Martina Zitterbart

Analysis of TCP/IP for high performance parallel implementations

An analysis of potential parallelism in TCP/IP is presented for two purposes. First, it is used as a basis for an implementation on a multiprocessing platform. Second, it reveals the areas where protocol modifications and/or tuning would yield more efficient implementations on multiprocessors. The focus is on the analysis, and detailed dependence graphs among the atomic TCP/IP protocol functions are presented. No specific implementation design is shown, since this highly depends on the particular hardware chosen.<<ETX>>

On the Design of a Multigigabit IP Router

The emergence of gigabit speed networks hinges upon the existence of high performance internetworking units, such as IP routers. In this paper, we present an architecture and we discuss the implementation of a multigigabit IP router. For this implementation, two special purpose VLSI chips are required; the rest can be built using off-the-shelf components. IP header processing of received packets is handled by a specialized chip. Memory management, another well-known performance bottleneck, is simplified and efficiently implemented using special VLSI support. Searching for the next hop address in the routing table, which is the major contributor to the delay in traditional IP protocol implementations, is significantly reduced by using a special configuration of Content Addressable Memories (CAMs).

Parallel TCP for high performance communication subsystems

Emerging applications and rapid advances in communication technology increasingly require high-performance communication subsystems. Parallelizing protocols in order to efficiently make use of multiprocessor platforms is one approach such systems. A function-based parallel structure of the widely used TCP protocol, applicable to a large variety of implementation platforms, is presented. An example of a multiprocessor implementation of the parallel TCP is discussed. General-purpose processors as well as specialized VLSI implementations of performance critical functions are considered as a basis for high-performance communication subsystems being able to sustain gigabit rates.<<ETX>>

A scheme for high-performance LAN interconnection across public MAN's

The proliferation of local area networks (LANs) and the emergence of high-bandwidth public networks will create an environment in which it will become increasingly simple and attractive to communicate among different organizations across a common public networking infrastructure. The open bridging scheme proposed, which employs inherent features of forthcoming cell-based metropolitan area networks (MANs) in order to achieve efficient interenterprise communications is introduced. It uses enhanced remote LAN bridging techniques to ensure router-level functionality with bridge-level simplicity and performance. The practicality of open bridging is shown by describing how it can be incorporated in networks using current standard protocol stacks, such as open systems interconnnection (OSI) and transmission control protocol/internet protocol (TCP/IP). >

A system framework for open distributed processing

Emerging distributed applications increasingly require adequate tools and techniques for system- and application-level management. The integration of both aspects in an overall system framework is an important issue. This paper presents such a framework supporting advanced distributed applications in the context of the evolving Open Distributed Processing reference model. For system-level communication, an efficient subsystem providing advanced service capabilities is presented to cope with the increasing diversity of application service requirements. For application-level processing, a distributed object-based environment is offered. It implements location-independent invocation and object mobility and provides a high level of distribution transparency. The approach is augmented with tools and techniques for managing an overall application configuration.

A scheme for remote LAN bridging across SMDS MANs

A remote bridging scheme for the interconnection of local area networks (LANs) across cell-based public metropolitan area networks (MANs), such as SMDS (switched megabit data service) networks, is briefly described. The scheme makes use of some characteristics of cell-based networks such as the hierarchical E.164 addressing scheme, which makes routing at the MAC level through the public environment possible. This leads to a less complex and more efficient bridging scheme called open bridging. In contrast with standard media access control (MAC) bridging approaches, open bridging emphasizes the notion of independent domains to reflect the practical aspects of interconnecting unrelated private bridged LANs across a public network. Each domain is allowed to independently manage and operate its internal connectivity. Full support for current bridging standards is provided within each domain.<<ETX>>

Internetworking across public ATM networks

The emergence of public ATM networks, such as B-ISDN and SMDS, is considered to have a high impact on future internetworking environments. This paper presents a new bridging architecture designed for that purpose and discusses its feasibility in TCP/IP networks. The new architecture, called Open Bridging, uses special capabilities of public ATM networks, such as the hierarchical addressing mechanism and the routing support inside the network, to provide for efficient interconnection of remote LANs across public ATM networks. The operation of Open Bridges does not interfere with current TCP/IP protocols. However, the efficiency of Open Bridging can be obtained in the interconnection of TCP/IP networks if a slight extension to the address resolution protocol (ARP) is implemented to push the routing of data packets down to the MAC level. Moreover, the architecture of Open Brouters is also proposed to enable full interoperability among domains using either Open Bridging or IP routing.<<ETX>>

Transport service and protocols for high-speed networks

Several physical communication networks currently available can offer very high bandwidth. At the same time, many potential applications are in need for all that bandwidth. Unfortunately, the communication subsystems used today do not preserve the raw network throughput up to the application level. In this paper, we give an overview of the major current trends and approaches proposed to solve the so-called transport bottleneck problem. We discuss issues related to protocol design and implementation as well as the provision of new features at the transport service interface to accommodate emerging applications.