JPH01175431A - Packet communication system - Google Patents

Abstract

PURPOSE:To shorten packeting delay PD and to improve packeting efficiency by replacing headers of packets with one header, and connecting those packets into one data according to their generation order. CONSTITUTION:A terminal sends out information which is generated successively to a network in as short packets as possible. Three-byte packets are stored in different LCN buffers 303 by their destinations by a demultiplexer(DEMUX) 302 at an exchange in the network and waited for to be processed. Packets 305 and 306 to be processed which have the same LCN are coupled by a packet connector 304 into a packet 307. The connected packets stored by LCNs are read out of the LCN buffers 303 in the order of processing, multiplexed by a multiplexer(MUX) 308, and then switched by a switch 309. Consequently, the packeting efficiency is improved and traffic is reduced.

Description

[Detailed description of the invention] [Industrial application field] The present invention relates to a communication method in a packet switching network.

[Conventional technology and its problems]

In conventional packet switching, the longer the data length of Paqueso I□, the smaller the header overhead is, so it is more efficient. However, such as audio, which has high time transparency,
In other words, when transmitting packetized information from a call source that requires a guaranteed time interval between information occurrences and a property that requires high real-time performance, that is, a small delay between transmission and reception, it is difficult to use long buckets. The delay PD of buffering for packet creation is extremely large, which has become a problem as a cause of poor communication quality compared to circuit switching. In order to packetize continuously occurring information, it is necessary to buffer the information until it becomes one packet long. As shown in FIG. 5, the packet length and the packet creation delay PD are generally proportional.

To deal with this delayed PD light generation, there is a method to reduce the delay PD of packet creation using short packets, but when short packets are used, the packetization efficiency (data length / (header first child data length)) is reduced due to header overhead. Downstream traffic also increases, resulting in a decrease in the efficiency of use or throughput of network resources such as exchanges and transmission lines, and there is a problem in that queuing delay increases.

SUMMARY OF THE INVENTION An object of the present invention is to provide a bucket communication system that eliminates the drawbacks of the conventional system, reduces the packet creation delay PD, and efficiently uses network resources.

[Means for solving problem C] The packet communication system of the present invention divides continuously generated information into blocks at the terminal and sends them out as packets to the network, and within the network, the information is transmitted by queuing in the exchange or transmission path. Concatenation of packets by replacing multiple headers of these packets with one header and concatenating multiple data of these packets according to the order of generation to make one data for multiple packets with the same destination stored in the same hasher. It is characterized by doing the following.

[Operation] FIG. 1 shows the principle of the present invention. In the present invention, as shown in FIG.
to the network 104. When the load on the network is high and queuing occurs in the resource 105 inside the network 104, and multiple packets that have been sequentially packetized at the terminal 103 are simultaneously accumulated in the processing waiting sofa, these multiple packets are concatenated. .

FIG. 2 shows an example in which short packets t, p, 201 and short packets) P, , +202, which are continuously generated by the terminal 103, are concatenated to form a packet 203. It is assumed that the packet format is composed of a header including a destination (LCN) and a cyclic redundancy code (CRC), and data. The short bucket 201 is composed of a header consisting of LCNa and CRCa and data a,
The short packet 202 is composed of a header consisting of LCNa and CRCa and data b.

The destinations of short packets L-201 and 202 are both LCN a
are the same, and when connecting, short bakenos 1 to 201 and 2
02 is replaced with one header. At this time, the LCNs of the short packets 201 and 202 before concatenation and the packet 203 after concatenation are the same, that is, LCNa, and the CRC is recalculated and set as CRCab.

On the other hand, data a of short bucket 201 and short packet 202
The data b are concatenated into one data according to the order of generation.

The above concatenation forms a concatenated formant packet 203 as shown in the figure.

According to the packet communication method of the present invention, since the terminal transmits short packets, the packetization delay PD, which is proportional to the packet length, becomes extremely short. In addition, when concatenating packets within the network, headers are shared, which reduces header overhead, making it possible to increase packetization efficiency and reduce traffic.

In addition, by concatenating packets, short packets are recombined into long packets while queuing for processing, so long packets can be created without packetization delay PD for creating long buckets at the terminal.

〔Example〕

Next, examples of the present invention will be described.

The terminal divides the continuously generated information into the shortest possible packets and sends them to the network.

