Search

Network on-Chip (NoC) is an interconnect fabric to connect sub-system blocks on a chip. The NoC should provide high bandwidth and low latency, should consume low energy, and should be compact. However, all these requirements are at odds and require tradeoffs at all levels. In this chapter, we discuss issues and challenges for future NoCs with demands for high bandwidth and low energy. Next, we present details of how coupling packet-switched arbitration with circuit-switched data transfer can achieve these goals. In this hybrid network, packet-switched arbitration is used to reserve future circuit-switched channels for the data transfer, eliminating the performance bottlenecks associated with pure circuit-switched networks while maintaining their power advantage. Furthermore, proximity-based data streaming increases network throughput and improves energy efficiency. Measurements of this NoC in 45 nm CMOS are described to analyze de-sign tradeoffs.

Computing the blocking probability of bursts at the core nodes in an OBS network is important for many reasons. First, it helps to understand the loss process of bursts and provide exact analytical expression for the loss rate of bursts at the core nodes. Second, it helps to give bounds for the loss rate which is an important measure of QoS in OBS network. Finally, it helps to design new loss minimization mechanisms at various places in the network in order to improve the performance of the network at higher layers. Computation of blocking probability in OBS networks is challenging due to the lack of optical buffers at the core nodes and the unacknowledged nature of switching. Contention losses which are different from the traditional losses due to buffer overflow (congestion) lead to an additional complexity in formulating analytical expressions for blocking probability of bursts. In the circuit switching, arriving connection requests are blocked at the source due to lack of bandwidth. In packet switching, packets are buffered at the intermediate nodes if bandwidth is not available. Packets are dropped only when the buffer overflows. Unlike these two switching paradigms, which are analyzed with traditional queueing systems, losses in OBS network occur whenever two or more bursts request the same wavelength for the same duration. These losses are completely random and do not necessarily indicate congestion in the network. To a certain extent such contention is resolved either in the space domain by switching onto an alternate wavelength (wavelength conversion) or in the time domain by delaying the optical burst (using FDLs). When both these techniques cannot resolve the contention for a wavelength, bursts are dropped. The probability of bursts getting dropped due to contention at a node is commonly referred to as the blocking probability or the BLP.

The idea behind the proposition of Networks-on-Chip (NoCs) for modern and future systems on chip capitalizes on the fact that busses do not scale well when shared by a large number of cores. Even if NoC research is a relatively young field, the literature abounds with propositions of NoC architectures. Several of these propositions claim providing quality of service (QoS) guarantees, which is essential for real time and multimedia applications. The most widespread approach to attain some degree of QoS guarantee relies on a two-step process. The first step is to characterize application performance through traffic modeling and simulation. The second step consists in tuning a given network template to achieve some degree of QoS guarantee. These QoS targeted NoC templates usually provide specialized structures to allow either the creation of connections (circuit switching) or the assignment of priorities to connectionless flows. It is possible to identify three drawbacks in this two-step process approach. First, it is not possible to guarantee QoS for new applications expected to run on the system, if those are defined after the network design phase. Second, even with end-to-end delay guarantees, connectionless approaches may introduce jitter. Third, to model traffic precisely for a complex application is a very hard task. If this problem is tackled by oversimplifying the modeling phase, errors may arise, leading to NoC parameterization that is poorly adapted to achieve the required QoS. This Chapter has two main objectives. The first one is to evaluate the area-performance trade-off and the limitations of circuit switching and priority scheduling to meet QoS. This evaluation will show where such implementations are really suited for QoS, and when more elaborate mechanisms to meet QoS are needed. The second objective comprises proposing a method, called rate-based scheduling, to approach QoS requirements considering the execution time state of the NoC. The evaluation of circuit switching and priority scheduling show that: (i) circuit switching can guarantee QoS only to a small number of flows; the technique do not scale well, and can potentially waste significant bandwidth; (ii) priority-based approaches may display best-effort behavior and, in worst-case situations, may lead to unacceptable latency for low priority flows, besides being subject to jitter. In face of these limitations, rate-based scheduling arises as an option to improve the performance of QoS flows when varying traffic scenarios are used.

