JP3392533B2 - Connectionless data transmission method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for transmitting connectionless data in a connection-based optical network constructed by optical wavelength division multiplexing technology.

[0002]

2. Description of the Related Art With the development of an information-oriented society, there is an increasing demand for a network capable of communicating a larger amount of data at a higher speed. The optical wavelength division multiplexing technology for simultaneously transmitting a large number of lights having different wavelengths in one optical fiber is expected to be capable of constructing a network whose total throughput exceeds 100 Gbps.

Optical wavelength division multiplexing networks are currently considered to be first applied to computer networks. In a computer network, burst data, that is, very large data when transmitted, is transmitted at high speed, but the transmission and reception of data is such that one transmission time is short and the frequency is low.

Optical wavelength division multiplexing technology can also be used in backbone networks. It is considered that the information transmitted in this case is mainly data requiring a connection such as a telephone and a video signal.

Recently, the following protocol based on connection has been considered (Japanese Patent Application No. 5-096).
196). Each node is connected to the star coupler as shown in FIG. 10A, and the general nodes (n1 to nn) are shown in FIG.
As shown in (b), a set of data transceivers and access means to the control channel (here, the control channel is one of wavelength-multiplexed light, and the access means is the WDM coupler and the optical transmission / reception for the control channel). It consists of a vessel). In addition to general nodes, the star coupler includes a network controller (N
WC) is connected. Each data wavelength is divided into frames as shown in FIG. 11, and the frames are further divided into slots. In addition, the transmission wavelength of each node is fixed and varies from node to node.

Connection-oriented calls, such as telephones, continue to use some of the bandwidth in the network for several minutes. In order to secure this band, a certain slot in the frame is repeatedly occupied. For example, node n
It is assumed that a call to i is made to nj. Also, at that time ni
To nj have not sent any other calls. Then, ni must newly reserve a slot addressed to nj. Through the control channel, ni requests the NWC that a new slot for nj is needed and how many slots are needed. N
WC is the slot usage status of the transmitter of the transmitting node ni,
The slot usage status of the receiver of the receiving node nj is checked to find a common empty slot for both. When the requested number of common empty slots are found, the NWC notifies ni and nj of items necessary for communication such as slot positions through the control channel. Upon receiving the notification, the ni communicates with the nj in the slot designated by the NWC. Ni is used when the slot is no longer required to be used due to a call termination or the like.
Notifies NWC and releases. The ni can use the slot for communication to nj in all frames until it is released after permission to use the slot.

In such a system, all transmission and reception are performed based on the connection. Therefore, even in the case of burst communication (connectionless communication) such as file transfer that does not require a connection, it is necessary to apply to the NWC and once establish a connection. In a network such as a LAN using optical communication, the span of the network can exceed several tens of km, and with the development of an optical fiber amplifier, a span of more than 100 km can be realized. In this case, NW
The time it takes to interact with C is very large due to the propagation delay. In connectionless communication, there are many applications that want to send with the lowest possible delay because the discard rate may be somewhat high. The delay due to the time required for communication with the NWC is very useless, especially when the network load is small and the success probability is likely to be high even if a connection is not established.

Humblet et al., "An Eff
icient Communication Prot
ocol for High-Speed Packe
t-Switched Multichannel N
networks ”(IEEE Journal on S
elected Areas in Communic
ations, Vol. 11, No. 4 p568,1
In 993), a connection-based protocol in a distributed system without NWC is proposed, and a method for performing connectionless communication is also described therein. Since it is a distributed processing system, it always broadcasts the reception slot status of its own node. Therefore, although the transmission side is slightly behind in time, it is possible to know the slot status of the other party and select a slot that the other party can receive to transmit. However, the distributed system has a large amount of control signal processing at each node, and
Inefficient use of bandwidth. However, there is no such defect N
In the centralized system with the WC, the transmitting node cannot directly know the slot status of the destination node, and the same protocol as in the distributed system cannot be used.

[0009]

As described above, in the connection-based wavelength division multiplexing network of the centralized control system, a method for efficiently transmitting connectionless data with low delay without establishing a connection has not been considered. The present invention provides a method for efficiently transmitting connectionless data in such a network with the lowest possible delay.

