The Dynamic Management of Guaranteed Performance Connections in Packet Switched Integrated Service Networks
Abstract: The communication infrastructure of the future must provide efficient support for applications with diverse traffic characteristics and performance requirements. Currently these applications are supported using several specialized networks that accommodate the different services (e.g., cable networks for video, phone networks for voice, and so on); however, technological advancements in the fields of microelectronics and optics have made it possible to integrate these services on a single network. These Integrated Services Networks support a wide range of qualities of services to the client and provide many advantages, which include large economies of scale, increased network management capabilities, improved statistical multiplexing, and ubiquity of access. The qualities of service offered by these networks include guarantees on various performance indices for a client's specified traffic characteristics; these services are often referred to as Guaranteed Performance Communication (GPC) services.
These GPC services provide performance guarantees in terms of throughput, delay, delay jitter and loss rates, and adopt a connection-oriented, fixed-routing, reservation-oriented approach to achieve these guarantees. In such an approach, resource allocation and route selection decisions are made before the start of the communication on the basis of resource availability and real-time network load at that time, and are usually kept for the duration of the communication. This rather static resource management approach has certain limitations: it does not take into account (a) the dynamics of the communicating clients; (b) the dynamics of the network state; and (c) the tradeoff between quality of service and network availability, thus limiting the flexibility or adaptability of these guaranteed-performance services. In order to accommodate the dynamics of client demands and network state, it is necessary that the GPC services be flexible.
In this thesis, we examine this problem of flexibility of GPC services in wide-area packet-switched networks, and present a solution by proof of concept; that is, we designed a dynamic resource management scheme, analyzed its behavior through simulation experiments, and implemented a prototype of the scheme. This dynamic resource management scheme, called the Dynamic Connection Management (DCM) scheme, provides the network with the capability to dynamically modify the traffic characteristics, the performance requirements, and the routes of any existing guaranteed-performance connection. We begin this examination by providing several examples of the dynamics of the client demands and of the network state to motivate our work, and we continue with a review and critique of various proposed solutions. We then present the concept of Dynamic Connection Management and discuss its components: namely, the DCM scheme and the DCM Policies. The DCM scheme is a collection of algorithms and mechanisms that permit the runtime modification of the traffic and performance parameters, and the route of a guaranteed-performance connection. The DCM scheme is guided by high-level management policies, called the DCM policies, that determine when modification is permissible in the network and the values of the appropriate parameters to be modified. The focus of this thesis is the DCM scheme.
The DCM scheme is an enhancement of the Tenet GPC service; it is based on three algorithms: the DCM channel administration algorithm, the DCM transition algorithm, and the DCM routing algorithm; and it is subject to the DCM modification contract. This contract specifies the degree of disruption that a client may experience during a modification. This degree can range from no disruption to a bounded number of performance violations. The channel administration algorithm conducts the admission control tests and reserves the appropriate network resources to ensure that the performance guarantees of the modified channel are satisfied during and after modification. The DCM transition algorithm ensures that the performance violations specified in the DCM modification contract are adhered to during modification. The DCM routing algorithm determines a route from the source to the destination host according to the traffic and performance requirements and other administrative factors.
The DCM scheme also supports mechanisms that enable modifications to a connection to be made to a segment of the connection (local control) or to the entire connection (global control). Furthermore, faster establishment and modification is also possible as the DCM scheme uses the intelligent restart establishment mechanism, which utilizes the real-time network and client state to compute the path before establishment and the time value of the network state information to bypass unavailable links during establishment. The DCM scheme was verified and analyzed by a series of simulation experiments. These simulation experiments indicated that the scheme is functionally correct and that the performance of the scheme is very acceptable given our time scales of interest.
In completing the proof of concept, we implemented the DCM scheme, and conducted several initial measurement experiments on this prototype implementation. The implementation provided the basic management mechanisms, using a standardized "open" management framework, namely the Simple Network Management Protocol version 1 (SNMPv1), by which guaranteed performance connections can be monitored and controlled. In our implementation the data delivery protocols used were the transport-level message protocol and network-level protocol of the Tenet Real-Time Protocol Suite; namely, the Real-Time Message Transport Protocol (RMTP) and, the Real-Time Internet Protocol (RTIP). Management Information Bases (MIBs) were designed for these data delivery protocols and the DCM scheme, and are used provide the monitoring and control capabilities. Results from the initial monitoring and control experiments, which were conducted on a local-area testbed, indicated that the prototype is effective in supporting the monitoring and control of guaranteed-performance connections.