| D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE JSAC, vol. 11, no.4, May 1993, pp. 489--506. |
....the issue of programmable transport, but do not consider issues of QoS. Reference [10, 11] addresses both QoS and programmability issues; 12] presents the x kernel, an operating system environment that provides an explicit architecture for constructing and composing network protocols. Reference [13] also provides a taxonomy of key transport system services and illustrates the concepts with a survey of four operating system transport architectures, namely, System V and BSD UNIX, the x kernel, and Choices. Finally, 14] addresses many of the issues related to the implementation of protocol ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE JSAC, vol. 11, no.4, May 1993, pp. 489--506.
....by N A. While multimedia middleware is intended to support a wide range of multimedia applications, flexible middleware introduces an orthogonal concept for communications to support adjustable protocol processing for high performance applications and high speed networks, as done within ADAPTIVE [9] or FCSS [10] Generally spoken, to facilitate a flexible approach requires to structure protocols in a modular fashion, where separate building blocks can inter operate efficiently. Da CaPo offers a set of protocol functions implemented in terms of software modules that run in an efficient ....
D. Schmidt and T. Suda, "Transport system architecture services for highperformance communication subsystems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....codesign. Our work relates to and builds upon the following areas of research. Communication subsystem design and performance: Several researchers have studied the issues affecting the design and performance of network adapters [1, 4, 17, 18] and communication subsystems in general [19, 20]. While many of these studies have influenced our work, we have focussed as much on the design process, as the design itself. Simulation based evaluation: Performance evaluation via simulation can be conducted at various levels of detail, and hence, accuracy. Recently, significant attention has ....
D. C. Schmidt and T. Suda, "Transport system architecture services for high-performance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....the CPU and link and any coupling between CPU and link bandwidth allocation. In a recent paper [5] we have studied the extent to which these factors affect admissibility in the context of real time channels. 3 A QoS Sensitive Communication Subsystem Architecture In the process per message model [24], a process or thread shepherds a message through the protocol stack. Besides eliminating extraneous context switches encountered in the process per protocol model [24] it also facilitates protocol processing to be scheduled according to a variety of policies, as opposed to the software interrupt ....
....in the context of real time channels. 3 A QoS Sensitive Communication Subsystem Architecture In the process per message model [24] a process or thread shepherds a message through the protocol stack. Besides eliminating extraneous context switches encountered in the process per protocol model [24], it also facilitates protocol processing to be scheduled according to a variety of policies, as opposed to the software interrupt level processing in BSD Unix. However, the process per message model introduces additional complexity for supporting per channel QoS guarantees. Creating a distinct ....
[Article contains additional citation context not shown here]
D. C. Schmidt and T. Suda, "Transport system architecture services for high-performance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....asynchronous timeouts can be significantly reduced. Process architecture Subsystems dictate how protocol entities like messages, layers, connections, buffers, and timers are bound to the underlying unit of scheduling. The manner in which they are bound is called a subsystem s process architecture [19]. HotLava defines a vertical process architecture in which packets are escorted through the protocol graph via non preemptable threads. When an application wishes to send data, a buffer is created and a thread is attached to it to escort it through the protocol graph in much the same manner as the ....
Douglas C. Schmidt and Tatsuya Suda. Transport system architecture services for high-performance communications systems. IEEE Journal on Selected Areas in Communications, 11(4):489--506, May 1993.
....above functionality. It is clear that how this functionality is implemented in a particular workstation depends to a great extent on the transport system infrastructure which encompasses the application interface, network protocols, and particular operating system (OS) framework which support them [13]. In addition, we also need to consider presentation and application layer processing, such as the synchronization and playout operations described above, and the OS support and scheduling for these operations. In our simulation, however, we do not attempt to emulate any particular implementation ....
....described above, and the OS support and scheduling for these operations. In our simulation, however, we do not attempt to emulate any particular implementation of a OS and transport system. It may, however, be helpful to understand it in the context of the BSD UNIX OS and transport system [13]. The application interface, such as the socket layer, is not explicitly modeled but costs such as user to kernel copies at this interface are included as part of the per byte protocol processing costs (see section 3) The network protocol processing is modeled as a single monolithic unit of ....
Douglas C. Schmidt and Tatsuya Suda. Transport system architecture services for highperformance communications systems. IEEE J.Select.Areas Commun., pages 489--506, May 1993.
....the communication subsystem must exercise fine grain control over management of resources involved in communication, such as processors and the network adapter. Within the communication subsystem, the degree of sharing of processors amongst connections depends on the transport system architecture [3] and resource management policies employed. Resources such as processors and buffers must be allocated to individual connections in order to meet their QoS requirements. This allocation must attempt to reduce load imbalance, achieve high resource utilization, and maximize the number of connections ....
....multiple protocol processors facilitates scalable server design by increasing the processing capacity of the communication subsystem and allowing concurrent handling of different connections. Protocol processing is based on a vertical process architecture employing the process per connection model [3]. Each connection has associated with it a unique process (the connection handler ) and a first in first out (FIFO) queue of messages waiting to be processed and transmitted; the connection handler processes the queued messages one at a time. Maintaining a FIFO queue of messages per connection ....