[0010]

In order to solve the above problems, in the first invention of the present application, a plurality of general nodes are connected to a star coupler, and the general node is a wavelength fixed data transmitter and wavelength variable data. receiver, and have a means of access to the control channel, each data wavelength is time-divided into frames, the frame is time-divided into slots, the general node application via the control channel in the connected band management node to the scan Takappura Then, the slot is reserved to communicate with other general nodes, and the reserved slot is composed of one or more packets in the centralized control type optical wavelength multiplexing network that can be used in all frames until the reservation is canceled. The general node that intends to send connectionless data is connected to the control channel through the connection. Data, its amount of information, and the destination node are declared, and then the connectionless data packet is transmitted using the transmission empty slot of the own node, and when the destination node receives the declaration, the destination node receives it. The empty slot is tuned to the wavelength of the node that made the above declaration, and for the received packet of the connectionless data packet, ack is returned through the control channel. There is provided a connectionless data transmission method characterized in that a packet to be transmitted is retransmitted in a slot in which a connectionless data packet returned by ack is transmitted.

Further, in the second invention of the present application, the general node, which is going to transmit the connectionless data consisting of one or more packets, has a According to an algorithm used by the band management node when setting a slot, a connectionless data transmission method is characterized in that a destination node of the connectionless data also uses a slot having a high probability of being an empty slot.

[0012] In the third aspect of the invention, are connected to the plurality of general nodes in a star cup la, the general node data transmitter fixed wavelength and tunable data receiver, and,
Has access means to the control channel, each data wavelength is time-divided into frames, the frame is time-divided into slots, the general node to apply through the control channel in the connected band management node to the scan Takappura slot
In the centralized optical wavelength division multiplexing network, the reserved slot communicates with other general nodes and the reserved slot can be used in all frames until the reservation is cancelled. The general node that intends to transmit data declares that it is connectionless data on the control channel, its information amount, and destination node, and then transmits the connectionless data using its own transmission empty slot. After that, the connectionless data is transmitted once or more so as not to send the same packet in the same slot by changing the order of the connectionless data packets or changing an empty slot for transmission. Provide a transmission method.

In the fourth invention of the present application, a plurality of general nodes are connected to the star coupler, and the general node has a fixed wavelength data transmitter, a variable wavelength data receiver, and a control channel access means. each data wavelengths are time-divided into frames, the frame is time-divided into slots, the general node to apply through the control channel in the connected band management node to the scan Takappura slot
By communicating with other general nodes,
In all Ru current in the control type optical wavelength multiplexing network can be used in frame to the reserved slot is canceled reservation, the general node tries to send connectionless data consisting of one or more packets, through the control channel After the band management node is told about the common empty slot with the destination node of the connectionless data, after declaring that it is connectionless data on the control channel, its information amount, and the destination node, A connectionless data transmission method is provided, wherein the connectionless data is transmitted using a slot.

[0014]

In the centralized control type network as the premise of the present application, the empty slot status of nodes other than the own node is not known in a normal use state. Therefore, in the first invention of the present application, connectionless data is transmitted using the empty transmission slot of the connectionless data transmission node. In the network on which the present application is based, there are usually some slots in use that are performing connection-oriented communication in the frame, and even if the slots are sent by the sending node, it is not always possible for the receiving side to receive them. Absent. Therefore, the receiving side returns ack (acknowledgement: receipt notification) for the data that can be received on the control channel. It is assumed that the control channel does not require an access request and uses a pre-assignment method that does not cause collision. The reason why ack is returned on the control channel is that if you try to return it on the data channel, you need to establish a connection in order to surely return it, and if you try to return it by the same means as connectionless, it may not surely return. Because.

The connectionless data transmission node which has received the ack can know in which slot the data transmitted was received and in which slot the data transmitted was not received. Therefore, the unreceived portion is retransmitted using the slot in which the received portion is sent. By doing so, the probability of successful retransmission is much greater than the probability of successful initial transmission. Of course, during the meantime, the slot for which ack is returned may be filled up due to the occurrence of other connection-oriented calls, so the receiving node should receive the received data regardless of the initial transmission or retransmission. returns ack.

In the second invention of the present application, there is a rule when a network controller (NWC: equivalent to a bandwidth management node in claims) sets a slot (accepts a reservation) for a connection-oriented call. Is a method to increase the probability of success.

