Commercial virtual-desktop computing is well established using computers optimized to serve as thin clients connected to centralized computing systems. Some common residential applications impose more stringent requirements on both communication bandwidth and latency than those of typical commercial applications. This article describes an objective study of residential applications accessed through thin-client virtual desktops for the purpose of investigating the feasibility of applying virtual-desktop computing to residential users. New metrics are introduced to quantify user-received application performance. The results suggest that certain commercial solutions with a commodity datacenter server show a strong potential for being adapted to residential virtual-desktop computing.

Large-volume scientific data is one of the prominent driving forces behind next generation networking. In particular, Software Defined Network (SDN) makes leveraging path-based network multicast services practically feasible. In our prior work, we have developed a cross-layer architecture for supporting reliable file-streams multicasting over SDN-enabled Layer-2 network, and implemented the architecture for a meteorology data distribution application in atmospheric science. However, it is challenging to determine an optimal rate for this application with the varying type, volume, and quality of meteorological data. In this paper, we propose a Quality of Service (QoS)-driven rate management pipeline to determine the optimal rate based on the input traffic characteristics and performance constraints. Specifically, the pipeline employs a feedtype classifier using Multi-Layer Perception (MLP) to recognize the type of meteorological data and a delay prediction regressor using stacked Long Short-Term Memory (LSTM) to predict per-file delay for the file-streams. Finally, we determine the optimal rate for the given file-streams using the trained regressor. We implement this pipeline to test the real-world file-stream data collected from a trial deployment, and the results show that our regressor outperforms all baselines by selecting the optimal rate in the presence of varying file set sizes.

The time- and space-varying nature of residual bandwidth in inter-datacenter networks (inter-DCNs) makes conventional end-to-end connections difficult to fully utilize the residual bandwidth. A promising solution is to introduce datacenter storage into data-plane paths. Delay-tolerant bulk data can be temporarily stored at intermediate sites and forwarded (SnF) at a later time when inter-DCN is less congested. However, the conventional methods attempt to involve multi-dimensional state information of the entire network in scheduling, which results in high computational complexity. In this paper, our studies reveal that there exist redundant states in scheduling, which cannot provide any performance benefit while imposing extra computational burden. Inspired by this finding, we propose a state-merging scheduling (SMS) method. By merging the state information of the pre-selected links, the SMS method naturally reduces the redundant states involved, which greatly improves the efficiency of SnF scheduling. Simulations demonstrate that the SMS method can outperform the conventional scheduling method, given a limit of the computational cost.

Increasingly bulk data transfers impose an unprecedented challenge on inter-datacenter (DC) wide area networks (WANs). To overcome this challenge, DC storage is exploited and introduced into the data-plane path so that delay-tolerant bulk data can be temporarily stored and forwarded (SnF) at a later time when network is less congested. However, the use of storage introduces additional complexity into the transfer process. In this paper, we evaluate the computational complexity of the SNF scheduling problem. Studies show that desirable performance can be attained by considering only a few alternate routes rather than dynamically routing over the entire network topology. Inspired by this finding, we propose a time-space decoupled (TSD) SNF scheduling method. By decoupling the problem into its spatial and temporal components and solving these components separately, the TSD method obtains good performance while maintaining low complexity. Simulations show that the TSD method can outperform the conventional joint method when the traffic load is medium or higher.

In prior work, we proposed a cross-layer architecture called Multicast-Push Unicast-Pull (MPUP) for Software Defined Networks (SDN) to support a reliable file-stream multicast application. In this work, we improved the algorithms used to set parameters: transport-layer sender retransmission timer, VLAN rate (which is also the sending rate) and sender-buffer size. Experimental evaluation using feeds with metadata collected from real meteorology file streams was conducted. A significant finding is that the throughput achieved is smaller than the VLAN/sending rate even though file blocks are multicast continuously in UDP datagrams. Sender-buffer waiting times and propagation delays are the main reasons for the degraded throughput. For example, increasing the VLAN rate from 20 Mbps to 500 Mbps, reduced the degradation from 90% to 45%. However, the degradation increased from 45% to 58% when the VLAN rate was increased from 500 Mbps to 1 Gbps. We found an increase in the number of block retransmissions at the higher rates, which explains this increased degradation. Increasing RTT from 0.1 ms to 100 ms caused throughput to drop from 274.8 Mbps to 27.6 Mbps on a 500 Mbps VLAN. If transmission delay was a significant component in total latency, then throughput degradation relative to VLAN rate would be small; however, the meteorology file-streams used in our study have small-sized data products. Due to bandwidth borrowing between VLAN and IP-routed services, VLAN utilization is not important, and hence we recommend using the smallest rate at which sender-buffer waiting times are insignificant.

During the past decade, botnets have become one of the most significant threats in the field of network security. A botnet attack typically works by infecting a device with malware and then recruiting it into a network of infected devices controlled by an attacker, which may lead to severe economic and social consequences. As a result, a considerable amount of research has been conducted to detect and prevent such attacks. In this paper, we create a foundation for an anomaly-based intrusion detection system to improve network security and to help reduce human involvement and error. The network traffic we used is captured in the form of connection logs, which are gen- erated by a popular network monitoring framework called Bro. We use our proposed framework to compare the performances of multiple supervised learning approaches, including Logistic Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM), Random Forest RF and Neural Networks (NN) at anomaly detection. We evaluated these models using F1 Score and Area Under Curve (AUC). Our models are trained on malicious network traffic samples from pervasive botnets like Zeus, Miuref and Conficker, as well as benign traffic samples. Using traditional cross validation, we illustrate that Random Forest has the best performance for anomaly detection. To test each algorithm's ability to generalize to unseen bot types, we implemented a custom-designed Leave-One-Out Cross Validation (LOBO-CV). In this procedure, each algorithm is trained on all but one bot family and evaluated on their ability to detect the unknown botnet traffic. We then improved overall detection performance by ensembling our two best models. Our results demonstrate we can detect both previously seen bot families and unseen botnet families with a high degree of confidence.

In this paper we introduce a new framework to detect elephant flows at very high speed rates and under uncertainty. The framework provides exact mathematical formulas to compute the detection likelihood and introduces a new flow reconstruction lemma under partial information. These theoretical results lead to the design of BubbleCache, a new elephant flow detection algorithm designed to operate near the optimal tradeoff between computational scalability and accuracy by dynamically tracking the traffic's natural cutoff sampling rate. We demonstrate on a real world 100 Gbps network that the BubbleCache algorithm helps reduce the computational cost by a factor of 1000 and the memory requirements by a factor of 100 while detecting the top flows on the network with very high probability.

Nowadays, the cyber, social and physical worlds are increasingly integrating and merging. Especially, combining the strengths of humans and machines helps tackle increasing hard tasks that neither can be done alone. Following this trend, this paper designs a Quality aware Truthful Incentive mechanism for cyber–physical enabled Geographic crowdsensing called Geo-QTI. Different from existing work, Geo-QTI appropriately accommodates the utilities of various stakeholders: requesters, participants and the crowdsourcing platform, and explicitly takes the requesters’ quality requirements, and participants’ quality provision into account. Geo-QTI explicitly includes four components: requester selection, participant selection, pricing and allocation. Requester selection with feasible analysis removes the requesters whose job cannot be completed by all participants or suffers from the monopoly participant (without the participant’s contribution, others cannot cover requesters’ requirement), obtains winning requesters set and determines actual payments. In participant selection phase, the platform aggregates the requested tasks (submitted by all winning requesters) in the sensed geographic area, and chooses the appropriate participants satisfying the winning requesters’ quality requirements with total cost as low as possible. Pricing phase determines the payments to winning participants. The phase of allocation assigns the specific participants to minimally cover the quality requirements of those winning requesters. Rigid theoretical analysis demonstrates Geo-QTI can achieve both requesters’ and participants’ individual rationality and truthfulness, computational efficiency and budget balance for the platform. Furthermore, the extensive simulations confirm our theoretical analysis, and illustrate that Geo-QTI can reduce requesters’ expenses greatly and ensure the fairness of allocation.

WAN provider links are often operated at low utilization levels, which leaves large unused capacity (headroom). In this paper, we propose using Software Defined Networking (SDN) controllers to support novel Static Headroom (SH) and Dynamic Headroom (DH) services to allow customers to fill this headroom with Elephant Flows (EFs) without adversely affecting the provider's ability to meet its Best-Effort (BE) service-level agreements, and ability to absorb extra traffic load created during failure recovery periods. Our solution calls for the use of lower-priority service for EFs. We use simulations to compare SH service with BE service, and DH service with SH service. When EFs are sent on BE service, they could cause packet losses in general-purpose IP traffic, especially when the burstiness of the latter is high, while with SH service, this packet loss rate is reduced to 0. While DH service requires the added complexity of a provider SDN controller, the ability to dynamically route EFs on lower-utilized links results in higher average EF throughput. The higher the non-uniformity (from a node-pair perspective) in network traffic, the greater the DH gain factor.

Hybrid network architectures with both Electrical Packet Switches (EPS) and Optical Circuit Switches (OCS) have been proposed for datacenter networks to reduce energy consumption and for cost savings. However, the high reconfiguration delay of OCS makes it challenging to leverage the high-speed OCS part of the hybrid network. In prior work, we proposed modifying applications to create network traffic patterns that match the characteristics of OCS, and developed, but did not evaluate, a modified version of Hadoop, called Hadoop for Hybrid Networks (HHN). This paper presents a comparative evaluation of HHN on hybrid networks and original Hadoop on EPS-only networks. The hypothesis tested is that it is feasible to achieve the same system-level performance, makespan and CPU utilization, and user-level metrics, such as job response time and fairness, with HHN while simultaneously achieving power and cost savings with the hybrid network. When the percentage of Shuffle-Heavy (SH) jobs is small, e.g., 5%, the performance of HHN was the same as that of original Hadoop on an EPS-only network. When the percentage of SH jobs was large, e.g., 20%, the performance was almost the same even at high loads when we placed an upper bound on the per-job input-data size. Without the upper bound, we found that one very large job (which required 19000 map tasks to process a 2.375 TB input dataset) caused HHN performance to be lower than that of original Hadoop because SH jobs can only be handled in a limited set of racks to which their input datasets were concentrated.

Optical switching technologies offer a cost-and power-efficient approach for handling the DataCenter Network (DCN) oversubscription problem. We propose a hybrid DCN architecture named Optical Switch in the Middle (OSM), which offers increased flexibility (when compared to prior hybrid architectures) for supporting multiple simultaneous high-speed TOR-to-TOR paths through an Optical Circuit Switch (OCS) and a core-level Electrical Packet Switch (EPS). A multilayer SDN controller supports advanced-reservation scheduling of optical circuits, and the integration of storage in the core EPS increases the usage rate of optical circuits. To effectively use the OSM architecture, we propose four modifications to Hadoop, and illustrate the potential of this architecture for achieving higher compute-resource utilization while simultaneously offering users shorter job completion times.

The growing deployment of OpenFlow/SDN networks makes it increasingly possible to leverage network multicast services. This work proposes a novel cross-layer Multicast-Push Unicast Pull (MPUP) architecture that includes functionality in the application, transport and link layers to offer users a reliable file-stream distribution service to multiple subscribers. A prototype implementation of the MPUP architecture was realized in a new version of Local Data Manager (LDM), LDM7, a software program that has been in use since 1994 for real-time meteorology data distribution. LDM6, the currently deployed version, uses application-layer multicast. Experiments were run on the GENI infrastructure to compare LDM7 and LDM6. The two main findings are (i) LDM7 can be run at a higher sending rate than LDM6 allowing for improved performance (lower filedelivery latency), and (ii) to achieve the same performance, LDM7 uses significantly lower bandwidth and compute capacity. A three-fold improvement in performance improvement was possible with LDM7, and a bandwidth reduction from 350 Mbps to 21.4 Mbps was observed with 24 receivers.

This paper aims to design a Hadoop system and evaluates the performance of a task allocation scheme. The task allocation scheme splits each job into tasks using an appropriate splitting ratio, and assigns tasks to slave servers based on server processing performance and network resource availability. We experimentally evaluate the performance of the scale out of the task allocation scheme with five machines. We focus on the configuration of jobtracker and tasktracker in Hadoop. In cases with heterogeneous Hadoop clusters, we distribute task blocks to high-capability slaves with proportionally larger-sized tasks than to low-capability slaves. We create an environment in which high-capability slaves perform more work than low-capability slaves. The experimental testbed results indicate that the task allocation scheme is effective.

Software switches offer flexibility to service providers but potentially suffer from low performance. A software switch called Lagopus was implemented using Intel's Data Plane Development Kit (DPDK), which offers libraries for high-performance packet handling. Prior work on software switches focused on characterizing packet forwarding throughput. In this work, we evaluated the impact of certain parameters and settings in Lagopus on application performance and studied packet drop rates. The importance of receive-thread packet classification for load balancing and to send delay-sensitive flows to a different worker thread from high-throughput flows was first demonstrated. Next, we showed that a loop-count variable used to control packet batching should be kept small in case link utilization is low. Finally, we showed that packet drop rate could be non-zero when the OpenFlow table size is large and packet arrival rate is high, and interestingly, the packet drop rate was higher with four worker threads than with a single worker thread. This implies a need for careful calibration and planning of the parameters of parallelization.

This paper describes a new reliable transport protocol designed to run on top of a multicast network service for delivery of continuously generated files. The motivation for this work is to support scientific computing Grid applications that require file transfers between geographically distributed data enters. For example, atmospheric research scientists at various universities subscribe to real-time meteorology data that is being distributed by the University Corporation for Atmospheric Research (UCAR). UCAR delivers 30 different feed types, such as radar data and satellite imagery, to over 240 institutions. The current solution uses an application-layer (AL) multicast tree with uncast TCP connections between the AL servers. Recently, Internet2 and other research-and-education networks have deployed a Layer-2 service using OpenFlow/Software Defined Network (SDN) technologies. Our new transport protocol, FMTP, is designed to run on top of a multipoint Layer-2 topology. A key design aspect of FMTP is the tradeoffs between file delivery throughput of fast receivers and robustness (measure of successful reception) of slow receivers. A configurable parameter, called the retransmission timeout factor, is used to trade off these two metrics. In a multicast setting, it is difficult to achieve full reliability without sacrificing throughput under moderate-to-high loads, and throughput is important in scientific computing grids. A backup solution allows receivers to use uncast TCP connections to request files that were not received completely via multicast. For a given load and a multicast group of 30 receivers, robustness increased significantly from 81.4 to 97.5% when the retransmission timeout factor was increased from 10 to 50 with a small drop in average throughput from 85 to 82.8 Mbps.

In a meteorology data-distribution application, streams of files are served to hundreds of receivers every day on unicast TCP connections. Software Defined Networking (SDN) offers a more scalable solution in which a rate-guaranteed Layer-2 multipoint virtual topology can be provisioned to have switches perform Ethernet-frame multicasting to support this application. A characterization of the file streams shows that file sizes and file inter-arrival times are both right skewed. The objective of this work is to design an algorithm for determining the rate of the Layer-2 multipoint virtual topology, and the size of the sending-host buffer, based on traffic characteristics of the file streams and performance requirements. Furthermore, the traffic characteristics are not exactly the same from day-to-day. An empirical method is proposed to determine the ideal rate and buffer size based on a day's traffic, which are then used along with the current rate and buffer size in an Exponential Weighted Moving Average (EWMA) scheme to determine the rate and buffer size for the next day. Our method was evaluated using metadata obtained for the top five file-streams of this meteorological data distribution, and found to be effective.

This paper proposes a new two-wavelength design for large-enterprise access links. The goal of this design is to lower power consumption and equipment costs without having a significant impact on performance. In our design, one wavelength is used as part of a lower-rate static circuit for general-purpose IP traffic, while the second wavelength is dynamically configured into a high-rate circuit for large dataset transfers whenever needed. A few provider-router ports are shared among a larger number of customers given that large dataset transfers are relatively infrequent. This leads to potential start-time delays, but results in significant power and cost savings. Using analytical models, we provide a quantitative power-and-cost comparison of our dynamic solution with the static always- on high-rate single-circuit solution used currently. The number of shared provider-router ports is kept large enough based on the number of customers to ensure that the probability of start-time delay exceeding a threshold is less than 1%.

This work advances the state-of-art in path scheduling and route selection by considering multiple call classes and allowing users to provide multiple start-time options in their requests for bandwidth in advance-reservation systems. Our system model supports two call classes: User-Specified Start Times (USST) and Earliest-Start Time (EST). The USST class is suitable for applications such as remote visualization or 4K immersive video-conferencing, while the EST class is suitable for high-speed large file transfers. Two path-selection algorithms, Earliest Available Path (EAP) and Shortest Available Path (SAP), are considered. Given that most core network links are operated at low utilization to support failure-recovery, we focus our simulation study on low-load operation. At low loads, there is no significant difference in call-blocking rates between EAP and SAP, and therefore, we recommend choosing EAP to lower mean waiting time. Since file-transfers, unlike audio-video calls, do not have an intrinsic rate requirement, we studied the question of what rate to choose for EST calls, and found that if EST calls were assigned the full link capacity, call blocking rates and mean waiting times were higher than if EST calls were assigned half the link capacity. Finally, we studied inter-class effects, and recommend design choices that mitigate these effects.

Multi-homing is one of the primary causes for the exponential growth rate of the global routing table (GRT). This work proposes a stub multihoming solution called Multi-Addressing with Stub Tunnels (MAST). By combining IPv6 multi-addressing with backup tunnels between a stub and each of its providers, stubs can enjoy the reliability advantages of multihoming without adding to the size of the GRT. An analysis of the current Internet BGP RIBs was conducted. It showed that the current global routing table size of 450K prefix entries would be reduced to about 24K entries if the MAST solution was adopted by all stubs.

High-rate large-sized (α) flows have adverse effects on delay-sensitive flows. Research-and-education network providers are interested in identifying such flows within their networks, and directing these flows to traffic-engineered QoS-controlled virtual circuits. To achieve this goal, a design is proposed for a hybrid network traffic engineering system (HNTES) that would run on an external server, gather NetFlow reports from routers, analyze these reports to identify α-flow source/destination address prefixes, configure firewall filter rules at ingress routers to extract future flows and redirect them to previously provisioned intra-domain virtual circuits. This paper presents an evaluation of this HNTES design using NetFlow reports collected over a 7-month period from four ESnet routers. Our analysis shows that had HNTES been deployed, it would have been highly effective, e.g., > 90% of α-bytes that arrived at the four routers over the 7-month period would have been redirected to virtual circuits. Design aspects such as whether to use /24 subnet IDs or /32 addresses in firewall filters, and which router interfaces' NetFlow reports to include in the HNTES analysis, are studied.

This paper describes traffic analysis undertaken to answer certain questions needed to design a hybrid network traffic engineering system (HNTES). The hybrid network in question consists of an IP-routed network and a dynamic virtual-circuit network, and the role of HNTES is to identify and redirect α-flows, which are defined as flows in which the number of bytes exceeds a threshold H (1 GB) over at least one α-interval (1 minute). NetFlow data from ESnet was analyzed. Our findings show that raw IP α-flows (identified by the 5-tuple) are mostly short-lived (80% are shorter than 2 minutes), which implies that HNTES should use an offline mechanism for identifying α-flows and preconfiguring policy-based routes to redirect packets for these flows to virtual circuits since setup delay is about 1 minute. An offline mechanism using prefix flow identifiers (/32 source and destination IP addresses or /24 source and destination subnets) appears to be highly effective because α-flow generating sources and sinks typically have static public IP addresses. For example, an analysis of two months of NetFlow data from an ESnet router shows that more than 94% of the total number of bytes from α-flows would have been redirected had a /24 based offline HNTES solution been deployed.

The goal of this work is to characterize scientific data transfers and to determine the suitability of dynamic virtual circuit service for these transfers instead of the currently used IP-routed service. Specifically, logs collected by servers executing a commonly used scientific data transfer application, GridFTP, are obtained from three US super-computing/scientific research centers, NERSC, SLAC, and NCAR, and analyzed. Dynamic virtual circuit (VC) service, a relatively new offering from providers such as ESnet and Internet2, allows for the selection of a path on which a rate-guaranteed connection is established prior to data transfer. Given VC setup overhead, the first analysis of the GridFTP transfer logs characterizes the duration of sessions, where a session consists of multiple back-to-back transfers executed in batch mode between the same two GridFTP servers. Of the NCAR-NICS sessions analyzed, 56% of all sessions (90% of all transfers) would have been long enough to be served with dynamic VC service. An analysis of transfer logs across four paths, NCAR-NICS, SLAC-BNL, NERSC-ORNL and NERSC-ANL, shows significant throughput variance, where NICS, BNL, ORNL, and ANL are other US national laboratories. For example, on the NERSC-ORNL path, the inter-quartile range was 695 Mbps, with a maximum value of 3.64 Gbps and a minimum value of 758 Mbps. An analysis of the impact of various factors that are potential causes of this variance is also presented.

