摘要:
Algorithms and data structure are described for constructing and maintaining a clock distribution tree (“CDT”) for timing loop avoidance. The CDT algorithms and data structure allows a node to make an automated and unattended path switch to the most desirable clock source in the network. In response to a network topology change, a clock root node distributes new clock paths to all nodes in the network. In particular, the root node calculates a new clock path for each affected node by building a clock source topology tree, and identifying from that tree a path to the network node from a clock source of higher or equal stratum relative to that network node. The root node then sends a network message to each node indicating the new path that the node should use. Each node receives the message and compares the new path with the existing path. If the paths are different then the node acquires the new path just received in the message. If the paths are the same then the node does nothing and discards the message.
摘要:
A timestamp-based clock synchronization technique is employed for CES in packet networks. The technique is based on a double exponential filtering technique and a linear process model. The linear process model is used to describe the behavior of clock synchronization errors between a transmitter and a receiver. The technique is particularly suitable for clock synchronization in networks where the transmitter and receiver are not driven from a common timing reference but the receiver requires timing reference traceable to the transmitter clock.
摘要:
Network elements may be synchronized over an asynchronous network by implementing a master clock as an all digital PLL that includes a Digitally Controlled Frequency Selector (DCFS), the output frequency of which may be directly controlled through the input of a control word. The PLL causes the control word input to the master DCFS to be adjusted to cause the output of the master DCFS to lock onto a reference frequency. Information associated with the control word is transmitted from the master clock to the slave clocks which are also implemented as DCFSs. By using the transmitted information to recreate the master control word, the slaves may be made to assume the same state as the master DCFS without requiring the slaves to be implemented as PLLs. The DCFS may be formed as a digitally controlled oscillator (DCO) or as a Direct Digital Synthesizer (DDS).
摘要:
A novel beacon-based position location technique for efficient location discovery of untethered clients in packet networks is disclosed. The position location technique utilizes the time-difference-of-arrival (“TDOA”) of a first signal transmitted by a beacon of known location and a second signal transmitted by an untethered client. The TDOA of these two signals is measured locally by at least three non-collinear signal receivers. For each of the receivers, the TDOA is used to calculate a perceived distance to the client. A circle is then calculated for each receiver, centered on the receiver and having a radius equal to the perceived distance. At least two lines defined by points of intersection of the calculated circles are then calculated. The point of intersection of the lines represents the location of the client. To facilitate operation, the signal receivers may be arranged on vertices which define a convex polygon as viewed from above. The location system requires no time (time-of-day) synchronization of the signal receivers, and only the coarse frequency synchronization, on the order of, tens of parts-per-million (ppm). The technique even works for the case where the signal receivers are run asynchronously, provided the frequency accuracies of the signal receivers are on the order of about 50 ppm or better. The technique introduces no communication overhead for the beacon, client and signal receivers. Further, the computation overhead at the signal receivers is relatively low because the location detection algorithm involves only simple algebraic operations over scalar values.
摘要:
The TCP receiver's advertised window (i.e., the receive buffer of a TCP connection) limits the maximum window and consequently the throughput that can be achieved by the sender. Thus, the idea behind TCP rate control is to match the offered network load to the available resources by modifying at an intermediate network device, the receiver's advertised window in TCP acknowledgments returning to the sources. In this disclosure, we propose a new TCP rate control scheme for a shared buffer where the buffer is logically organized into multiple queues. In the scheme, dynamic buffer thresholds are used to ensure efficient and fair usage of buffer memory among the queues. Conventional schemes allocate buffer space to each queue through the use of static buffer thresholds. This can result in unnecessary packet drops which leads to poor network performance since congested or heavily loaded queues cannot gain access to buffers not utilized by lightly loaded queues.
摘要:
A technique for an improved active queue management scheme which dynamically changes its threshold settings as the number of connections (and system load) changes is disclosed. Using this technique, network devices can effectively control packet losses and TCP timeouts while maintaining high link utilization. The technique also allows a network to support a larger number of connections during congestion periods.
摘要:
Flow control in a network is implemented based on aggregate traffic measurements. For example, in an ATM network only the aggregate background (CBR/VBR) traffic rate and the aggregate ABR traffic rate are used, in contrast with other schemes that require per-connection rate measurements or variables. An explicit rate is calculated recursively at discrete time instances using a scaled error value which is generated in response to an aggregate ABR input rate and a desired traffic rate. Explicit rate computations can be performed entirely by software, and the interval between computations is large enough to keep the processing overhead required of the switch very low. In addition, methods consistent with the present invention achieve max-min fairness and MCR plus equal share in a natural way without any additional computation or information about bottleneck rates.
摘要:
A method, system and master service interface transfer differential timing over a packet network. The transmitting service interface receives a service clock and is coupled to a receiving service interface through a network backplane. A primary reference clock is provided to time the network backplane. The primary reference clock and the service clock are used to synthesize a copy of the service clock connected to the transmitting service interface. A first control word containing an error differential between the service clock and the synthesized copy of the service clock is generated and transmitted through the network backplane via a packet. The first control word, together with the primary reference clock, is used to recreate the service clock for timing the receiving service interface.
摘要:
In response to a network topology change, a clock root node calculates a new clock path for each affected node by building a clock source topology tree, and identifying from that tree a path to the network node from a clock source of higher or equal stratum relative to that network node. The root node then sends a network message to each node indicating the new path that the node should use. Each node receives the message and compares the new path with the existing path. If the paths are different then the node acquires the new path just received in the message. If the paths are the same then the node does nothing and discards the message.
摘要:
A timestamp-based all digital phase locked loop is utilized for clock synchronization for Circuit Emulation Service (“CES”) over packet networks. The all digital phase locked loop at a CES receiver includes a phase detector, a loop filter, a digital oscillator and a timestamp counter. The all digital phase locked loop enables the CES receiver to synchronize a local clock at the receiver with a clock at a CES transmitter, where indications of transmitter clock signals are communicated to the receiver as timestamps. The phase detector is operable to compute an error signal indicative of differences between the timestamps and a local clock signal. The loop filter is operable to reduce jitter and noise in the error signal, and thereby produce a control signal. The digital oscillator is operable to oscillate at a frequency based at least in-part on the control signal, and thereby produce a digital oscillator output signal. The timestamp counter operable to count pulses in the digital oscillator output signal, and output the local clock signal.