For example, if the following rules are used, the slots used by the connection-oriented call are concentrated in the front of the frame as a whole.
If there is a slot application and more than the requested number of common empty slots are found when setting, select the one near the front of the frame. Also, when multiple calls are communicated to the same party using multiple slots, the number of slots can be reduced by ending one of the calls, but in that case, the number of slots Free the slot near the back.

As a result, the empty slots are gathered backward, so that when transmitting connectionless data, if the empty slots of the local node at the rear of the frame are transmitted, it is possible that there will be a common empty slot. The probability of success is high.

The third invention, since the transmitting node if the sending by the sending empty slot of the own node by a slot may not receive destination node also transmits repeatedly the same data with different slot Is the way. In this method, there is no request for retransmission, but it is sent many times, but it seems wasteful, but in a system where the propagation delay between nodes is larger than the frame length, the ack-retransmission procedure Takes a very long time and causes a large total delay. Therefore, in such a system, the procedure of retransmitting without waiting for ack is effective.

In the fourth invention of the present application, in a network with a high usage rate, if connectionless data is sent in an empty slot of the own node, the success rate is very low and it takes time to retransmit. Therefore, this is a method in which the NWC is informed of the common empty slot in advance and the common slot is sent there. This is a method in which the delay can be made relatively small when sending over a heavily loaded network.

In the above procedure, it is assumed that the connection oriented slot allocation changes relatively slowly. That is, while transmitting connectionless data, slot allocation is performed, for example,
It is assumed that even if one or two are changed, they will not be in completely different states. This is because the network has a long hold time for connection-oriented calls.

[0022]

Embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a diagram for explaining an embodiment relating to the first invention of the present application. Suppose node ni wants to send connectionless data to node nj. FIG. 1 is a time chart showing a related transmission / reception schedule. The frame of the data channel is divided into slots, and in the example of FIG. 1, one frame is divided into 10 slots.

Slots indicated by ◯ in the figure are slots used for connection-oriented calls. The data transmission wavelength of each node is fixed and varies from node to node. Therefore, each node can use the fixed wavelength exclusively for transmission, and ni can send connectionless data in the empty slot of its own node that is not marked with a circle (whether or not it is received). , The transmission is always successful.

In FIG. 1, slot no. Of 3, 6, 9, 1
0 is an empty slot for transmission of ni. Where ni is n
Suppose that we want to send connectionless data for 6 slots to j.

On the other hand, the control channel is divided into frames shorter than the data channel as shown in FIG. In the example of FIG. 1, it is a half frame of the data frame. One control channel is TDM (Time Div)
(ion Multiplexing: time division multiplexing)
As a pre-sign, the time and time (slot) in the control channel frame that can be transmitted by each node are fixed in advance. After the connectionless data is generated, ni makes a declaration in the transmission slot of its own node of the control channel that is the closest in time (after the processing time until the declaration and the data can be transmitted). The content of the declaration is the transmission of connectionless data,
The destination node (nj in this example) and the information amount of data (the number of slots, six in this example). Hereinafter, the data sent in one slot will be called a packet.

Further, in order to teach the period during which the destination node needs to perform the connectionless data receiving operation, the number of slots from end to end (FIG. 1) or the end required for connectionless data transmission is included in the declaration. It is good to include the time from the end to the end. From this, it is understood that after receiving the declaration, the destination node receives at most the number of slots (or hourly can) +1 frame. As a simpler method, although the bandwidth of the control channel is somewhat wasted, the transmission node may inform that the transmission has been completed on the control channel when the connectionless data transmission is completed.

After the declaration, the transmitting node starts transmitting connectionless data using its own empty slot. In this example, as shown in FIG. 6, the transmission is started from the slot No. 6 of the next frame. Transmit with empty slots up to 9. At the time of transmission, the number in the connectionless data is assigned to each packet. here,
For example, like circled numbers in FIG. 1, numbers 1 to 6 are put in a header or the like and transmitted.

On the other hand, when nj receives the declaration of connectionless data transmission, ni receives ni in the reception empty slot of its own node.
Tune to the transmission wavelength of and receive. Here, the slot no. Tune to 3, 4, 8, 9, 10.

However, the slot no. The data received at 4 and 8 are not connectionless data addressed to nj.
In order to identify this, it is advisable that the connectionless data be provided with a header that clearly indicates that the data is connectionless data, and the destination and the ID number of the data so that the destination node can identify it. When nj receives a packet that is not addressed to its own node, the data is discarded.