A virtual circuit is a connection between two users, devices, or terminals that Junctions as if it were a direct connection. Forwarding, switching, and/or routing over directly-connected circuits takes place at intermediate devices within the virtual circuit, but the details are hidden from the end points. Virtual circuits are widely used in many telecommunications and networking architectures; viz.: in wireline (PSTN, ISDN, BISDN) and wireless (GSM) telephony; in packet-switched networks at various layers (e.g., transport layer connections); in Multiprotocol Label Switched (MPLS) networks where tunnels are a type of virtual circuit; in Asynchronous Transfer Mode (ATM) networks where permanent and switched virtual circuits are fundamental components; in optical networks where wavelength routing and burst switching are abstract forms of provisioned virtual circuits; and in emerging multi-service networks where managed Ethernet virtual circuits are used to bridge legacy technologies like frame relay and ATM.

Thus far in this book we have discussed the signal processing necessary for digital communications from one point to another over a communications medium. The remainder of the book will begin to address the realization of a digital communications network in which many users simultaneously communicate with one another. One of the key issues that must be resolved in moving from a single digital communication system to a network is how we provide access to a single transmission medium for two or more (typically many more) users. In moving from a single transmitter and receiver to the sharing of media by multiple users, we must address two primary issues. First, how do we resolve the contention that is inherent in sharing a single resource. This issue is discussed in this chapter. Secondly, how do we synchronize all the users of the network as an aid to resolving the contention. We have discussed synchronization of a point-to-point link in chapter 15, but now we must extend these concepts in chapter 17 to the synchronization of many users. In chapter 18 we describe one particular multiple access technique based on signal processing called echo cancellation that is specifically used for full-duplex transmission.

This chapter presents an overview of Multi-Level Precedence and Preemption and its architecture behaviors. It describes the key components of the MLPP architecture, and a list of issues to consider.

Rate adaptation scheduling with prioritized call admission control for UMTS WCDMA networks is analyzed. Admission is based on maximum base station transmit power and service class priority. Conversational and interactive services are considered. Real Time services with stringent QoS requirements are operated in circuit switched mode over the dedicated radio bearers. Downlink shared channels are used to transport Non Real Time data handled in packet switched mode. Rate adaptation based upon link quality for NRT services and system load is introduced to improve performance and increase capacity. Results indicate that rate adaptation scheduling uses radio resource efficiently and improves system stability. Compared with fixed rate scheduling, rate adaptation scheduling results in better overall QoS.

We evaluate the blocking probabilities in optical networks which support multicast (or multipoint) service. Due to the difficulty of the problem, we restrict our attention to the case of the completely connected networks, for which the blocking probabilities act as a lower bound for other topologies. Nodes may, or may not be equipped with wavelength converters. In both cases, routing can be done either directly, or indirectly, and even in the latter case the number of hops can be either restricted to two hops, or unrestricted. We evaluate the blocking probability by first estimating the load offered to each link, and then use the iterative reduced load approximation method to refine this estimate. Using this estimate, we then obtain the probabilities of blocking and success. Numerical examples are presented, and are compared to simulation.

We describe the architecture, performance analysis and simulation result of a novel switching paradigm for optical WDM networks called Just-In-Time Optical Burst Switching (JIT-OBS) designed for ultra-low-latency transport of data-bursts across an optical WDM network. It combines the desirable features of circuit-switching and packet-switching, and features an out-of-band signaling scheme on a separate control channel with explicit feedback on delivery of data-bursts. We provide a performance analysis and simulation of the JITOBS approach, and compare its performance with those of circuit-switching and packet-switching approaches. We find that it has the best latency performance among the different switching mechanisms, and it has a better throughput performance than circuit-switching, and its performance is insensitive to network propagation delays.

In this paper the concept of a PMP system based on multi-carrier transmission, developed recently within Bosch-Telecom has been presented. The use of advanced microwave RF-technologies, dynamic bandwidth allocation (DBA), adaptive channel coding & modulation and efficient cell planning tools allows the service providers to build efficient local distribution networks and to offer circuit switched (e.g. POTS, ISDN, E1) and packet data (e.g. IP) services.

An integrated switching-system architecture is proposed in which IP-packet, X.25-packet, ATM, and circuit-switching traffic can be handled simultaneously. It is based on a hierarchical memory system. The optimum boundary condition for frames sent over a TDM line is discussed from the viewpoint of minimum-cost requirements. Effectively utilizing transmission lines in an international-connection environment is particularly important when integrating various kinds of multimedia traffic to be sent over one line. Typical examples with key traffic-characteristic parameters are discussed and evaluated for determining optimum bandwidth allocation.

During the last thirty years, the telecommunications industry has witnessed a proliferation of technological developments which have made possible a number of new concepts of operation providing different types of services to customers. The Integrated Service Digital Network (ISDN) was formulated in an attempt to integrate in a single network these services of a different nature such as the classical circuit switched facilities and the newer packet switched facilities.

