Abstract: In a multiple access system, transmit authorizations are transmitted at a high rate in order to fulfil the requirements relating to the delay jitter. However, the transmit authorizations are transmitted only as long as is necessary. Following the reception of a data packet of a terminal, the transmission of further transmit authorizations for the same terminal is interrupted. Only after a specific time period, i.e. shortly before the expected transmission of a data packet of the same terminal, are transmit authorizations again sent to this terminal at a high data rate. In this way short time periods, in which transmit authorizations are sent to a terminal at a high rate, alternate with long time periods in which no transmit authorizations are sent to this terminal. In the long time periods, for example, transmit authorizations can be sent consecutively, in each case in short time periods, to one, two or more further terminals. In this way the bandwidth utilization is optimised.

Abstract: In the method according to the invention, to supply a service or an application, a service required by a customer or an application required by the customer is linked to a QoS category selected by the customer and the required service or the required application is supplied to the customer with the QoS category selected by the customer. In a first embodiment version of the method according to the invention a request to supply a service or an application is received from a terminal of the customer in the resources management unit and a request is transmitted to the customer to select one of at least two QoS categories. The selected QoS category is received from the terminal in the resources management unit and the requested service or the requested application is supplied to the terminal with the selected QoS category by the resources management unit.

Abstract: A communication system (KS) with a hierarchical server structure is indicated for a demand service, which makes possible a suitable compliance with requests for services from users (U1 . . . UN). To that end, the communication system (KS), as viewed by the respective user (U1 . . . UN), is divided into a central (I) and a local (II) distance zone. Movies (A1 . . . A10) frequently requested by the users (U1 . . . UN) are stored in local servers (SL1 . . . SLN) and seldom requested movies (B1 . . . B100) are stored in central servers (SM). According to the invention, in case the local server (SL1) is unable to provide a movie (B2) requested by a user (U11 . . . U13), the central server SM or another local server (SL2 . . . SLN) complies with the request through a switched connection between the requesting user (U11 . . . U13) and the central server (SM), or the other local server (SL2 . . . SLN).

Abstract: A method of operating a switching network, a switching network, and an exchange with such a switching network are described. The switching network consists of several switching modules and an interconnect network between these switching modules. The interconnect network includes a switch unit and interconnects the switching modules in accordance with a logic interconnection structure. The switch unit is simple and can be implemented by all-optical means. By means of the switch unit the logic interconnection structure can be reconfigured. This makes it possible to reconfigure the switching network in the event of a failure of a switching module or according to the load characteristic (video, voice, computer-computer communication). Further, a uniform switch unit can be used to meet different requirements ("flexible hardware").

Abstract: For traffic mixing, an implementation involves generating a random element in a device that detects nondeterministic events already occurring in operation of the device and mapping them to the random element by means of an assignment rule. The traffic mixing in an ATM switching element (SE) is achieved by cyclically writing data packets into a buffer (MEM) and cyclically reading data packets out of a buffer (MEM). The write cycle in this case is clocked by the arrival of a data packet, that is, by a nondeterministic event. The read cycle is clocked by a stable clock signal.

Abstract: A buffer or other communications resource in, e.g., an ATM switch element receives random data which is then used by different data sinks. After the data has been outputted to the data sinks, the communications resource (e.g., the memory locations of the buffer) are released, i.e., labeled as free again. However, a resource which is not marked as free as a result of an error (whether erroneously the data is not retrieved or whether the release procedure is erroneous) remains blocked. To avoid permanent blockage of those memory locations whose contents have been in the buffer so long that they definitely (or at least very likely) should have been called for, information on the time of entry is stored together with the data. All memory locations are checked at regular intervals for the age of their contents. Upon attainment of a predetermined age, the location is labeled as free.

Abstract: To be able to load an ATM switching network as efficiently as possible, the successive cells of one and the same connection are routed to the output by as many different paths as possible; however, mutual overtaking of successive cells must be avoided or corrected. Each cell is held at the output until it is certain that no older cell can be buffered in the switching network. Before the cell is passed on, a check is made to determine whether the cells received from the switching network later include an older cell which must be passed on before that cell. At the input end, consecutive numbers ("Sequence Numbers") are allocated to the cells. A buffer (1) at the output end is operated, at least in part (1 . . . d), in the manner of a shift register, thereby introducing a predetermined delay. Prior to the output of a cell, at least the shift-register-like portion of the buffer is examined as to whether a cell written in later is to be put out earlier: if so, the two cells are interchanged.

Abstract: Fixed numbers of packets are combined into frames within the switching facility, and all packets are divided into subpackets of equal length and distributed to subframes. Switching takes place on the basis of subframes using synchronous time-division multiplexing. The amount of storage space needed in the switching network is reduced in the ratio of frames:subframes. The buffers are those which are necessary for synchronous TDM switching anyhow.