandwidth assurance and domain throughput under various conditions: different round trip times (RTT), different numbers of micro-flows in an aggregate, different target rates, different packet sizes, and the presence of nonadaptive flows, compared to other mechanisms proposed in the literature.

In this paper we present a generalized scheme for quality of service routing with both delay and bandwidth constraints using a delay weighted flow approach. The objective is to minimize call blocking probability. A request is routed along the path capable of supporting the required quality of service and which most preserves the routing ability of the network for future requests. A general mathematical model has been formulated to solve this problem. Simulation results show that the proposed scheme outperforms existing schemes.

The congestion control mechanisms built into the transmission control protocol (TCP) use packet drops as a means to detect congestion occurring in the network. Unnecessary packet drops lead to poor performance for low-bandwidth delay-sensitive applications. Explicit congestion notification (ECN) is proposed as a mechanism to provide feedback to the sources about impending congestion in the routers without the need to drop packets. This requires the ECN bit of the IP packet to be marked at the router based on mechanisms like random early detection (RED) to identify congestion. We examine three different marking strategies, viz., mark-tail, mark-front and mark-random. The throughput performance of ECN flows and the unfairness among the ECN flows are examined. We also study the interaction between ECN and non-ECN flows.

In this paper we present bandwidth-delay constrained routing algorithms that use knowledge of the ingress-egress node pairs in the network in reducing the rejection rates for setting up new paths. Simulation is used to evaluate the algorithms and compare their performance against some existing algorithms for bandwidth constraints that have been modified to handle delay constraints. The results show that the proposed algorithms outperform all others under a wide range of workload, topology and system parameters.

In this letter, we analyze the performance of multiple input-queued asynchronous transfer mode (ATM) switches that use parallel iterative matching (PIM) for scheduling the transmission of head-of-line cells in the input queues. A queueing model of the switch is developed under independently, identically distributed, two-state Markov modulated Bernoulli processes bursty traffic. The underlying Markov chain of the queueing model is a quasi-birth-death (QBD) chain. The QBD chain is solved using an iterative computing method. Interesting performance metrics of the ATM switch such as the throughput, the mean cell delay, and the cell loss probability can be derived from the model. Numerical results from both the analytical model and simulation are presented, and the accuracy of the analysis is briefly discussed.

An analytical model for the performance analysis of a multiple input queued asynchronous transfer mode (ATM) switch is presented. The interconnection network of the ATM switch is internally nonblocking and each input port maintains a separate queue of cells for each output port. The switch uses parallel iterative matching (PIM) to find the maximal matching between the input and output ports of the switch. A closed-form solution for the maximum throughput of the switch under saturated conditions is derived. It is found that the maximum throughput of the switch exceeds 99% with just four iterations of the PIM algorithm. Using the tagged input queue approach, an analytical model for evaluating the switch performance under an independent identically distributed Bernoulli traffic with the cell destinations uniformly distributed over all output ports is developed. The switch throughput, mean cell delay, and cell loss probability are computed from the analytical model. The accuracy of the analytical model is verified using simulation.

An analytical model for the performance analysis of a novel input access scheme for an ATM switch is developed and presented in this paper. The interconnection network of the ATM switch is internally nonblocking and each input port maintains a separate queue for each output port so as to reduce the head-of-line (HOL) blocking of conventional input queuing switches. Each input is allowed to send only one cell per time slot, and each output port is allowed to receive only one cell per time slot. Using a tagged queue approach, an analytical model with an underlying two-dimensional Markov chain with a state space of size (queue capacity/spl times/switch size) is constructed for evaluating the switch performance under i.i.d Bernoulli traffic for different offered traffic loads. The switch throughput, mean cell delay, and cell loss probability are computed from the analytical model. The accuracy of the analytical model is verified using simulation.

The authors examine the problem of accurately modeling the software fault-tolerance technique based on recovery blocks. Analysis of some systems have investigated the correlation between software modules, which may be due to a portion of the functional specification that is common to all software modules, or to the inherent hardness of some problems. Three types of dependence which can be captured using measurements are considered. These are correlation between software modules for a single input, correlation between successive acceptance tests on correct module outputs and incorrect module outputs, and correlation between subsequent inputs. The authors' technique is quite general and can be applied to other types of correlation. In accounting for dependence, they use the intensity distribution introduced by D.E. Eckhardt and L.D. Lee (1985). A method of generating the intensity distribution that is based on the pairwise correlation between modules is discussed. This method is contrasted with the assumption of independent modules as well as the use of the beta-binomial density introduced by V.F. Nicola and A. Goyai (1990). The effects of dependencies were studied using a Stochastic Reward Network (SRN) that incorporates all of the above dependencies and a modeling tool called Stochastic Petri Net Package (SPNP).>