Large File Transfers Research Articles

The HALO network/sup TM/

The High Altitude Long Operation Network/sup TM/ is a broadband wireless metropolitan area network, with a star topology, whose solitary hub is located in the atmosphere above the service area at an altitude higher than commercial airline traffic. The HALO/Proteus airplane is the central node of this network. It will fly at altitudes higher than 51,000 ft. The signal footprint of the network, its "Cone of Commerce", will have a diameter on the scale of 100 km. The initial capacity of the network will be on the scale of 10 Gb/s, with growth beyond 100 Gb/s. The network will serve the communications needs of each subscriber with bit rates in the multimegabit per second range. A variety of spectrum bands licensed by the FCC for commercial wireless services could provide the needed millimeter wavelength carrier bandwidth. An attractive choice for the subscriber links is the LMDS band. The airplane's fuselage can house switching circuitry and fast digital network functions. An MMW antenna array and its related components will be located in a pod suspended below the aircraft fuselage. The antenna array will produce many beams, typically more than 100. Adjacent beams will be separated in frequency. Electronic beamforming techniques can be used to stabilize the beams on the ground, as the airplane flies within its station keeping volume. For the alternative of aircraft-fixed beams, the beams will traverse over a user location, while the airplane maintains station overhead, and the virtual path will be changed to accomplish the beam-to-beam handoff. For each isolated city to be served, a fleet of three aircraft will be operated in shifts to achieve around-the-clock service. In deployments where multiple cities will be served from a common primary flight base, the fleet will be sized for allocating, on average, two aircraft per city to be served. Flight operational tactics will be steadily evolved and refined to achieve continuous presence of the node above each city. Many services will be provided, including but not limited to T1 access, ISDN access, Web browsing, high-resolution videoconferencing, large file transfers, and Ethernet LAN bridging.

The Object-Oriented Design and Performance of JAWS: A High-performance Web Server Optimized for

This paper provides two contributions to the study of high-performance object-oriented (OO) Web servers. First, it outlines the design principles and optimizations necessary to develop efficient and scalable Web servers and illustrates how we have applied these principles and optimizations to create JAWS. JAWS is a high-performance Web server that is designed to alleviate overheads incurred by existing Web servers on high-speed networks. In addition to its highly extensible OO design, it is also highly efficient, consistently outperforming existing Web servers, such as Apache, Java Server, PHTTPD, Zeus, and Netscape Enterprise, over 155 Mbps ATM networks on UNIX platforms. Second, this paper describes how we have customized the JAWS OO design to leverage advanced features of Windows NT on multi-processor platforms linked by high-speed ATM networks. The Windows NT features used in JAWS include asynchronous mechanisms for connection establishment and data transfer. Our previous benchmarking studies demonstrate that once the overhead of disk I/O is reduced to a negligible constant factor (e.g., via memory caches), the primary determinants of Web server performance are its concurrency and event dispatching strategies. Our performance results over a 155 Mbps ATM network indicate that certain Windows NT asynchronous I/O mechanisms (i.e., TransmitFile) provide superior performance for large file transfers compared with conventional synchronous multi-threaded servers. Conversely, synchronous event dispatching performed better for files less than 50 Kbytes. Thus, to provide optimal performance, a Web server design should be adaptive, i.e., choosing to use different mechanisms (such as TransmitFile) to handle requests for large files, while using alternative I/O mechanisms (such as synchronous event dispatching) on requests for small files.

The implementation and evaluation of the use of CORBA in an engineering design application

Many computer applications today require some form of distributed computing to allow different software components to communicate. Several different commercial products now exist based on the Common Object Request Broker Architecture (CORBA) of the Object Management Group. The use of such tools, however, often requires the modification of existing systems, rather than the development of new applications. The objective of this research has been to integrate the use of a CORBA tool into an existing engineering design application for the purpose of (1) evaluating the amount of re-engineering that is involved to effectively integrate distributed object computing into an existing application, and (2) evaluating the use and performance of distributed object computing in an engineering domain, which often requires the transfer of large amounts of information. The results of this work demonstrate that CORBA technology can be easily integrated into existing applications. The ease of the integration as well as the efficiency of the resulting system, however, depends upon the degree of modification that developers are willing to consider in the re-engineering process. The most transparent approach to the use of CORBA requires less modification and generally produces less efficient performance. The less transparent approach to the use of CORBA can potentially require significant system modification but produce greater performance gains. This work outlines issues that must be considered for the partitioning of functionality between the client and the server, development of an IDL interface, development of client and server-side wrappers, and support for concurrent, multi-user access. In addition, this work also provides performance and implementation comparisons of different techniques for the use of wrappers and for the transfer of large data files between the client and the server. Performance comparisons for the incorporation of concurrent access are also presented. Copyright © 1999 John Wiley & Sons, Ltd.

