To efficiently transfer diverse traffic over highspeed links, modern integrated networks require more efficient packet-switching techniques that can capitalize on recent advances in switch hardware. Several promising approaches attempt to improve performance by creating dedicated "shortcut" connections for long-lived traffic flows, at the expense of the network overhead for establishing and maintaining these shortcuts. The network can balance these cost-performance tradeoffs through three tunable parameters: the granularity of flow end-point addresses, the timeout for grouping related packets into flows, and the trigger for migrating a long-lived flow to a shortcut connection. Drawing on a continuous one-week trace of Internet traffic, we evaluate the processor and switch overheads for transferring HTTP server traffic through a flow-switched network. In contrast to previous work, we focus on the full probability distributions of flow sizes and cost-performance metrics to highlight the ...

We suggest a new simple forwarding technique to speed-up IP destination address lookup. The technique is a natural extension of IP, requires 5 bits in the IP header (IPv4, 7 in IPv6) and performs IP lookup nearly as fast as IP/Tag-switching but with a smaller memory requirement and a much simpler protocol. The basic idea is that each router adds a "clue" to each packet, telling its downstream router where it ended the IP lookup. Since the forwarding tables of neighboring routers are similar, the clue either directly determines the best prefix match for the downstream router, or provides the downstream router with a good point to start its IP lookup. The new scheme thus prevents repeated computations and distributes the lookup process across the routers along the packet path. Each router starts the lookup computation at the point its up-stream neighbor has finished. Furthermore, the new scheme is easily assimilated into heterogeneous IP networks, does not require routers coordination.

This paper addresses a fundamental problem in resource management for flow-based hybrid switching systems. Such systems aim at efficient transport of layer-3 connectionless IP traffic over layer-2 connection-oriented ATM switching fabrics. One idea behind flow-based hybrid switching is to decompose individual IP packet streams into flows and then to classify them into short-lived and long-lived flows. While the short-lived flows are good for forwarding by the embedded software through permanent virtual connections (PVC's), the long-lived flows are more effectively transmitted by hardware through switched virtual connections (SVC's). Clearly the flow identification/classification mechanism will have great impact on the utilization of the system's resources. Unlike the traditional emphasis on resources such as link bandwidth and cell buffer size, our paper focuses on the resources which are directly associated with packet processing power, signaling capacity, and flow cache table size. Our study indicates that the presently available static flow classification methods have a vital shortcoming in balancing the utilization of the system's resources. We propose a novel approach for adaptive flow classification based on the min--max objective for the system resource utilizations to match with the time-varying traffic/resource characteristics. Based on the monotone properties and sensitivity analysis of the resource utilizations as functions of the control parameters, we first prove that the optimal solution of the static min--max problem is achieved at a unique balance point for the resource utilizations. With the intuition gained from the static results, we then design an adaptive controller formulated as a hierarchical stochastic automata control system with local search. T...

Workload distribution is critical to the performance of network processor based parallel forwarding systems. Scheduling schemes that operate at the packet level, e.g., round-robin, cannot preserve packet-ordering within individual TCP connections. Moreover, these schemes create duplicate information in processor caches and therefore are inefficient in resource utilization. Hashing operates at the flow level and is naturally able to maintain per-connection packet ordering; besides, it does not pollute caches. A pure hash-based system, however, cannot balance processor load in the face of highly skewed flow-size distributions in the Internet; usually, adaptive methods are needed. In this paper, based on measurements of Internet traffic, we examine the sources of load imbalance in hash-based scheduling schemes. We prove that under certain Zipf-like flow-size distributions, hashing alone is not able to balance workload. We introduce a new metric to quantify the effects of adaptive load balancing on overall forwarding performance. To achieve both load balancing and efficient system resource utilization, we propose a scheduling scheme that classifies Internet flows into two categories: the aggressive and the normal, and applies different scheduling policies to the two classes of flows. Compared with most state-of-the-art parallel forwarding schemes, our work exploits flow-level Internet traffic characteristics.