We consider the off-line determination of the offered load that maximizes throughput in circuit-switched, wireless multihop networks, subject to Quality of Service (QoS) constraints on circuit blocking probability. This problem is of interest in network design for “sizing” the service capabilities that can be provided. We consider a rigorous formulation of the problem as a nonlinear optimization problem with nonlinear inequality constraints, and employ Lagrangian techniques for its solution. We develop several heuristic variations of the basic technique that improve the speed and accuracy of the search, and we demonstrate that there can be significant increase of the achievable throughput value via the optimization method as compared to other ad-hoc techniques for the maximization of throughput.

The synthesis of switched telecommunication networks is the problem of finding the optimal transmission capacity on the links between switches and the optimal routing of communications on these links. For circuit-switched telephone networks, the first methods can be traced back almost 80 years ago in the pioneering work of Erlang. For data networks, the first analysis and synthesis techniques go back to the 1960’s with the work of Kleinrock (1964), Fratta, Gerla and Kleinrock (1973). For historical reasons, telephone and data networks methods were developed in isolation from each other.

Many sectors of society, business, and industry have become dependent on the availability of communication networks. Today virtually all forms of traffic, whether circuit switched as in telephony, cell-based, or packet-switched, such as in B-ISDN or the Internet, are combined into standard synchronous optical network (SONET) carrier signals for transport on fiber optic transmission systems. A pair of optical fibers operating at the SONET rate of 9.8 Gb/s (called an OC-192) carries the equivalent of 129,000 voice or data circuits. A single cable may contain 48 or more of these fibers. The sheer volume of traffic makes the impact of a cable cut widespread and severe. Such events are, however, surprisingly frequent in practice. They have not only caused widespread congestion effects, but more serious effects such as loss of 911 service, loss of credit card services, and even the shut down of a major air traffic control network (MacDonald, 1994).

The plasma opening switch(POS) is a promising pulsed power circuit switching device which uses an injected plasma, embedded in a low density gaseous dielectric background, first to conduct high currents for a period ranging from tens of nanoseconds up to several microseconds and then to convert them into a load by means of some conduction and opening mechanisms.1,2,3

Circuit switching techniques are preferred in optical multiprocessor interconnection networks because they do not require any optical to electronic signal conversion to route messages. Thus, they can provide all-optical paths between message sources and destinations and the large communication bandwidth of optical signals is available to the nodes attached to the network. However, circuit switching techniques increase the complexity of managing the network, and thus may increase the communication delay. In a massively parallel processor, this added delay may affect the performance of a parallel program. In this chapter we model the performance impact of multiplexed circuit switched network management techniques. We show how program performance is affected by the choice of circuit switching techniques, by the multiplexing degree, and by the characteristics of the network.

This paper presents a distributed IN archit ecture called @INGate [1], [2], that elaborates a Unified Media Communications Service (UMCS) providing both messaging and telephony services between circuit switching and packet networks. It represents an InterWorking Unit (IWU) in CTI networks allowing the user access to a Unified Message Store-Line (UMSL) via traditional PSTN/ISDN equipment such as telephones and fax machines on the one side and networked computers equipped with mail readers and Web browsers on the other side to enable both online and off-line communication.

We present a static list scheduling technique for assigning tasks of a parallel program described as a task graph onto a distributed memory multiprocessor (DMM). Our technique factors in the impact of inter-task communication delays in heuristics for processor selection and channel assignment. Unlike conventional list schedulers that schedule task graph nodes top-down, we schedule task graph nodes bottom-up to get a better estimate of the scheduling weight of a task in the face of finite communication delays [2]. In [3] and [4], we applied and evaluated our scheduler for message switched architectures for a variety of interconnection topologies. This paper describes our scheduling technique as applied to hypercube connected systems using wormhole and circuit-switched routing.

A wide area network (WAN) connects one node to another over relatively large distances via an arbitrary graph of switching nodes. For the transmission of digital data, then data is either sent through a public data network (PDN) or through dedicated company connections.

At the outset of the WIS System Design study it become apparent that the WIS switching subsystem must satisfy communications requirements arising from the functions of its individual users in war and peacetime conditions as well as in periods of crises. The switching system concept was therefore developed by taking into consideration the operational and performance requirements of the WIS network and services. The WIS network is conceived to be based on a common user, basically circuit switched, area grid network, providing situation oriented, low delay, voice and non-voice services, supporting packet and message switches, and capable of accommodating evolutionary trends in services.