If a 64 kbps PCM audio signal (8 kHz sampling, 8 bit quantization) is used as an information source using the packet format 1- as shown in FIG. 2, the shortest packet length is 1 data hide + 2 headers.

This 3-hide packet 300 is as shown in FIG.
A demultiplexer (DEMUX) 302 is used in a switch 301 in the network to
N) is accumulated in a different LCN number sofa 303 and waits for processing. Packet 30 waiting to be processed with the same LCN
5 and 306 are concatenated by a bucket concatenator 304 to form a packet 307. The concatenated packets stored for each I and CN are read out from the LCN processor 303 in accordance with the processing order, multiplexed by a multiplexer (MUX) 30B, and then exchanged by a switch 309.

FIG. 4 is a diagram showing an example of the correspondence between short packets successively sent from a terminal and packets concatenated and multiplexed within the exchange.

The short bucket 401 has header HA and data D, the short packet 402 has header H9 and data D, and the short packet 403 has header HX and data D, and short packet number 0.
4 is a header Hβ and data Dβ, and the short packet 405 is composed of a header HB and data DB, respectively. It is assumed that the destinations of headers H7 and HB are the same, and the destinations of headers H7 and Hβ are the same. Therefore, the short packets 401 and 405, the short packets I-402 and 404, and the short bucket 403 are each separate 1. , C.N.
It is accumulated in the HaSofa and is waiting for processing. Concatenation is performed during this processing queuing. That is, short buckets 401 and 405 are concatenated to form a packet 406 with header HAR, data DA, and DB, and short bucket 4
02 and 404 are concatenated to form a packet 407 having a header Hαβ, data D, and Dβ. The packets are read out from the LCN processor in the order of processing, multiplexed by a multiplexer in the order of 406, 407, and 403 and output.

An example of the packetization delay PD that is shortened by the present invention will be described in the case where audio is assumed to be the information source.

When the information generation rate is 64 kbps, when a packet with a packet length of 128 bytes is created using the conventional method, the packet creation delay PD is 16 mm, but with the present invention, a short packet of 8 bytes is created. In this case, the packet creation delay PD was 1 millisecond, and when a 1-byte short packet was created, the packet creation delay PD was 0.125 milliseconds.

In this way, the packet creation delay PD can be reduced from several 1/10 to several 1/100 by changing the audio packet length, which conventionally had a packet data length of about 100 bytes, to a short packet with a packet data length of several bytes. Can be shortened.

Furthermore, when short packets I' P h and Pk+I are waiting for processing in the resources within the network, for example, in the case of 8 headers and 8 data, the packetization efficiency is 50%, but the concatenation of 2 buckets is If this is done, the header will be 8 hides and the data will be 16 high, and the packetization efficiency will increase to 66%.

Also, two 16-height packets are concatenated and 24
The traffic is 7 because it becomes 1 pakeso l- of Hyde.
5% decrease.

As described above, concatenation of packets increases packetization efficiency,
Reduce traffic.

Further, according to the present invention, packets are concatenated only for resources with high traffic and queuing to some extent within the network. In other words, the operation of concatenation corresponds to performing a type of flow control in which traffic is reduced in a resource with high traffic.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to provide a packet communication method that reduces the packet creation delay PD and efficiently uses network resources.

[Brief explanation of the drawing]

FIG. 1 is a diagram for explaining the present invention in detail, FIG. 2 is a diagram showing an example of packet concatenation, FIG. 3 is a diagram for explaining one embodiment of the present invention, and FIG. The figure shows an example of multiplexing of concatenated packets, and FIG. 5 shows the packet creation delay time in the packet switching method. 101 ... Information 102, 300, 307 ... Packet 103 ...
...Terminal 104 ...Network 105 ...Network resource 302
... Demultiplexer 303 ..., ...
...LCN buffer 304 ...Packet concatenator 308 ...Multiplexer 309
···· ··switch

Claims (1)

[Claims] (1) At the terminal, continuously generated information is divided into blocks and sent as packets to the network, and within the network, these packets are processed against multiple packets for the same destination that have been accumulated in the same buffer by queuing in the exchange or transmission path. multiple headers of a packet
1. A packet communication method characterized in that packets are concatenated by replacing a header with one header and concatenating a plurality of data of these packets according to the order of generation to form one data.