Multiple paths protocol for a cluster type network

Nowadays we can access the Internet by carrying a portable computer and using wireless communication. The wireless network with personal handy phone system (PHS) and portable cellular telephone has only rates of tens of Kbps to a few Mbps. Compared with the cable network, the transfer rate cannot generally satisfy a highly developed communication service such as large file transfer and real-time communications. This paper proposes a protocol, SHAKE, for sharing multiple paths in a cluster type network that is a kind of LAN in which some mobile hosts temporarily connect mutually. SHAKE provides the functions for composing a cluster type network, and dispersing traffic efficiently by measuring transfer rate and round-trip time. As a mobile host has only a low transfer capacity for individuals to communicate with the outside, if whole capacities of other hosts which compose a cluster type network are shared, we can get larger transfer capacity and satisfy the required communication services. Copyright © 1999 John Wiley & Sons, Ltd.

Transport protocols for Internet-compatible satellite networks

We address the question of how well end-to-end transport connections perform in a satellite environment composed of one or more satellites in geostationary orbit (GEO) or low-altitude Earth orbit (LEO), in which the connection may traverse a portion of the wired Internet. We first summarize the various ways in which latency and asymmetry can impair the performance of the Internet's transmission control protocol (TCP), and discuss extensions to standard TCP that alleviate some of these performance problems. Through analysis, simulation, and experiments, we quantify the performance of state-of-the-art TCP implementations in a satellite environment. A key part of the experimental method is the use of traffic models empirically derived from Internet traffic traces. We identify those TCP implementations that can be expected to perform reasonably well, and those that can suffer serious performance degradation. An important result is that, even with the best satellite-optimized TCP implementations, moderate levels of congestion in the wide-area Internet can seriously degrade performance for satellite connections. For scenarios in which TCP performance is poor, we investigate the potential improvement of using a satellite gateway, proxy, or Web cache to "split" transport connections in a manner transparent to end users. Finally, we describe a new transport protocol for use internally within a satellite network or as part of a split connection. This protocol, which we call the satellite transport protocol (STP), is optimized for challenging network impairments such as high latency, asymmetry, and high error rates. Among its chief benefits are up to an order of magnitude reduction in the bandwidth used in the reverse path, as compared to standard TCP, when conducting large file transfers. This is a particularly important attribute for the kind of asymmetric connectivity likely to dominate satellite-based Internet access.

Performance study of buffer management schemes under multicasting traffic in ATM switching nodes

Multimedia applications in broadband ISDN present various types and classes of traffics such as video teleconferencing, broadband telephony and large file transfers. Multicasting is important in multimedia applications. Since cell loss in a multicast packet will cause the retransmission of the entire packet causing excessive load in the network, the ATM switch should provide for the reduction of cell loss of multicast traffic. The reduction in cell loss of a particular type of traffic should be done in coordination with the QoS parameters required from other classes within each traffic type. Buffering schemes are used in ATM switches to achieve the required levels of QoS. In this paper, we develop a new buffering scheme which considers traffic types that include multicasting and nonmulticasting components with two classes of cells within each traffic type. We compare the new buffering scheme with two existing schemes, No Priority and Push Out. We show that the new buffering scheme performs better than the two other schemes by reducing cell loss of multicast cells. We also introduce a new cell scheduling policy that produces lower delay for the delay sensitive class. The new buffering and scheduling schemes provide the ability to tune the cell loss and delay of each class of cells within each traffic type.

Routing high-bandwidth traffic in max-min fair share networks

We study how to improve the throughput of high-bandwidth traffic such as large file transfers in a network where resources are fairly shared among connections. While it is possible to devise priority or reservation-based schemes that give high-bandwidth traffic preferential treatment at the expense of other connections, we focus on the use of routing algorithms that improve resource allocation while maintaining max-min fair share semantics. In our approach, routing is closely coupled with congestion control in the sense that congestion information, such as the rates allocated to existing connections, is used by the routing algorithm. To reduce the amount of routing information that must be distributed, an abstraction of the congestion information is introduced. Using an extensive set of simulation, we identify a link-cost or cost metric for "shortest-path" routing that performs uniformly better than the minimal-hop routing and shortest-widest path routing algorithms. To further improve throughput without reducing the fair share of single-path connections, we propose a novel prioritized multi-path routing algorithm in which low priority paths share the bandwidth left unused by higher priority paths. This leads to a conservative extension of max-min fairness called prioritized multi-level max-min fairness. Simulation results confirm the advantages of our multi-path routing algorithm.

A novel approach to estimating the cell loss probability in an ATM multiplexer loaded with homogeneous on-off sources

Estimating the cell loss probability in an ATM multiplexer is one of the most important problems concerning congestion control and bandwidth management in an ATM-based BISDN. We propose a new approach to estimating the cell loss probability in an ATM multiplexer. We use the Markov modulated deterministic process (MMDP) to approximate the actual arrival process and then model the ATM multiplexer as an MMDP/D/1/K queueing system. Using queueing analysis, we derive a formula for the cell loss probability expressed in terms of the limiting probabilities of a Markov chain. We propose two approximation methods based on the results of the analysis. The actual arrival process is approximated by an (M+1)-state MMDP in the first method and by a two-state MMDP in the second. The major advantages of both methods are simplicity, computational efficiency, and numerical stability. The most attractive feature of the second method is that the cell loss probability can be expressed in closed form. Numerical and simulation results show that the first method is sufficiently accurate for all cases in which burst-level congestion is the main contributing factor to cell loss, while the closed-form formula is sufficiently accurate for applications where the average burst length is large (such as large file transfers, image retrievals, etc.).< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Hybrid ARQ schemes employing coded modulation and sequence combining

In this paper, we propose and analyze two hybrid automatic-repeat-request (ARQ) schemes employing bandwidth efficient coded modulation and coded sequence combining. In the first scheme, a trellis-coded modulation (TCM) is used to control channel noise; while in the second scheme a concatenated coded modulation is employed. The concatenated coded modulation is formed by cascading a ReedSolomon (RS) outer code and a block coded modulation (BCM) inner code. In both schemes, the coded modulation decoder, by performing sequence com- bining and soft-decision maximum likelihood decoding, makes full use of the information available in all received sequences corresponding to a given information message. It is shown, by means of analysis as well as computer simulations, that both schemes are capable of providing high throughput efficiencies over a wide range of signal-to-noise ratios. The schemes are suitable for large file transfers over satellite communication links where high throughput and high reliability are required.

Distributed Bandwidth Management in FDDI Networks with Multiple Isochronous Services

The Fiber Distributed Data Interface (FDDI) is a proposed American National Standard for a 100 Mbps token ring using an optical fiber medium. The timed token protocol used in FDDI allows the integration of digital voice, video, interactive data communications and large file transfers over the same network. The performance of the protocol has been studied in the past for asynchronous class of service with a single isochronous service.In this paper, we focus on the performance of the protocol when multiple class of isochronous services are to be provided along with the asynchronous class of service. It is found that the performance of the protocol is affected by the different delay requirements of different isochronous services. The performance of the protocol is studied through simulation with two classes of isochronous services viz video and voice. We also propose two distributed algorithms to improve the performance of the protocol under multiple isochronous services.

DQDB networks with and without bandwidth balancing

It is explained why long distributed queue dual bus (DQDB) networks without bandwidth balancing can have fairness problems when several nodes are performing large file transfers. The problems arise because the network control information is subject to propagation delays that are much longer than the transmission time of a data segment. Bandwidth balancing is then presented as a simple solution. By constraining each node to take only a certain fraction of the transmission opportunities offered to it by the basic DQDB protocol, bandwidth balancing gradually achieves a fair allocation of bandwidth among simultaneous file transfers. Two ways to extend this procedure effectively to multipriority traffic are proposed.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Distributed source control: a network access control for integrated broadband packet networks

The authors propose a congestion control strategy called distributed source control (DSC), designed to address the integration of diverse traffic types in broadband packet networks. DSC is a rate-based network access control implemented at the asynchronous transfer mode (ATM) layer as a feedforward control, in conjunction with an adaptive end-to-end control between network edges. The authors examine the performance of high-speed data traffic such as large file transfers, still images, and document retrievals. The authors study DSC's integrability with real-time traffic such as voice and video. With the help of a simulation model of a broadband packet network (150 Mb/s), the authors quantify the improvement in network performance due to DSC. An analytic model for an access node under DSC is developed, and guidelines for sizing of buffers in the switch and for choosing the end-to-end window size are provided. >

A concatenated coded modulation scheme for error control

A concatenated coded modulation scheme is presented for error control in data communications. The scheme is achieved by concatenating a Reed-Solomon outer code and a bandwidth efficient block inner code for M-ary phase-shift keying (PSK) modulation. Error performance of the scheme is analyzed for an additive white Gaussian noise (AWGN) channel. It is shown that extremely high reliability can be attained by using a simple M-ary PSK modulation inner-code and a relatively powerful Reed-Solomon outer code. Furthermore, if an inner code of high effective rate is used, the bandwidth expansion required by the scheme due to coding will be greatly reduced. The scheme is particularly effective for high-speed satellite communications for large file transfer where high reliability is required. A simple method is also presented for constructing block codes for M-ary PSK modulation. Soome short M-ary PSK codes with good minimum squared Euclidean distance are constructed. These codes have trellis structure and hence can be decoded with a soft-decision Viterbi decoding algorithm. Furthermore, some of these codes are phase invariant under multiples of 45 degrees rotation. >

Dynamic bandwidth allocation technique for high speed LANs

A modified time division multiplexing scheme using a slotted ring local area network combined with token passing is proposed. The hybrid scheme allows for multiple classes of data to be transmitted over a single optical fibre simultaneously. The bandwidth allocation can be changed dynamically, thereby accommodating real-time traffic as well as large file transfers. Simulation experiments have been carried out to demonstrate the feasibility of such a scheme.

MBRAM-a priority protocol for PC based local area networks

A protocol called modified broadcast access method (MBRAM) is presented that is based on the broadcast recognizing access method (BRAM). The goal was to develop a protocol that would support large file transfers from priority nodes (such as image workstations) and still provide reasonable access for nonpriority nodes for normal traffic, using standard Ethernet hardware. Initial results and future-simulation and experiments on a prototype testbed are reported, showing that MBRAM meets the requirements.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

Decomposition approaches to modelling LAN contention and host computer performance

Networks consisting of special-purpose workstations connected by local area networks have become very popular. Their use may grow to the point where LAN contention degrades the performance of the application, e.g. in graphics workstations whose images must be updated frequently or in applications involving large file transfers. Methods of predicting the level of LAN contention and its effect on host performance are discussed, and the way in which this problem affects token rings and CSMA-type networks is examined.

Some observations on the performance of a 56 Kbit internet link

A 56 Kb gateway between two Ethernets was installed and subjected to heavy traffic loads. While large file transfers were being done between the networks, less demanding internetwork commands were issued and the response time measured. The effect of the heavy network load on the smaller tasks is presented in this report. It appears that the smaller tasks are unfairly penalized for the network load caused by the larger tasks. If so, this could be a problem for the design and use of the backbone network of the NSFnet. No conclusions are drawn concerning the applicability of these results in a general case. The data is presented for consideration by those who are planning to rely on such links and have some knowledge of the level of traffic that can be expected in their own environment.

A Broadcast Protocol for File Transfers to Multiple Sites

A retransmission Protocol for a broadcast connection (point-to-multipoint) is proposed and its performance characteristics are considered. The protocol is designed for transfers of large files Over a satellite channel that is time-shared to carry both the data from the broadcasting transmitter and the set of acknowledgments from the multiple receiver sites. A mathematical model of the transmission system that includes separate error processes for uplink and downlink errors on data transmissions, and similar processes for errors on the acknowledgment frames as well, is used to analyze the performance of the scheme. Exact analytical expressions for the relative throughput of the channel are obtained for two special cases: 1) the uplink is error-free; and 2) the acknowledgments are error-free. For the general case, upper and lower bounds are derived and are shown to be virtually indistinguishable for many practical sets of system parameters. The results demonstrate that the broadcasting of large files to multiple receivers can be done both efficiently and reliably.

Design Choices for Selective-Repeat Retransmission Protocols

The design of selective-repeat retransmission protocols to provide reliable transfer of large data files is discussed. The transmission system is assumed capable of losing, distorting, or reordering the transmitted data frames and acknowledgment frames. The principle design considerations center on methods for managing the finite sequence number space and methods for managing the finite-sized receiver buffer. Under the assumption that there exists a maximum time from transmission of a frame to receipt of an acknowledgment for that frame, methods are described for safe reuse of frame sequence numbers. Previous results by Metzner and Morgan on buffer overflow control are extended to cover non-FIFO transmission systems. A two-dimensional space of design parameters is explored by way of Monte Carlo simulations. For appropriate designs, it is shown that significant increases in throughput over that achieved by the Go-Back- N protocol are possible on high-rate long-delay channels.