This paper presents a discussion of system design issues related to the development of next-generation multimedia personal communication networks (PCN). It is proposed that the baseline definition of emerging PCN systems be broadened to provide multimedia transport capabilities qualitatively similar to those offered by BISDN/ATM. A number of system-level design and technology selection issues are discussed. Specific topics considered include: multimedia PCN traffic handling requirements, service cost targets, system design criteria, choice of packet switching vs. circuit switching, strategies for ATM compatibility, media access control (MAC) selection, etc. The presentation is concluded with an outline of a dynamic TDMA based multimedia PCN system architecture consistent with the preceding discussion. Simulation based performance evaluation results are also given for such a dynamic TDMA PCN channel operating in example mixed voice/data traffic scenarios.

In chapter 9, we formulated the multi-stage circuit switching problem as a multi-commodity resource allocation problem, for which a terminal pair requests the simultaneous allocation of bandwidth on a set of links In contrast, packet switching does not require simultaneous allocation but allocation on a link-by-link basis. Local conflicts in resource allocation result in queueing. Based on results for single stage queueing developed in the previous chapter, we approximate the queueing behavior in multi-stage packet networks in this chapter. The first two sections deal with more general results such as reversibility and product form solutions for networks. The last three sections discuss performance analysis methods for buffered banyan networks under different traffic assumptions.

The previous chapters have addressed general topics related to digital switching, such as general switching networks and their controls, switching definitions, traffic theory, and switching system control and software. Many of these concepts are independent of whether the switch is analog or digital, space or time, or circuit- or packed-oriented. As the topics were discussed, areas of uniqueness were highlighted for the various types of networks. This chapter introduces the concept of time division circuit-switching networks and indicates how a circuit-switching function is performed in a time division system. The time division circuit-switching function is illustrated first for analog time division systems and then for digital time division systems, which forms the basics for digital switching. The digital time division networks then become the basis for the digital-switching systems discussed in Chapter 9. The remainder of this chapter describes the techniques and components associated with the networks characterized as time division circuit switches, as distinct from the packet switches described in Chapter 6. This chapter concludes with a discussion of potential future technologies for time division circuit switches, including sub-and super-rate, and optical switching. Later chapters will discuss the analog and digital terminations for the switch fabric, the impact of ISDN on the network, and some of the switching system architectures developed for public and private switching systems.

In practice, many large capacity switching systems have switching networks with more than three stages. The complexity of the control algorithm is a major bottleneck on the rate at which circuits can be established. The circuit rearrangement algorithms studied in chapter 3 require excessive computation. Circuit rearrangement algorithms are seldom used in practice. Instead a depth-first-search circuit hunting algorithm is usually employed. This is described in section 5.1.

In the previous chapter, we were concerned with making point-to-point connections. Formally, let I be a set of M inputs. and O be a set of N outputs. A point-to-point connection pattern is defined as a set of pairs C = {(i,o)} for which the i E I are distinct, and o E O are also distinct. In Mathematics, C is also known as a one-to-one mapping (Figure 1).

Quality of service and network performance parameters for voice and data transmission on existing packed and circuit switched networks are reviewed.

The three major elements in multiuser telecommunication networks are the station or terminal equipment, the transmission media and repeaters, and the centralized or distributed switching entities, including the signaling required for them to function.

In Chapters 3 and 5, we discussed the complexity of path hunting and time slot matching algorithms for multi-stage circuit switching networks. While these algorithms have manageable complexity for circuit switching for telephony, they are not fast enough to allocate bandwidth for multi-rate and bursty services. To alleviate the processing bottleneck of bandwidth allocation, we introduced distributed algorithms using self-routing addresses and queueing at switch nodes for dealing with local congestion.

In this chapter, we shall deal with the problem of providing point-to-point interconnection between terminals with the same bit-rate such as telephony.

Previous chapters have dealt with general topics related to switching, such as circuit switching networks, traffic theory, and control, which are independent of whether the switch is analog or digital. As the topics were discussed, the areas where these technologies were unique to digital switching were highlighted.

CCITT Recommendation X.21 has been developed as “The General Purpose Interface between Data Terminal Equipment (DTE) and Data Circuit-terminating Equipment (DCE) for Synchronous Operation on Public Data Networks” [1]. The only “general purpose” part, however, is the designation of the physical elements which include the electrical (X.26/X.27), functional (X.24), and mechanical (ISO 4903) characteristics described in Chapter 3 (Bertine). Additionally, the basic family of quiescent signals and states for the interface is specified. These provide the fundamental components of X.21 which will apply to all modes of operation in new data communications apphcations for circuit-switched, packet-switched, and general purpose integrated services [2].