If an ID number is assigned to the data, ni includes the number in the declaration contents when the control channel is declared. Alternatively, the declared obvious ID number from content, may be determined by Determinictic (deterministic) algorithm, which is installed in common to each other node. The former is simple and the latter saves bandwidth on the control channel.

Nj is the slot no. Since the packets Nos. 2, 3, 4, and 6 (circled numbers) received at Nos. 3, 9, and 10 were able to be received, ac at the nj transmission slot of the control channel
returns k. The form of ack may be such that an ack is generated and returned for each packet, but for example, the source node (ni),
If the connectionless data ID and packet number (2, 3, 4, 6) are returned together (if any), the bandwidth of the control channel can be saved. Alternatively, an intermediate value may be taken, and every time the transmission slot of the nj control channel comes around, the packet numbers received so far may be grouped together and an ack may be returned.

Ni is a control channel and ack from nj
When the packet is received, the packet for which ack is not returned is retransmitted. However, if the ack is returned several times, or it is not received at all and one ack is received.
In some cases, it may not be returned, so you have to decide the deadline before starting the retransmission procedure.

Here, in the case where no packet is received at all and the number of packets is relatively small, ni may not use all empty slots of its own node at the time of transmission. In such a case, there is a possibility of success if retransmission is performed using an empty slot of the own node that was not used in the initial transmission. However, in the case as shown in FIG. 1 in which all the empty slots of the own node are used, there is little possibility of success even if it is immediately retransmitted. Depending on the delay time limit of connectionless data and the status of the memory of the node, wait for a time to change the usage status of the connection oriented call slot, or discard the data as untransmittable and send it to the terminal that sent the data, Take a method such as notifying the transmission failure.

There are several possible methods for determining the deadline for starting the retransmission procedure. The maximum value of the processing time required for a node that receives connectionless data to generate ack and make it ready for transmission is relatively small, and
In a system that can always send the generated ack in the next transmission slot of the control channel of the node, for example, the following is performed. As shown in Fig. 2, after sending the last packet, wait a round trip of the propagation delay time from the sending node to the receiving node and the maximum value of the processing time required to generate ack, and then send ack The ack up to the transmission slot on the control channel of the node to be activated may be regarded as the successful portion.
In this way, the delay can be reduced. FIG. 2 is a graph in which the horizontal axis is time and the vertical axis is distance, and the specific operation is described.

In the above case, the transmitting node must know the propagation time between the connectionless data transmitting node and the receiving node. For example, when the system introduces information about the propagation delay from the star coupler of each node in the memory, and when a new node joins the system, the time (or distance) from the star coupler of that node is set to another value. Inform the nodes through the control channel that each node knows the propagation time from the star coupler for all nodes. Since the propagation time changes due to changes in the temperature of the optical fiber, the deadline for starting the retransmission of connectionless data may be calculated with a value that allows for the maximum value of the change.

In the system on which the present invention is based, as shown in FIG. 11, all wavelengths must be synchronized in frame and slot. This can be achieved by transmitting the light transmitted from each node at the time when the lights are mixed by the star coupler so that the timing is as shown in FIG. 11.
Each node is provided with means for knowing the propagation time from the own node to the star coupler relatively accurately (described later). Therefore, it will use a lot of bandwidth on the control channel, but if the time to the star coupler is measured and it deviates to some extent from the previous data, broadcast a new value on the control channel to notify other nodes. You can By doing so, it is not necessary to look at the maximum value of the change as described above, the time until the retransmission procedure is started is shortened, and as a result, the total delay of connectionless data transmission is reduced.

Alternatively, each node has a slightly large delay, but each node has a propagation time t of a node farthest from the star coupler.
Knowing only max, the time from the transmission node to the star coupler + tmax or tmax + tmax may be used instead of the propagation time from the transmission node to the reception node. This method saves memory and access time when the distance to the star coupler of each node is relatively uniform.

The above example is based on the premise that the processing time required to generate an ack and be ready for transmission is always within a relatively short finite time, and transmission is always possible in the next transmission slot on the control channel. . However, the system for generating ack is not limited to the one for generating ack, and there is a system in which it may take a very long time to generate ack for performing other processing having higher priority. . Further, in some systems, the amount of information that can be transmitted at one time on the control channel is relatively small, and because other information having a higher priority is transmitted on the control channel, ack may not be transmitted immediately.