Trainet, a new scheme to extend MPLS (Multi Protocol Label Switching) is presented. The scheme works much like the subway system in a large metropolitan area. Each (unidirectional) subway line corresponds to a labeled path, and a route in the network is defined by either a pair !label,count-value?, where count specifies how many hops a packet still has to take in the specified train, or a route may be defined by a sequence of such pairs. A sequence of such pairs specifies that the packet has to take a number of hops in one train-line, and then continue for a certain number of hops on another train-line and so forth. While slightly increasing the number of labels in a header and adding a counter to each label, the scheme reduces the total number of different labels necessary in the network, and in each switch. Thus, for a given number of labels it may support a larger number of flows. Moreover, our scheme considerably simplifies the path set up cost while still providing all the feature...

In this work we discuss the flow classification in IP switched networks. Previous work done with flow classification methods has concentrated on optimizing the IP switch performance. We briefly examine the performance of several previously introduced flow classification methods and then we introduce the use of Learning Vector Quantization (LVQ) in flow clas- sification. The LVQ classifier has the ability to offer the user an intuitive traffic profile. The LVQ classifier is found to successfully classify traffic flows with feasible performance requirements while also providing the user with an unambiguous traffic profile.

We propose the use of "selectors for (active) packet flows" similar to tags employed in the IP world. Their impact on the performance of an active network node is significant, as active packets have to be demultiplexed not only to the network layer, but all the way to an application level Execution Environment. We have built an Active Network Node that implements the selector-based Simple ActivePacket Format (SAPF). Our measurements show that SAPF packets can be processed 30% faster than regular IP packets that use the traditional Active Network Encapsulation Protocol (ANEP) header.

: The efficient utilization and management of bandwidth in broadband networks is of paramount importance since bandwidth costs are likely to dominate network operational costs, especially in the wide area scenario. However, relatively little attention has been given to this issue in developing the current schemes for the transport of Internet data traffic over ATM. In this paper we describe a bandwidth allocation and management approach for the transport of Internet traffic over ATM that uses Virtual Private Networks, exploits the ABR service and implements bandwidth guarantees. We also address the efficient sharing of bandwidth between Internet data traffic and other media traffic using ATM. 1 Introduction We all have noticed the unprecedented growth of the Internet over the last few years, especially after the World Wide Web became so widely popular. As a result, there is an ever increasing volume of data traffic injected into the Internet, which is continuously pushing the install...

This is a survey of the state of the art in delivering IP services over ATM networks, as it stands in the second quarter of 1997. It also includes a look at the alternatives to that set of technologies. The technology and the choices are changing "on the fly", and have evolved significantly during the course of this project. Moreover, the issues are not exclusively technical, but in many respects reflect the great schism in the data communications world: connection-oriented versus connectionless networks. We have tried to present the technical issues and solutions along with an unbiased overview of the more "philosophical" issues. We indicate how we think the technology and the installed base of equipment is going to develop over the next few years, in order to give a picture of the future of ATM in data networking. Contents 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2 Supported Services . . . . . . . . . . . . . . . . . . . . . . . . . 9 3 The Standar...

. We describe some of the issues in the design of a new packet switched network protocol. Adaptation to various network technologies along the dimensions of speed, error model, robustness, etc., is a goal for this new protocol. We look at the adaption in size of the packet header to the speed and robustness of the underlying network to allow efficient communication on low-speed wireless networks, for example. We also explore issues in resource reservation and multicast for real-time multimedia, the notion of a network "flow", a hybrid of datagrams and virtual circuits, and suggest common solutions for both mobile and multicast routing. The authors are engaged in the design of a network protocol, NP++, whose goal is flexibility over a wide dynamic range of speeds and varying kinds of hardware switching elements. 1 Introduction The last quarter century has produced mainly two sorts of packet-switched networks: virtual circuit and datagram. Virtual circuit networks emulate circuit switch...