A probe data service has been defined as a means for vehicular on-board units (OBUs) to transmit the data collected by their in-vehicle sensors in real time via wireless communication links to road-site units (RSUs). Sensor data fusion algorithms are developed in this work to aggregate data collected through this probe data service for computing single-RSU and multi-RSU average per-division speeds and average route travel times, where a division is defined to be a small segment of a road. These algorithms are evaluated using an integrated simulation testbed that consists of a vehicular simulator and a wireless communication link simulator, which are interconnected via a TCP/IP connection. Our key findings are that the probe data service offers an excellent means for computing average per-division speeds and average route travel times when the market penetration level of OBUs is relatively high. Privacy constraints negatively impact the accuracy of route travel time estimates as routes often span roads under the wireless coverage area of many road-side units requiring frequent vehicle identity changes. The impact of wireless communication losses is quite significant, requiring the use of accurate models in any simulation-based studies.

This paper puts forth a straw man proposal for future diverse internets. Guided by a principle that we call Align-and-Decouple (AD), our AD protocol architecture design calls for lightweight transport protocols that allow for decoupled operation of networks that are either owned by different organizations, or are networks of different types/protocols. TCP's fragmentation and congestion control mechanisms modify the original data stream so significantly that all new networks are essentially designed to just carry “TCP transformed” traffic. In contrast, with the AD principle, an AD-API is defined to consist of atomic transfer and pipe communication services, which can be supported by different providers with their own specific networking technologies, and yet allow for internetworking.

Wireless Access in Vehicular Environments (WAVE) networks operating in the Dedicated Short Range Communications (DSRC) band have been deployed as testbeds on which trials have been conducted recently. The goal of this paper is to analyze field-test measurements from these trials, and to use the data to validate assumptions made in simulation models. First, we analyzed the field-test measurements for message success rate and characterized its dependence on distance. Next, using trial-and-error methods, we found the “best” channel model assumptions that fit this field-test data. It consists of using the Nakagami model for fast fading, free space and shadowing model for slow fading, and a shadowing correlation model for retransmissions and acknowledgments. As the field tests were conducted in a variety of environments (tall and short urban-canyon as well as open field), this best-fit model is a generic model that will account for the variability that can be expected in actual deployments as vehicles move between different environments.

Recently, both research-and-education networks (RENs) and commercial networks have added a dynamic circuit service. With this service, users can request and obtain dedicated bandwidth for short durations (on the order of minutes to hours). Host-to-host (cluster-to-cluster) circuits are used for large file transfers. The network nodes used for this circuit service are not pure TDM or WDM circuit switches; instead they are hybrid nodes that include Ethernet interfaces that allow for the connection of host Ethernet NICs. While as with ordinary circuit switches, cross-connections have to be provisioned prior to data transfer, the technology allows for the creation of mismatched rate circuits, in which the wide-area portion of the circuit can be matched to the (typically lower) disk-access rates, while the access links from hosts to these switches are Gb/s Ethernet. In this work, an in-depth experimental investigation is performed to gain insights into the complex interactions between the TCP layer, ON/OFF flow control at the Ethernet layer, and switch buffer sizes (the Ethernet line cards have buffers). Using a novel tool and transport protocol designed for circuits, these interactions are characterized. An interesting new dynamic was found between flow control at the data-link layer and congestion control at the transport layer. With this in-depth characterization, it is clear that automated mechanisms are necessary to not only configure end-host TCP send and receive buffer sizes as is required for high throughput across IP-routed networks, but for circuit networks, additionally, Ethernet-layer output queue (called qdisc in Linux) size needs to be set, along with flow-control related parameters within switches.

In prior work, we proposed a hybrid architecture that complements the connectionless Internet with a high-speed, dynamically shared circuit-switched network for file-transfer applications. "Smallrdquo files are directed to the connectionless network while for "largerdquo files, a high-speed circuit setup is attempted. If the attempt is unsuccessful (because all circuits are already in use), the application will fall back to using the connectionless path. This paper addresses the question of how to select appropriate values for operational parameters in this hybrid architecture. Specifically, we study the questions of what sized files to direct to the circuit-switched network and how much bandwidth to allocate per file transfer. To answer these questions, we construct a model of this hybrid architecture by combining the Erlang-B call blocking model with TCP delay models. Our model captures a combination of utilization and delay considerations in the selection of minimum file size above which transfers should be directed to the circuit-switched network. The optimal per-call circuit rate and the optimal minimum file size are determined to maximize the benefits of using the circuit network, which is quantified with an average-delay-reduction metric.

High-speed optical circuit-switched networks are being deployed to support highly demanding applications in the eScience research community. File transfers were identified as an ideal application for high-speed circuits because such transfers can utilize any rate, limited only by end-host capabilities. This paper addresses the question of how a circuit-switched network can be shared efficiently amongst many users for file transfers. Specifically, we study the question of how much bandwidth should be allocated per file transfer so that the system performance averaged across all flows is optimal in terms of mean response time. We explore a heterogeneous bandwidth-allocation scheme, in which bandwidth is allocated based on file sizes, as well as a homogeneous scheme, in which all transfers are allocated the same amount of bandwidth. Our analytical results show that bandwidth allocation should be based on loading conditions. Under low loads, we recommend a homogeneous scheme and each call should be allocated a high-bandwidth circuit. However, under high loads, small files should be allocated low-bandwidth circuits and large files high-bandwidth circuits; this heterogeneous bandwidth-allocation scheme results in a one-to-three orders of magnitude reduction in mean response time.

We are interested in exploring how to virtualize network switches in order to create multiple logical switches from one physical switch. The motivation behind this is to be able to support multiple research-and-education (R&E) projects by offering each of them a logical switch that could coexist with other R&E projects' logical switches, so that each would have the equivalent of their own switch. We propose an approach for virtualizing off-the-shelf Ethernet switches that have built-in support for creating isolated bandwidth partitions on their data-plane interfaces. Our solution is to implement two software modules that are run external to the switches, a slice scheduler and a slice administration controller (SAC). We applied our approach to virtualizing a specific Ethernet switch, i.e. the Force10 E600 model. We describe our implementation, and show how a slowdown of 3% to 26%, based on the type of administrative command issued, is experienced when using the SAC.

In this article, we present a novel discrete-time Markov chain model of book-ahead bandwidth-sharing mechanisms. We use this analytical model and a simulation model to understand the benefits of book-ahead (BA) bandwidth-sharing when compared to the immediate-request (IR) call-blocking mode of bandwidth-sharing in circuit-switched networks. We study two different BA schemes, BA-all, in which the caller accepts any set of available timeslots, and BA-n, in which the caller specifies n call-initiation time options. Numerical results show that the BA-all achieves 95% utilization with a call-blocking probability of only 1%, while in the IR mode, call blocking probability is 23% even when utilization is only 80%. The BA-n schemes perform as well as the BA-all scheme if the call-initiation time options are restricted to fall on timeslot boundaries separated by the minimum call holding time. The length of the advance reservation horizon, K, is shown to increase linearly with call holding time, H. The ratio K/H is primarily dependent on the link capacity in channels. For example, if the link is divided into 10 channels, to achieve a 2% call blocking probability, the advance-reservation horizon needs to be a factor of 4 times the call holding time. In other words, the extra data storage and processing required to accept and maintain advance reservations is not significant.