[Article contains additional citation context not shown here]
D. C. Schmidt and T. Suda, "Transport system architecture services for high-performance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....various operating system resource management functions such as timer services, buffer management, and device management (i.e. interrupt servicing and low level device operations) as shown in Figure 2.2. Depending upon the protocol architecture, kernel architecture, and process architecture [152], other resource management functions such as process management and scheduling may also occur during data transfer. On multiprocessor hosts, processor allocation may also occur in addition to the above. An extensive survey of transport system architecture services and implementation strategies in ....
....data transfer. On multiprocessor hosts, processor allocation may also occur in addition to the above. An extensive survey of transport system architecture services and implementation strategies in a variety of operating systems such as System V UNIX, BSD UNIX, x kernel, and Choices is presented in [152]. 20 Network Buffer management Resource Management API Network Adapter Protocol Stack Application tasks Timer management Process management Scheduling Processor allocation Device management Figure 2.2: Interaction between protocols and resource management functions 2.3 Factors ....
[Article contains additional citation context not shown here]
D. C. Schmidt and T. Suda, "Transport system architecture services for highperformance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....and are only discernible by examining sample code and or supporting documentation. Process Architecture: The manner in which protocol entities, like messages, layers, connections, buffers, and timers, are bound to the underlying unit of scheduling is often referred to as a process architecture [9]. Process architecture differences can impose serious portability and performance problems. Protocol Graph: Variations in how the protocol graph is constructed and accessed may also pose portability and performance problems. Buffers: Because communication protocols require memory buffers in ....
Douglas C. Schmidt and Tatsuya Suda. Transport system architecture services for high-performance communications systems. IEEE Journal on Selected Areas in Communications, 11(4):489--506, May 1993.
....reinventing the services (e.g. integrity checking) that they require. Services similar to those provided in the Transport layer are also reused in lower layers, e.g. routing control messages need integrity. The demonstration focuses on modularity and dismisses other transport system design issues [27, 28] with simplistic kludges: Implementation is entirely in software, as a single process in the user address space, and the configuration is coded into the organs. The only stimuli to the process are the expiration of a timer, and arrival of a message from the channel. The channels extend between ....
D. C. Schmidt and T. Suda: "Transport system architecture services for high-performance communications systems", IEEE J. Sel. Areas in Comm., 11(4):489-505, May 1993
....captures significantly more details of adapter design and interaction with the protocol stack. Communication subsystem design and performance: Several researchers have studied the issues affecting the design and performance of network adapters [2, 16 18] and communication subsystems in general [19, 20]. All of these efforts are geared towards best effort adapters with the primary goal of maximizing data transfer throughput. In contrast, we focus on adapter design to provide QoS guarantees to applications. Modeling of real time operating systems: Several researchers have developed accurate ....
D. C. Schmidt and T. Suda, "Transport system architecture services for high-performance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489--506, May 1993.
....of such interfaces are opening and closing a session, and sending and receiving data. The protocol model also defines the manner in which protocol entities like messages, layers, connections, and buffers are bound to the underlying unit of scheduling, called a subsystem s process architecture [26]. Consequently, protocol encapsulation modules must also translate between the respective process architectures. Example process architectures include horizontal (scheduling protocol layers) and vertical (scheduling connections or messages) The protocol model also defines protocol graph ....
Douglas C. Schmidt and Tatsuya Suda. Transport system architecture services for high-performance communications systems. IEEE Journal on Selected Areas in Communications, 11(4):489--506, May 1993.
.... Research has indicated that performance bottlenecks are shifting from the network infrastructure to the transport system infrastructure resulting in a throughput preservation problem, where only a limited fraction of the available bandwidth is actually delivered to applications (see Schmidt[12]) One of the significant contributors to this problem is the excessive memory to memory copying that takes place amongst the various layers within a networking protocol suite. For example, in BSD UNIX, a piece of data is read and written at least twice in its entirety while traversing from the ....
D.C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems" IEEE J. Select. Areas Commun., May 1993.
....embedded in the multimedia mechanism layer, and the QoS description is done in DSL (Device Specification Language) language at the application layer. The QoS description applies to the DSL streams. Most architectures at the end points consider the QoS handling only in transport subsystem layers [15], 16] 9] 13] Current research in [17] 2] also goes in the direction of a provision of a QoS architecture which considers different protocol functions for QoS guarantees in different layers. In the Lancaster QoS architecture [17] the application layer includes scheduling in order to satisfy ....
D.C.Schmidt, Tatsuya Suda, "Transport System Architecture Services for High-Performance Communications Systems", IEEE Journal on SAC, May 1993, Vol.11, Nr.4, pp.489-506
....protocol stack layer, responsible for all protocol processing for that layer. In some cases, one process may perform both send and receive processing, while in other cases, transmission and reception may be handled by separate processes. This so called horizontal or process per layer architecture [18] is not only inefficient, it also makes protocol processing unpredictable. The inefficiency arises from having to copy data across address spaces and having to switch contexts between processes. The unpredictability arises because once a process starts processing a low priority (incoming or ....