In such a system, until the retransmission procedure is started in anticipation that an ack is sent and arrives, for example, the control channel of
You just have to wait an extra frame or two frames.

On the other hand, the side that sends ack needs to know how long the other party will wait. In the system that determines the processing order by priority as described above,
How close to the deadline for processing can be one priority. Therefore, it is better to know the deadline accurately. At this time, there is no problem in the system having the distance between the nodes in the memory, but in the system using the time tmax from the star coupler to the farthest node, the time from the node transmitting the connectionless data to the star coupler I don't know, so I don't know how long the acker will wait.

In such a case, the time from the node that sent the packet to the star coupler is preferably sent together when sending the declaration on the control channel before sending the packet. Alternatively, if a time stamp indicating a transmission time is put in a header of a packet to be transmitted and the packet is transmitted, if each node has a common clock, the distance between the nodes can be known and the deadline can be known.

Further, there is a time from the start of the retransmission procedure to the actual transmission. Therefore, if the ack is sent back even after the deadline for sending the ack expires, it is possible to arrive before the actual resending and stop the resending procedure. By doing this, it is possible to avoid unnecessary transmission.

In a system in which one connectionless data is divided into several packets and retransmission control is performed for each packet as described above, the arrival order of the packets is out of order, and therefore a process for reassembling them is necessary. For example, packets are stored in a buffer or memory, and when they are all complete, they are read in the correct order. The order is the total number of packets included in the declaration contents at the time of declaration and the number assigned to the header etc. in advance at the packet transmission side. Read and know. Also, the transmitting side performs a procedure of deleting the packet for which ack is returned from the buffer or the memory.
This is, for example, after the initial transmission is completed, temporarily move the connectionless data from the transmission buffer to the memory, delete it from the memory when ack is returned, and retransmit the remaining one even if the deadline expires. Is.

Retransmission control is performed by, for example, packet numbers 1 and 3
When the ack is returned, the procedure may be started assuming that 2 was not received. In this way, the total delay can be reduced. If the system has a sufficient processing capacity, a deadline until the start of the retransmission procedure may be set for each packet.

Since the process of reassembling the order as described above does not meet the cost, there is a system in which the entire packet is retransmitted when one of the packets cannot be received. In such a case, since each packet is not retransmitted, the node that receives the connectionless data should return ack only when all the packets of the connectionless data are received in order to save the bandwidth of the control channel. good. Then, when all the slots have not been received, if the information about which slot can be received is returned instead of ack, the success rate of retransmission becomes high. Such an approach, since the load of the network is very small system has a high success rate, it is effective to cost savings.

The retransmission control may be repeated until all the packets have been sent, but if the connectionless data has a transmission deadline, it is stopped at that time and the terminal that issued the connectionless data is notified of that. Notifications may be left to higher layer protocols to reduce node processing. Alternatively, in order to save the memory and processing of the node, the network may decide the time to discard the connectionless data unless it is sent by the time it arrives at each node.

In order to synchronize the slot and frame at the network level, each node frequently measures the time required for light to propagate from the star coupler to its own node. FIG. 3A shows an example of the measurement system. When sending a part of the transmission data from the output of the buffer to the optical transmitter of the transmitter, a part of the power is branched and put into the register. The counter is turned on at the time when the data is stored in the register. The one sent to the optical transmitter is transmitted and returns through the star coupler. In the receiver, the received data is input to the shift register. As shown in FIG. 3, the received data is branched and input to the shift register, or the shift register is arranged in the path of the data. The shift register and the register take an exclusive OR in bit units, and the result of the exclusive OR of each bit is or'd, and when the result of or becomes 0, the data in the register , The data in the shift register match. The value of the counter at that time is the delay from the transmission of the same data as the data stored in the register to the reception through the star coupler. The delay of the electronic circuit is also included in it, so that part is corrected. This measurement should be done on the data channel. Since the system requires slot and frame synchronization, and on the TDM control channel, it is possible to know at what timing each node can transmit without overlapping the transmission of other nodes unless the time from the star coupler of its own node is known. Absent. In the control channel measurement system, a node that newly joins the network cannot even transmit for measurement (unless another node suspends transmission / reception for that node for a while).

On the other hand, since the transmission wavelength of the data channel is only for that node, even if a newly added node outputs data for measurement at a random timing at the start-up stage, it is transmitted by another node. There is no annoyance due to overlapping with. Also, the data channel has a higher transmission rate than the control channel. In the method of measuring with the counter as described above, the higher the transmission rate, the better the measurement accuracy.

When measuring with the data channel, the output data must be able to be received by the own node. Therefore, it is performed in the empty slot of the receiver of the own node.

Further, by using a slot which is a common empty slot of the transmitter / receiver of the own node, a predetermined pattern may be transmitted and the time until it returns may be similarly counted. . This method is simpler in hardware. In this case, the register of FIG. 3 does not branch the data to be sent to the optical transmitter, but contains a predetermined pattern. Alternatively, instead of the register and the OR, a multi-input AND circuit is used as shown in FIG. 3B, and the output becomes 1 only when it matches a certain pattern determined with or without the input NOT. If you use it, it will be easier hardware.

The first aspect of the present invention can transmit connectionless data with a very small delay when the network load is very small, and the processing amount of the node is relatively small. Next, the second invention will be described.

Slot setting for connection-oriented calls is performed according to a certain rule. For example, if a slot setting request is made and a common empty slot is searched for, and more than the requested number of common empty slots are found, the NWC selects the slot in front of the frame. In addition, each node transmits to the same destination node in several slots as shown in FIG. 4, and if one of the calls transmitted in that node ends to reduce the number of slots, the frame For example, release from the rear side. According to such a rule, the set slots are easily gathered in the front of the frame, and the empty slots are concentrated in the rear of the frame, and the common empty slot is easily taken.

In such a network, even in the empty slot of the connectionless data transmission node, the empty slot near the front is surrounded by the slots for connection oriented calls all around like the slot of the arrow in FIG. There is a low probability that it will be a common empty slot with the loss data receiving node. If such a slot with a low probability of success is sent avoiding the slot from the beginning, the probability of successful packet transmission will increase, and the retransmission procedure will decrease, reducing the load on the node.

For example, suppose ni wants to send connectionless data having an information amount of 3 slots to nj.
When the slot setting situation of i is as shown in FIG. 5, ni selects and sends the three closest to the rear of the frame as shown.

Further, when ni declares on the control channel, it can be said that the connectionless data transmission is performed in the slot number and after, and nj is the reception empty slot of the own node of the slot number before that. There is no need to tune to the ni wavelength.

The operation of receiving nj, the method of returning ack, etc. are the same as in the first invention. Network load is low,
A system with a very high probability of success will not return ack,
If the retransmission control in case of failure is left to the upper layer protocol, the processing of the node can be reduced.

The second aspect of the invention can reduce the total delay including the retransmission when the load on the network becomes slightly heavy (compared to the most efficient load situation of the first aspect).

Next, the third invention will be described. Figure 6
FIG. 3 is an explanatory diagram showing an example thereof. It is assumed that the nodes ni to nj transmit connectionless data having an information amount of four slots (divided into four packets). The number of slots in one frame in FIG. 6A is 10, and there are exactly four empty transmission slots in the connectionless data transmission node ni.

In the case of (a), first, after declaring on the control channel, four packets are sent by using the four empty transmission slots of the own node, and the packet sequence is continued as it is, as shown in the figure. It has been changed twice and sent a total of three times in a row. In three transmissions, the same packet is not sent in the same slot. Packet destination node n
If the reception slot of j is as shown in (b),
First, packets 3 and 4 are received in the first transmission, 2, 3 are received in the second transmission, and 1 and 2 are received in the third transmission.
Although 2 and 3 are received in duplicate, all packets are received once. nj is a received packet, and the redundant excess is discarded. In a system with a poor signal-to-noise ratio, among the duplicated ones, the one with the poor reception condition may be discarded. Then, the order of the entire packet is reassembled. Although a packet number, an ID number, etc. are used, a specific method is the same as that of the first invention.

The number of slots in one frame is larger,
If the number of connectionless data packets is also large, instead of shifting the order one packet at a time as in (a),
It is better to divide the packet into about three groups and rotate the position within the frame of the group as shown in (c). This is because the position in the frame of the slot occupied by the connection-oriented call is NW.
This is because, depending on the algorithm used by C, there is a possibility that there is some tendency such as concentration in the front of the frame.