Optical networks with book-ahead bandwidth schedulers are being deployed to meet the high-speed and predictable-service networking requirements of applications in the scientific research community. We present an analytical model for a single-link book-ahead bandwidth scheduler, which responds to advance reservation requests with the first-available time interval in which a channel is available. The link is assumed to be divided into m channels, and time is discretized into intervals. Our proposed model is a non-homogeneous continuous-time Markov chain, which has an embedded discrete-time Markov chain. We solve the model for call congestion, mean scheduling delay, and link utilization. This model can be used by network designers to select the size of the reservation window, K, in time-interval units, corresponding to a desired set of values for the output metrics, for a given value of m. For example, when m is 8, increasing the reservation window beyond 4 intervals does not affect the call congestion or system utilization, but causes the mean scheduling delay to increase. We show with comparative simulations that our analytical model can be used as a solution for an M/D/m/p queueing system at moderate-to-high loads.

In this paper, we address the question of how to introduce new Quality-of-Service enhancements into the existing Internet. Specifically, dynamically shared circuit-switched/virtual-circuit (VC) networks are being deployed in some of the backbone research and education networks. We describe an overlay solution for enabling access of end hosts that are not physically connected to these networks to nevertheless benefit from their services. Our solution is to use Web proxy servers that are dually connected to the Internet and these backbone circuit/VC networks. We implemented and demonstrated our solution on a wide-area experimental GMPLS network called CHEETAH. Building on our prior results, which show the benefits of using this network for wide-area transfers of large files, we demonstrated performance benefits for Web clients. For example, in an experimental setup, we showed that for a 10-MB file, the Web throughput improves from 17 Mb/s to 61 Mb/s by using this overlay solution. In addition, this technique offers a method for introducing new QoS-oriented network services in a backbone network to non-connected users.

Signaling protocols for GMPLS networks have been standardized and implemented in switch controllers. Most switch vendors allow for signaling messages to be carried over in-band signaling channels as well as through out-of-band networks. In this paper, we compare these two signaling transport options. In carrying out this analysis, we allow for both software-implemented signaling protocol processors, as is common in most off-the-shelf switches today, as well as hardware-accelerated signaling protocol engines. Our motivating application is file transfers, which have high call arrival rates and low call holding times. The resulting high signaling message load makes the question of which signaling transport solution to use important. Network delays are lower in the in-band option given that there are no IP routers on these paths, but the per-channel transmitter rates are lower for in-band channels. To study this tradeoff, we set up queueing models and obtained delay estimates. Our analysis shows that with hardware-accelerated signaling engines, in-band signaling is the better option to keep end-to-end call setup delays low. With software signaling protocol processors, the most significant component is the message processing delay, which is likely to include queueing delays, making the question of in-band or out-of-band transport less important.

In this article, we describe our experiences in implementing an experimental wide-area GMPLS network called CHEETAH (circuit-switched end-to-end transport architecture). The key concept is to add a complementary end-to-end circuit based service with dynamic call-by-call bandwidth sharing to the connectionless service already available to end hosts via the Internet. The current CHEETAH experimental network consists of off-the-shelf GMPLS-capable SONET switches (with Ethernet interfaces) deployed at three locations, Research Triangle Park, North Carolina, Atlanta, Georgia, and Oak Ridge, Tennessee. We describe our solutions to various problems relating to control-plane design, IP addressing and control-plane security. We designed and implemented a CHEETAH software package to run on Linux end hosts connected to the CHEETAH network. Among other functions, this software package includes an RSVP-TE module to enable end users and applications to dynamically initiate requests for dedicated end-to-end circuits and receive/respond to requests for circuits. We present measurements for typical end-to-end circuit setup delays across this network. For example, end-to-end circuit setup delay from a Linux end host in NC to a host in Atlanta is 166 ms

After identifying that current GMPLS specifications only allow for the implementation of a call blocking mode of operation that handles immediate-request calls (not book-ahead calls), this paper addresses the question of suitability of different types of applications for such networks. Two types of applications are considered, ones in which the per-circuit bandwidth and mean call holding time are independent and ones in which they are dependent (file transfers). GMPLS networks are often cited for high-bandwidth applications and for large file transfers. This work, through analysis, reasons that GMPLS networks are not well suited both for applications in which the required per-circuit bandwidth is high, on the order of one-tenth the shared link capacity, or low, on the order of one-thousandth the shared link capacity. Ideal is the one-hundredth range. Further for file transfers, we show that the best range of operation to achieve high utilization and low call blocking rates with an acceptable number of ports on switches for traffic aggregation, requires call holding time to be small, in the range of seconds.

Connection-oriented networks that offer guaranteed-rate circuits are being proposed as a solution to meet the high-end networking needs of distributed scientific research. This paper describes a transport protocol for data transfer across dedicated end-to-end circuits. We call this protocol Circuit-TCP (C-TCP). The key feature of C-TCP is to maintain a fixed sending rate, as closely matched to the circuit rate as possible, with the aim of achieving high circuit utilization. C-TCP is implemented by modifying TCP code in Linux. Experiments on a wide-area experimental circuit-switched testbed show that C-TCP is able to fully utilize circuit bandwidth and sustain a high circuit utilization. The challenge in this work lay in matching the variability inherent in general-purpose end hosts with the monotonic nature of circuits.

In this article, we describe our experiences in implementing an experimental wide-area GMPLS network called CHEETAH (Circuit-Switched End-to-End Transport Architecture). The key concept is to add a complementary end-to-end circuit based service with dynamic call-by-call bandwidth sharing to the connectionless service already available to end hosts via the Internet. The current CHEETAH experimental network consists of off-the-shelf GMPLS-capable SONET switches (with Ethernet interfaces) deployed at three locations, Research Triangle Park, North Carolina, Atlanta, Georgia, and Oak Ridge, Tennessee. We designed and implemented a CHEETAH software package to run on Linux end hosts connected to the CHEETAH network. Among other functions, this software package includes an RSVP-TE client module to enable end users and applications to dynamically initiate requests for dedicated end-to-end circuits and receive/respond to requests for circuits. We present measurements for typical end-to-end circuit setup delays across this network. For example, end-to-end circuit setup delay from a Linux end host in NC to a host in Atlanta is 166ms.

High-speed optical networking infrastructures such as National LambdaRail promise to revolutionize the way that we approach grid computing. With optical-fiber speeds doubling every 9 months and computing speeds doubling only every 18-24 months over the past few decades, we have finally reached a crossroads where network speeds have outstripped the ability of processors to keep up. Thus, we may be entering a new world where the central architectural element in grid computing is the optical network, not the end-host computer(s). The goal of this panel is to discuss whether such a new world is emerging, and if so, what the research challenges will be.

Recently, polling has been included as a resource sharing mechanism in the medium access control (MAC) protocol of several communication systems, such as the IEEE 802.11 wireless local area network, primarily to support real-time traffic. Furthermore, to allow these communication systems to support multimedia traffic, the polling scheme often coexists with other MAC schemes such as random access. Motivated by these systems, we develop a model for a polling system with vacations, where the vacations represent the time periods in which the resource sharing mechanism used is a non-polling mode. The real-time traffic served by the polling mode in our study is telephony. We use an on-off Markov modulated fluid (MMF) model to characterize telephony sources. Our analytical study and a counterpart validating simulation study show the following. Since voice codec rates are much smaller than link transmission rates, the queueing delay that arises from waiting for a poll dominates the total delay experienced by a voice packet. To keep delays low, the number of telephone calls that can be admitted must be chosen carefully according to delay tolerance, loss tolerance, codec rates, protocol overheads and the amount of bandwidth allocated to the polling mode. The effect of statistical multiplexing gain obtained by exploiting the on-off characteristics of telephony traffic is more noticeable when the impact of polling overhead is small

Signaling protocols, primarily used to set up and teardown connections, are essential in connection-oriented networks. Up to now, signaling protocols are mostly implemented in software for two reasons: complexity and the requirement for flexibility. Adversely, the price paid is in performance. Software implementations of signaling protocols are rarely capable of handling over 1000 calls/s. Corresponding call setup delays per switch are in the order of milliseconds. To improve performance for high-speed networks, we implemented a subset of the resource reservation protocol-traffic engineering signaling protocol in reconfigurable field programmable gate array hardware. Our implementation demonstrates the feasibility of 1000x speedup vis-a/spl grave/-vis software implementations. The impact of this work is far-reaching in that it enables connection-oriented networks to support new applications that require rate guarantees but have short call holding times.

Scheduling resources on Grids is a well-known problem. The extension of Grids, to LambdaGrids requires scheduling of lambdas, i.e., end-to-end high-speed circuits. In this paper we propose an heuristic for the scheduling of lambdas specifically for file transfers, given that many eScience applications require high-throughput transfers of large files. We call this heuristic varying-bandwidth list scheduling (VBLS) because the scheduler returns a time-range-capacity (TRC) allocation vector with varying banchwidth levels assigned for different time ranges within the duration of a transfer. The advantage of VBLS over a fixed-bandwidth allocation scheme is that it allows the scheduler to backfill any holes left in resource allocations. Such a scheme is enabled by end host applications specifying the file size in their transfer requests. To characterize VBLS, we used analytical models and ran simulations. Our results show that VBLS performance is close to packet-switching performance, which means that file transfers can take advantage of bandwidth that becomes available subsequent to the start of transfers, a critical drawback of typical fixed-bandwidth allocation schemes in circuit-switched networks.

Time-Division Multiplexing/Frequency-Division Multiplexing (TDM/FDM) schemes are typically used in a fixed-bandwidth allocation mode, which means a call is assigned a fixed amount of bandwidth for its whole duration. For file transfers, such schemes compare unfavorably against statistical multiplexing schemes such as packet switching. This is because in fixed-bandwidth TDM/FDM schemes, once a file transfer is allocated a certain bandwidth, it cannot take advantage of bandwidth that becomes available as a result of other transfers completing. In this paper, we propose a Varying-Bandwidth List Scheduling (VBLS) heuristic for TDM/FDM networks in which a sender specifies the file size, maximum bandwidth limit and a desired start time, and the network returns a time-range-capacity allocation vector assigning varying bandwidth levels in different time ranges for the transfer. Simulation results show that VBLS performance is indistinguishable from packet-switching performance, and hence superior to the fixed-bandwidth allocation mode.

Signaling protocols are primarily implemented in software for two reasons: protocol complexity and the requirement for flexibility. While these are two good reasons for implementing signaling protocols in software, the price paid is in performance. Software implementations of signaling protocols are rarely capable of handling over 1000 calls/sec. Corresponding call setup delays per switch are in the order of milliseconds. To improve performance for high-speed networks, we implemented RSVP-TE signaling protocol in reconfigurable FPGA hardware. Our implementation demonstrates the feasibility of 100x and potentially 1000x speed-up vis-a-vis software implementation. The impact of this work can be quite far-reaching by allowing connection-oriented networks to support a variety of new applications, even those with short call holding times.

We propose a service concept in which high-speed Ethernet interfaces from end hosts are dynamically cross-connected to equivalent-rate Ethernet-over-SONET (EoS) optical circuits for transport across metro-area networks and/or wide-area networks. We call our service concept reconfigurable Ethernet/SONET circuits to end users (RESCUE). We describe how RESCUE can be used for two applications: dial-up service to Internet service provider routers and file transfers. We propose to deploy RESCUE service as an "add-on" to current Internet access for many reasons. Primary among these is that it allows a metro optical circuit-switched network to be operated at a high utilization, which is important to achieve a cost-effective bandwidth-efficient network. Given that end hosts with access to RESCUE service will have a choice of two paths, the primary Internet path and a secondary RESCUE option, end-host applications will need to make a routing decision. We carry out a quantitative analysis to provide a basis for this routing decision for both dial-up service and file transfers. For example, with the file-transfer application, if call-blocking probability on the optical circuit-switched network is 30% and the packet-loss rate on the transmission control protocol/Internet protocol path is 1%, a circuit setup should be attempted for files 180 KB or larger in low-propagation delay environments.

We propose a dial-up optical access architecture that allows an end host dynamically to request and receive a high-speed circuit for connection to its Internet service provider's router. We carry out an analytical comparison of the current leased-line architecture and our proposed dial-up architecture for access networks. It shows significant capacity savings, for example, savings of 39% at 50% traffic load for a call blocking probability of 0.5% assuming 3000 hosts share the optical link. With a call queueing model, we demonstrate that even if the circuit is held for long durations in the dial-up architecture, bandwidth savings are possible over the leased-line architecture.

The IEEE 802.11 MAC protocol supports two modes of operation, a random access mode for non-real-time data applications, and a polling mode for real-time applications. We design and analyze a system that uses the polling mode for interactive voice traffic. With larger inter-poll periods, more voice calls can be accommodated, but at the expense of increased delay. For example, our analysis shows that with an inter-poll period of 90 ms, a maximum of 26 voice calls can be handled with a worst-case delay of 303 ms, whereas with an inter-poll period of 60 ms, a maximum of 17 voice calls can be handled with a worst-case delay of 213 ms. We also carry out an error analysis that demonstrates the need for error correction of voice packets.

We examine the problem of address interworking that arises when two networks of the same type (that use the same addressing format) are interconnected by a different type of network (one that uses a different addressing format). Solutions to this problem fall broadly into four categories: (i) schemes that send address resolution messages when a call/packet arrives ("pull" information), (ii) schemes in which address resolution information is sent in routing or routing-like protocols ("push" information), (iii) schemes that use administered cross-office address translation tables at gateways, and (iv) schemes that use encapsulated/mapped addresses. Internetworking problems that are of interest today include PSTN/ATM, IP/ATM and PSTN/IP internetworking. The PSTN uses 8-byte E.164 addresses (telephone numbers), IP networks use 4-byte IPv4 addresses, and ATM networks use 20-byte ATM End System Addresses (AESAs). Current solutions to the address interworking problem for these internetworks fall into the first three categories of solutions. We propose use of the last solution category for these three internetworking problems. We provide a comparison of our proposed scheme with other solutions.

This paper presents an algorithm for optimizing the route of a connection that becomes suboptimal due to operations such as handoffs and location-based reroutes, and applies this algorithm to the handoff management problem in networks with hierarchical link state routing protocols such as PNNI (Private Network-to-Network Interface)-based ATM (Asynchronous Transfer Mode) networks. The route optimization algorithm uses hierarchical route information of the connection and summarized topology and loading information of the network to determine a "crossover node" such that adjusting the connection from that crossover node results in an optimally routed connection. Handoff management schemes that perform local rerouting of connections have been proposed in order to support fast handoffs. These methods result in suboptimally routed connections. In this paper, we demonstrate how this route optimization algorithm can be used to optimize the route of a connection after such a handoff is executed, as the second phase of two-phase handoff scheme. This route optimization procedure can also be executed as part of the handoff procedure resulting in a one-phase handoff scheme. Applying this route optimization algorithm, we propose two one-phase schemes, the one-phase optimal scheme and the one-phase minimal scheme. A comparative performance analysis of one- and two-phase handoff schemes is presented. Measures of comparison are handoff latency and the amount of network resources used by a connection. Handoff latency in the one-phase optimal scheme is greater than that in the two-phase schemes, and handoff latency in the one-phase minimal scheme is smaller than that in the two-phase schemes. The one-phase methods show a significant increase in efficiency of the connection compared to the two-phase methods.

In this paper, we propose an algorithm for optimizing the route of a connection that becomes suboptimal due to operations such as handoffs and location-based reroutes, for mobile ATM (asynchronous transfer mode) networks based on the PNNI (private network-to-network interface) standard. This algorithm uses hierarchical route information of the connection and summarized topology and loading information of the network to determine a "crossover node" such that adjusting the connection from that crossover node results in an optimally routed connection. We then apply this algorithm to the mobile location management problem. Location management schemes have been proposed in which an incoming call to a mobile is first routed to its home switch (based on summarized reachability data) and then rerouted to the mobile's current location. If we consider such rerouting a "first phase" of mobile connection setup, a "second phase" is required to optimize the paths of such rerouted connections. Such an approach is considered a two-phase mobile location/connection setup scheme. Alternatively, an incoming call to a mobile can be first routed to its home switch based on summarized reachability data, and then "cranked back" to an optimal crossover node before rerouting the connection to the mobile's current location. Such a scheme is a one-phase mobile location/connection setup scheme since it directly results in an optimal path. A comparative performance analysis of the one- and two-phase connection setup schemes is presented. Measures of comparison are call setup delay and the amount of network resources allocated to a connection. The maximum call setup delay (worst case call setup delay) is lower in the two-phase scheme, but the average call setup delay is lower in the one-phase scheme. The amount of resources required for a connection in the two-phase scheme (prior to route optimization) is significantly greater than that in the one-phase scheme.

In this paper, we investigate the issue of how multiparty-to-multiparty communication should be realized. In order to quantify the different techniques, we examine the network-layer (bandwidth and buffer) allocations of four models for realizing multiparty-to-multiparty communication. An example application considered in this paper, where such communication will be needed, is video conferencing based on MPEG video streams. Analytical comparisons are carried out based on actual data obtained from empirical measurements of the source characteristics of MPEG video (determined to be a 40 state Markov modulated fluid source). Our results show that in order to minimize bandwidth and buffer (network-layer resources) allocations, we must exploit the effect of statistical multiplexing by using shared buffers for such multiparty-to-multiparty communication. Using these results we identify the best technique for multiparty-to-multiparty communications.

This paper presents two mobile location management algorithms for ATM (asynchronous transfer mode) networks based on the PNNI (private network-to-network interface) standard. The first solution, referred to as the mobile PNNI scheme, builds on the PNNI routing protocol by using limited-scope (characterized by a parameter S) reachability updates, forwarding pointers and a route optimization procedure. The second solution, referred to as the LR (location registers) scheme, introduces location registers (such as the cellular home and visitor location registers) into the PNNI standards-based hierarchical networks. This scheme uses a hierarchical arrangement of location registers with the hierarchy limited to a certain level S. A comparison of the two schemes indicates that at low CMRs (call-to-mobility ratios), the LR scheme incurs a lower average total cost (consisting of the average move and average search costs) than the mobile PNNI scheme, while at high CMRs, the mobile PNNI scheme incurs a lower cost.

In earlier papers, we proposed the PCC (parallel connection control) algorithm for the setup and release of on-demand ATM (asynchronous transfer mode) connections. This paper describes an implementation and analysis of PCC. The PCC algorithm is implemented on a testbed of three types of ATM switches from which service time measurements (software execution times) are obtained. These measured service times are used as input data for an analytical queueing network model to characterize the PCC end-to-end setup delay and maximum throughputs. Assuming all connections pass through ten switches, for this measured data, PCC has a theoretical maximum throughput of 343 calls/sec/switch, and at about 90% of this maximum limit (i.e., at 310 calls/sec/switch), the mean end-to-end connection setup delay is only 43 ms. We also compare PCC performance against an equivalent sequential connection setup/release approach. We observe that PCC end-to-end setup delay is much smaller than the sequential setup delay (34% at low load and even smaller at higher load). Also, the per-switch maximum throughput of the PCC approach is 80% more compared to the sequential approach.

Increasingly, there is a need for context-aware networking. For example, in multimedia networks with heterogeneous end equipment, if two or more communicating nodes do not have the same application-layer interface, the network should be aware of the "context" of the communication in order to select an appropriate converter through which to route connections. In this paper, we motivate the need for logical connections, where a logical connection consists of one or more network-layer connections routed through application-layer resources, such as bridges and/or format/rate converters. We then describe the application-aware routing protocol (AARP) for managing logical connections.

To support mobile endpoints in ATM (Asynchronous Transfer Mode) networks, location management schemes have been proposed in which an incoming call to a mobile is first routed to its home switch and then rerouted to the mobile's current location.In this paper, we propose an algorithm for optimizing the route of a connection that becomes suboptimal due to such location-based reroutes in mobile ATM networks based on the PNNI (Private Network-to-Network Interface) standard.This algorithm uses the hierarchical route information of the connection and the summarized topology and loading information of the network to determine a "crossover r&e" such that adjusting the connection from that crossover node results in an optimally routed connection.This route optimization procedure can be executed during mobile connection setup resulting in a one-phase mobile location and connection setup scheme.We also demonstrate how this route optimization procedure can be executed after a connection is set up to a mobile as a secondphase of a two-phase mobile location and connection setup scheme.A comparative performance analysis of the one-and two-phase connection setup schemes is presented.Measures of comparison are call setup delay and the amount of network resources allocated to a connection.The one-phase scheme incurs higher call setup delays, but results in an optimal route during connection setup.On the other hand, the two-phase scheme incurs lower call setup delays, but results in connections with suboptimal routes, prior to the execution of the route optimization phase. -.

This paper presents two mobile location management algorithms for ATM (asynchronous transfer mode) networks based on the PNNI (private network-to-network interface) standard. The first solution is called the mobile PNNI scheme because it builds on the PNNI routing protocol. It uses limited-scope (characterized by a parameter S) reachability updates, forwarding pointers, and a route optimization procedure. The second solution is called the LR (location registers) scheme because it introduces location registers (such as the cellular home and visitor location registers) into the PNNI standards-based hierarchical networks. This scheme uses a hierarchical arrangement of location registers with the hierarchy limited to a certain level S. Analytical models are set up to compare the average move, search, and total costs per move of these two schemes for different values of the CMR (call-to-mobility ratio), and to provide guidelines for selecting parameters of the algorithms. Results show that at low CMRs (CMR<0.025), the LR scheme performs better than the mobile PNNI scheme. We also observe that the two schemes show a contrasting behavior in terms of the value to be used for the parameter S to achieve the least average total cost. At low CMRs, the parameter S should be high for the mobile PNNI scheme, but low for the LR scheme, and vice versa for high CMRs.

This paper proposes algorithms for handoff, location, and connection management in a wireless asynchronous transfer mode (ATM) local-area network (LAN). Fast handoffs while maintaining cell sequence and quality-of-service (QoS) guarantees are achieved by distributing switching functionality to base stations, and using a networking scheme based on provisioned virtual trees. A new distributed location management scheme using a minimal registration procedure and broadcasts on wired links is proposed for this LAN. The detailed signaling procedures that support the algorithms for mobility and connection management are described. Finally, an implementation of these procedures and an analysis of the measured data is presented. Measurements of service times obtained from this implementation indicate that over 100 calls/s. can be handled by each node in 50-node network with a high-percentage of mobiles (75%) relative to fixed endpoints. This is comparable to current wired ATM switch call handling throughputs, in spite of the fact that these nodes perform additional handoff and location management functions. The data also indicates handoff latency times of 1.3 ms. This validates our proposal for maintaining cell sequence while performing handoffs.

A common approach used for decreasing end-to-end connection setup delay in ATM networks is to provision partial segments a priori using virtual path connections (VPCs). In this paper, we present an analysis to study the effect of such provisioning. The analysis demonstrates that significant savings in bandwidth, and hence network costs, are achieved if the number of provisioned VPCs are minimized, especially if there is uncertainty in the traffic pattern characterization. For a given example, we show that to achieve a 0.1% blocking probability requirement, if link capacities are shared (i.e., not partitioned a priori to provisioned VPCs), resource savings of 14.4%, 47.9% or 80% are possible under exact, approximate and poor traffic characterizations, respectively. If the above bandwidth savings are to be achieved in networks that use traditional node-by-node connection setup approaches, end-to-end connection setup delay will increase. In this paper, we propose a new connection setup algorithm that allows the savings in bandwidth obtained through sharing while at the same time controlling the connection setup delay.

This paper presents a Wireless Virtual Local Area Network (WVLAN) to support mobility in IP-over-ATM networks.Mobility is handled by a joint ATM-layer handoff for insession mobility and MAC-layer handoff for inter-session mobility, such that the effects of mobility are localized and transparent to the higher-layer protocols.Different functions, such Address Resolution Protocol (ARP), mobile location, and ATM connection admission are combined to reduce front-end delay for connectionless packet transmission in connection-oriented ATM networks.The proposed WVLAN, through the use of ATM technology, provides a scalable wireless virtual LAN solution for IP mobile hosts.

We present a comparison of the control signaling load of two vastly different architectures for providing personal communication services (PCSs). One architecture is based on current cellular networks. The other architecture, called the wireless distributed call processing architecture (WDCPA), distributes processing from the mobile switching centers and cell sites and executes new procedures for tracking mobile users and locating mobile users to deliver calls. We determine the signaling load generated within each system to support mobility management and call control based on standard assumptions about the operating parameters of a cellular network. Our results show that, when compared to current cellular systems, for simple single-connection services, WDCPA has marginally reduced cross-network signaling loads. For multiconnection calls, WDCPA incurs 35% less total signaling load for mobility management, has reduced cross-network signaling load for mobility management by up to 65%, and depending on the user model (e.g., data or telecommunication), has reduced total cross-network signaling load, including procedures for call/connection and mobility management, by up to 55% when compared to current cellular systems, while more flexibly supporting services.

The paper defines new performability-based QOS (quality of service) metrics to characterize the quality of services offered over telecommunication and computer communication networks. Performability metrics used to characterize fault-tolerant computer systems are not directly applicable to services. This paper discusses the various challenges that arise in defining performability metrics for services. These metrics are defined both from a service provider perspective as well as a customer perspective. In order to model real-world service offerings, which often involve multiple vendor networks, we define service segment metrics from which overall service metrics can be obtained. Additionally, these service segment metrics can be used in negotiations between the different vendors participating in the joint service offering.

Abstract Several algorithms have been developed to solve the reliability problem for nonseries‐parallel networks using thesum of disjoint products (SDP)approach. This paper provides a general framework for most of these techniques. It reviews methods that help improve computer time and memory requirements in reliability computation. These parameters are generally used to compare SDP algorithms. We also overview three multiple variable inversion algorithms that result in sum of disjoint products expressions with fewer terms than that of algorithms that use only a single‐variable inversion. One common network is solved for two‐terminal network reliability using each of these algorithms. Finally, we have provided a comparison among these techniques.

The need for the combined performance and reliability analysis of fault tolerant systems is increasing. The common approach to formulating and solving such problems is to use (semi-)Markov reward models. However, the large size of state spaces is a problem that plagues Markovian models. Combinatorial models have been used for modeling reliability and availability of complex systems without paying the price of large Markov models. However, assumptions of two-state behavior of components (and that of the system), independence assumptions of component state transitions, and restrictive repair assumptions decrease the potential of combinatorial models for realistic systems. The authors propose a combinatorial algorithm for the combined performance and reliability analysis of coherent repairable systems with multistate components, allowing interdependent component state transitions. An example illustrating the algorithm is also presented.>

The common approach to formulating and solving combined reliability/availability and performance problems is to use Markov reward models. However, the large size of state spaces is a problem that plagues Markovian models. Combinatorial models have been used for modeling reliability and availability of complex systems without paying the price of large Markov models. Yet, assumptions of two-state behavior of components (and that of the system), independence assumptions of component failure behavior, and restrictive repair assumptions decrease the potential of combinatorial models for realistic systems. A combinatorial approach is proposed for the combined performance and availability analysis of coherent repairable systems with multi-state components, allowing inter-dependent component state transitions. Examples showing the usefulness of the approach are presented.>

An efficient Boolean algebraic algorithm for the symbolic reliability and sensitivity analysis of coherent two-terminal networks with s independent components is described. The algorithm is also applicable to a fault tree model without NOT gates. The algorithm uses the concept originally proposed by A. Grnarov, L. Kleinrock, and M. Gerla (1979). After the algorithm is presented, the errors in the original technique are illustrated by two examples. The algorithm is extended t compute the reliability importance of a given component (sensitivity of system reliability to a given component's reliability). A computer program implementing the modified algorithm is used to solve and obtain measured time complexities for a large set of network and fault tree models.>

A study is made of several methods for reducing complex fault tree models of fault-tolerant distributed systems. For each method the authors provide bounds on the estimate of unreliability that is obtained from the reduced model. They discuss methods for truncating the solution of a model expressed as a fault tree and then develop techniques that apply to the construction of the fault tree model. The emphasis is on producing approximate (but bounded) results applicable to realistic systems. The authors also discuss methods for incorporating dynamic system behavior (error handling and redundancy management) into fault tree models, and the corresponding truncated solution. The methods are presented as they are used in modeling two distributed systems, the Cm* system and AIPS (the Advanced Information Processing System).>