....This leads to priority inversion [4] a high priority message being delayed by a low priority one. Various techniques have been proposed to make protocol processing efficient and predictable as described next. 2. 1 Increasing Efficiency The vertical or process per message architecture [18, 19] circumvents many of the problems associated with horizontal architectures. In this scheme, the kernel maintains a pool of threads (lightweight processes) One thread is assigned to each outgoing and incoming message. This thread shepherds the message through the entire protocol stack. As a ....
D. C. Schmidt and T. Suda, "Transport system architecture services for high-performance communications systems," IEEE Journal on Selected Areas in Communications, vol. 11, no. 4, pp. 489-- 506, May 1993.
....(e.g. the transport layer, network layer, data link layer, etc. The ASX framework incorporates concepts from several other modular communication frameworks including System V STREAMS [17] the x kernel [13] and the Conduit [18] a survey of these and other communication frameworks appears in [19]) These frameworks all contain features that support the flexible configuration of communication subsystems by inter connecting building block protocol components. These frameworks encourage the development of standard reusable protocol components by decoupling protocolspecific processing ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
.... Reactor APP SPECIFIC SERVICES Concurrency global IPC SAP Figure 1: Class Categories in the ASX Framework other modular communication frameworks including System V STREAMS [14] the x kernel [15] and the Conduit [9] a survey of these and other communication frameworks appears in [16]) These frameworks all contain features that support the flexible configuration of communication subsystems by inter connecting building block protocol and service components. In general, these frameworks encourage the development of standard reusable communication related components by ....
.... and presentation conversions) and operating system mechanisms (such as process management, asynchronous event invocation, message buffering, and layer to layer flow control) that support the implementation and execution of protocol stacks that contain hierarchically related protocol functions [16]. Advances in VLSI and fiber optic technology are shifting performance bottlenecks from the underlying networks to the communication subsystem [25] Designing and implementing multi processor based communication subsystems that execute protocol functions and OS mechanisms in parallel is a ....
[Article contains additional citation context not shown here]
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
No context found.
D. C. SCHMIDT and T. SUDA, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, 11, pp. 489--506 (May 1993).
....Furthermore, the ACE Reactor is typically used to develop distributed applications. The major sources of overhead in distributed systems result from activities like caching, latency, network host interface hardware, presentation level formatting, memory to memory copying, and process management [21]. Therefore, the additional indirection caused by dynamic binding is typically insignificant in comparison [22] In addition, good C compilers can optimize virtual method overhead away completely via the use of adjustor thunks [23] 12 5 Design Rules for Using the Reactor Effectively The ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
No context found.
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
....systems. A careful examination of the major sources of overhead in distributed systems reveals that most performance bottlenecks result from activities such as caching, latency, network host interface hardware, presentation level formatting, memory to memory copying, and process management [12]. Therefore, the additional memory reference overhead caused by dynamic binding is insignificant in comparison [13] To justify these claims empirically, an upcoming article in the C Report will present the results of a benchmarking experiment that measures the performance of IPC SAP and the ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
....interface. The ASX framework incorporates concepts from several modular communication frameworks including System V STREAMS [39] the x kernel [40] and the Conduit framework [41] from the Choices object oriented operating system (a survey of these and other communication frameworks appears in [42]) These frameworks all contain features that support the flexible configuration of communication subsystems by inter connecting building block protocol and service components. In general, these frameworks encourage the development of standard communication related components (such as message ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
No context found.
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
....and (4) Message Parallelism, which associates a process per message. The ADAPTIVE framework also facilitates experimentation with the various process architecture alternatives. Several studies have compared the advantages and disadvantages of these process architectures via qualitative analysis [2, 12]. However, few studies have quantitatively compared the performance of the alternative process architectures via controlled, empirical experimentation. In particular, existing research that measures the performance of process architectures focuses on only one or two approaches [7, 9, 13, 14] ....
....network environment (such as high speed vs. low speed and large frame size vs. small frame size) This section outlines the distinguishing features of four process architectures supported by ADAPTIVE. These process architectures fall into three general categories: horizontal, vertical, and hybrid [12]. Although each process architecture has different structural and performance characteristics, it is possible to implement the same protocol family functionality (such as the OSI, TCP IP, and F CSS [16] with any approach. 4.1 Horizontal Process Architectures Horizontal process architectures ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
....interface. The ASX framework incorporates concepts from several modular communication frameworks including System V STREAMS [35] the x kernel [36] and the Conduit framework [14] from the Choices object oriented operating system (a survey of these and other communication frameworks appears in [37]) These frameworks all contain features that support the flexible configuration of communication subsystems by inter connecting building block protocol and service components. In general, these frameworks encourage the development of standard communication related components (such as message ....
D. C. Schmidt and T. Suda, "Transport System Architecture Services for High-Performance Communications Systems," IEEE Journal on Selected Areas in Communication, vol. 11, pp. 489--506, May 1993.
First 50 documents
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC