Computationally-efficient traffic shaper

Details Computationally-efficient traffic shaper Computationally-efficient traffic shaper

1999-02-23

2003-09-16

Rao, Seema S. (Department: 2666)

Multiplex communications

Data flow congestion prevention or control

Control of data admission to the network

C370S395610, C370S412000

Reexamination Certificate

active

06621792

ABSTRACT:

TECHNICAL FIELD
This invention relates generally to packet-switching systems, such as asynchronous transfer mode (ATM) systems, and specifically to traffic shapers of such systems.
BACKGROUND OF THE INVENTION
Today's business communications environment consists of two separate network infrastructures: a voice network (such as a private branch exchange (PBX)) characterized by real-time, high-reliability, constant bit-rate (CBR) connections; and a data network (such as a packet network) characterized by high-bandwidth variable bit-rate (VBR) connections. Business needs for simplified maintenance, management, and access to information on diverse networks are forcing the convergence of these networks along with a new class of real-time multimedia networks. Asynchronous transfer mode (ATM) provides a single infrastructure that cost-effectively and flexibly handles both switching and transmission for the traffic types mentioned above (voice, video, and data) for both local-area networks and wide-area networks. The evolving network convergence requires the adaptation of the legacy PBX voice traffic to ATM. Voice telephony over ATM (VTOA) specifications allow adaptation of compressed or uncompressed voice pulse-code modulated (PCM) data streams into streams (virtual circuits) of CBR cells.
An ATM cell, regardless of the traffic it carries, is a packet 53 octets long: 48 octets of payload attached to a 5-octet header. The header contains addressing and management information used to direct the cell from source to destination and to ensure that the negotiated aspects of the traffic-flow through the ATM network are met. CBR traffic is assembled into cell payloads using ATM Adaptation Layer
1
(AAL
1
). The AAL
1
cell constructor layer uses the first octet of the payload for its header and-the remaining 47 octets to carry CBR information. ATM cell construction is then completed by attaching the ATM header to the payload.
An individual ATM cell carries traffic of a single narrowband or wideband channel (a virtual circuit). Illustratively, a narrowband channel is represented by a single time slot of each successive frame of a TDM bus, while a wideband channel is represented by a plurality of time slots of each successive-frame of a TDM bus. If a transmitter handles a to plurality of virtual circuits, different ones of the ATM cells that it transmits carry traffic of different virtual circuits. Likewise, if a receiver handles a plurality of virtual circuits, different ones of the ATM cells that it receives carry traffic of different virtual circuits. A transmitter may transmit the traffic of different virtual circuits to different receivers. Likewise, a receiver may receive the traffic of different virtual circuits from different transmitters.
The traffic of each virtual circuit is processed by its own instance of the AAL
1
cell constructor layer; the number of active instances varies as virtual circuits are added and removed. Assuming a channel carrying uncompressed pulse-code modulated (PCM) traffic having an 8 KHz sampling rate, each instance of the cell constructor that is processing a narrowband virtual circuit is supplied one PCM octet every 125 us frame period. It therefore requires 5.875 ms to fill a cell (47 octets * 125 us/octet). Fully constructed (mature) cells from all instances of the cell constructor are then transmitted.
Unlike the constant cell rate from narrowband virtual circuit, a wideband virtual circuit generates a constant average cell rate but a variable instantaneous cell rate. For example, a 32-narrowband-channel-wide wideband virtual circuit is supplied 32 octets during each frame period. The processing sequence effected by its cell constructor instance is as follows: during a first frame period, fill the first 32 octets of a first empty ATM cell; during a second frame period, fill the remaining 15 octets of the first ATM cell, transmit the first ATM cell, and fill the first 17 octets of a second empty ATM cell; during a third frame period, fill the-remaining 30 octets of the second cell, transmit the second cell, and fill the first 2 octets of a third empty cell; etc. This sequence shows an instantaneous cell rate of one cell every 250 us (two frame periods) sometimes and one cell every 125 us (one frame period) at other times. Variable bit-rate virtual circuits, whether narrowband or wideband, provide even greater diversity in the instantaneous cell rate.
ATM switches police cell traffic to ensure that it does not exceed the cell-traffic capacity of the switches. The average cell rate is policed on a per-switch basis, but the peak (instantaneous) cell rate is policed on a per-virtual-circuit basis. Receiving ATM switches delete received cells of each virtual circuit that exceed that virtual circuit's negotiated peak cell rate. And transmitting ATM switches meter out cells from each virtual circuit to ensure that they do not exceed that virtual circuit's negotiated peak cell rate.
The ATM standards specify a Generic Cell Rate Algorithm (GCRA) for use by both hardware and software-implemented devices, called traffic shapers, for metering out cells. The GCRA requires the traffic shapers to periodically process the output queue of each virtual circuit. The period may be different for each virtual circuit, and is a function of the negotiated peak cell rate for that virtual circuit. In the case of PCM traffic having an 8 KHz sampling rate, the processing period is the 125 us frame period. The queues of all virtual circuits are processed during each period. If the queue of a virtual circuit is empty, the traffic shaper does not transmit a cell from that queue. If a cell is available in the queue of a virtual circuit, the traffic shaper transmits it if it meets the GCRA criterion, i.e., if the period between the last transmitted cell from this queue and now is greater than or equal to the minimum administered cell-rate period. If more than one cell is available in the queue of a virtual circuit, the traffic shaper transmits one cell from the queue if it meets the GCRA criterion, and the next cell must wait at least until the subsequent processing period to be transmitted. A spacing of at least 125 us between transmitted ATM cells from any virtual circuit is thus ensured, thereby guaranteeing that the virtual circuit does not exceed its negotiated peak cell rate.
A problem with using the GCRA for multiple virtual circuits is that the traffic shapers must test the output queue of each virtual circuit during each processing period to determine if it has a cell to transmit. This is a particularly onerous task for GCRA implemented in software that supports large numbers (e.g., hundreds) of virtual circuits. Statistically, however, only one in 47 GCRA queue checks will result in the transmission of a cell during each processing period.
SUMMARY OF THE INVENTION
This invention is directed to solving these and other problems and disadvantages of the prior art. Generally according to the invention, traffic shaping for a plurality of streams of traffic (e.g. ATM virtual circuits or other packetized channels) utilizes a sequence of a first plurality of queues that are shared by the plurality of traffic streams, an enqueuing arrangement, and a dequeuing arrangement. Each of the traffic streams has at least one of the queues designated as its serving queue. Dequeuing is effected by cyclically transmitting contents of the sequence of the queues, i.e., by transmitting the contents of a different sequential one of the queues during each sequential period of the first plurality of such periods. During each period, a queue having a fixed offset in the sequence from the queue whose contents are being transmitted forms a current queue for that period. Preferably, the offset is zero, i.e., the current queue is the one whose contents are being transmitted. Enqueuing is effected as follows. In response to receipt of traffic (e.g., an ATM cell or another packet) from any said traffic stream during a period before a serving queue of that traffic stream becomes the current queue

Affiliated with

Also associated with

Rating

Computationally-efficient traffic shaper does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Computationally-efficient traffic shaper, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Computationally-efficient traffic shaper will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3109062