(D) is the case where the number of empty slots of ni is larger than the number of packets. It can be seen that when repeatedly transmitting, even if the order of the packets is not changed, the slots for transmitting the packets are changed by continuously transmitting the packets. The same thing happens when the number of empty slots is smaller than the number of packets. Than in case (a)
The processing performed by ni is small.

(E) is the case where the number of packets is one,
It is sent 3 times by changing the location within one frame. Again, in consideration of the tendency of the NWC algorithm, transmission at slot positions that are too close to each other is avoided.

In the above, the transmitting node selects the appropriate method in view of the relative relationship between the number of packets and the number of empty slots, and the same packet is transmitted without being transmitted in the same slot. The number of transmissions may be fixed, or may be appropriately selected depending on the load status of the network. Receiving node n
When j has received all the packets declared at the time of declaration, the reception of connectionless data from ni is stopped. How long the receiving node nj continues the receiving operation is the same as in the first invention.

Further, there may be no common empty slot at all. When the receiving node nj tries to receive one frame and no packet is received, the common empty slot becomes 1
You can see that it's unbearable. In such a case, it is advisable that nj stops the receiving operation and informs ni through the control channel that it cannot receive. By doing this, unnecessary transmission is avoided.

In the case of this protocol, ack may be returned in the same manner as in the case of the first invention, and if it is not received, the network may perform retransmission control. Also, ack may be returned only when all packets are received. However, since the success rate is very high when the load on the network is relatively small, in such a network, retransmission control may be left to the protocol of a higher layer and ack may not be returned. Alternatively, the terminal that outputs connectionless data may be allowed to select whether to perform retransmission control on the network or on the terminal side.

The third aspect of the present invention is effective in a system in which retransmission is performed without waiting for ack, so that the propagation span is large because the span of the network is large, and it takes a long time to exchange ack. Further, connectionless data transmission with low delay can be performed with a larger load than in the load situation in which the protocols of the first and second inventions are most effective.

Next, the fourth invention will be described. The method so far can transmit connectionless data with a small delay when the load on the network is small or medium. However, when the load is large, many packets are not received, and as a result, the total delay or discard rate increases. In a heavily loaded network, using a method similar to establishing a connection to secure bandwidth will result in a smaller delay. The fourth invention is such a method.

A slot for a connection-oriented call is applied to the NWC and reserved by searching for a common empty slot, but a similar method can be used for connectionless data. Suppose ni wants to send connectionless data to nj. The ni notifies the NWC of that fact through the control channel, searches for a common empty slot, and tells through the control channel. ni transmits the connectionless data by the same procedure (declaration and transmission) as the first invention in the common empty slot that was taught.

In this way, it takes time before the initial transmission because of the exchange with the NWC. However, the success rate of packet transmission is very high, and it hardly depends on the network usage. It is wasteful to perform this protocol on a very lightly loaded network because it takes a long time, but on a heavily loaded network, the total delay is reduced as the retransmission probability becomes smaller, which is very effective.

On the receiving side nj, as in the first aspect of the invention, tuning is performed to ni for all the reception empty slots of its own node. Alternatively, if nj also hears that the NWC teaches ni a common empty slot, then nj can know in which slot ni is transmitting. If reception is performed based on this, tuning can be omitted in unrelated slots. Alternatively, the NWC may directly instruct the receiving slot to the nj, or the ni may inform the nj of the transmitting slot at the time of declaration.

Further, the NWC may instead make the declaration that the ni should make on the control channel. That is, ni declares information such as the destination node and the amount of information (the number of packets) of connectionless data to the NWC.
C searches for a common empty slot and instructs ni to send connectionless data to nj in which slot from when to when, and also to nj to receive connectionless data from ni. "From when" is necessary because there is a propagation delay from the NWC to each node. In this way, the bandwidth of the control channel can be saved.

This protocol is "from when to when"
It is almost equivalent to establishing a connection except that there is a limitation. However, since the period is limited, it is possible to reserve another connectionless data transmission / reception before and after that.

In the above-mentioned fourth invention, particularly in the protocol in which only the common empty slot is taught, there is a possibility that the declarations to the same destination node as other nodes may not be received because they may not be received, so return ack. Retransmission control may be performed. However, since the success probability is very high, ack may not be returned in the network, and transmission of ack and retransmission control may be left to the upper layer.

In the embodiments described above, it is assumed that the frequency of occurrence of connectionless data is low, and when the declarations to the same destination node as other nodes overlap, or a plurality of connectionless data to be transmitted are simultaneously generated in one node. If not, it is not stated. In systems where connectionless data transmission occurs very frequently, countermeasures are needed.

In case of the former case, if the empty slot number of the own node or the slot number for transmitting connectionless data is declared in the declaration, even if the destination node overlaps with the connectionless data transmission of another node. The destination node side can arbitrarily calculate a common empty slot with each transmitting node and receive only that slot.
Common empty slots may overlap, but in order to deal with such cases, the one declared in the early control frame (not the slot) has priority, the random number is given to determine, and the priority is declared at the time of declaration. A method such as deciding based on that can be considered. In the fourth invention, it is not particularly necessary when the receiving node is instructed to receive a slot.

When a plurality of connectionless data are simultaneously generated in one node, they are declared at the same time, and the transmitting node appropriately allocates the transmission empty slots of its own node to the plurality of connectionless data and transmits them. . Further, if the system is capable of instructing the delay deadline, priority, etc. when the terminal outputs connectionless data, a method of processing in time series based on such information can be considered. In the case like the fourth invention,
Only when the designated common empty slot overlaps, the above may be done for the overlapped slot.

In all of the above embodiments, the data transmission wavelength of each node is fixed and different for each node. However, in the optical wavelength division multiplexing network in which a large number of nodes are required to be connected, the number of nodes is not limited in the wavelength range. As described above, some transmission wavelengths are variable.

The connection-oriented call protocol, which is the premise of this patent described in the conventional example, is modified as follows in the case of variable wavelength. After the NWC receives the slot setting request and searches for a common empty slot for transmission / reception, it searches for an unused wavelength in the common empty slot. In most cases, there are more than the number of common empty slots required, so
It suffices if the number of common empty slots having unused wavelengths is the required number (FIG. 7). The NWC also indicates the wavelength to be used when instructing transmission / reception on the control channel.

In order to perform the connectionless data transmission protocol of the present invention in such a system, at the time of declaration,
You can specify the transmission wavelength, but if you select the appropriate wavelength, it will overlap with the slot for transmitting connection-oriented calls, and not only connectionless data but also connection-oriented calls that should have guaranteed bandwidth will be destroyed. Will end up.

In the case of the fourth aspect of the invention, the wavelengths usable in the NWC may be designated. But in other cases,
I do not know which slot of which wavelength is in use and which is not. As in the case of distributed systems, the active slots of each wavelength may be broadcast using some of the control or data channels, but that requires a lot of bandwidth, and It is not a good idea to read each and decide which wavelength to use in which slot because it requires a lot of processing.

As a solution to this problem, it is considered that the frequency of connectionless data transmission is not so high in a connection-based network, so it is advisable to prepare a wavelength dedicated to connectionless data transmission. In this way, even if a collision occurs, only the connectionless data will collide with each other and the connection-oriented call will not be broken. A collision does not destroy all packets of the connectionless data.

When there is one wavelength for connectionless data, it is not necessary to specify the wavelength in the declaration, but when there are multiple wavelengths, it is declared which one to transmit. The receiving side tunes to the wavelength when performing the receiving operation.

When the receiving side detects a collision, it may assume that the packet could not be received. Alternatively, a notification may be issued through the control channel that where and where the slot has collided. For example, let us say that what ni put out to nj overlaps what np put out to nq in the slot sa. This is because only one of the empty slots for transmission of ni and np overlaps with the common empty slot of ni and nj.
The common empty slots of p and nq do not always overlap. That is, even if nj can receive at sa and detects a collision, nq may not have performed receiving operation at sa because sa is not in the reception empty slot of its own node.
This can be understood by carefully listening to the control channel for ni and np. For example, for ni, nj is ni
You can tell if someone says that you have detected a collision in the same slot that you said you detected a collision in. If no one, other than nj, says there was a collision detected in that slot, then it can be determined that the slot was not a common empty slot for the other pair that had a collision. Therefore, using the slot for retransmission (unless it hits another new connectionless data transmission) does not cause a collision.

If two sets of ni and nj and np and nq are common empty slots, and if it is found that both of them have detected a collision, several countermeasures can be considered. The simplest way is that neither ni nor np uses that slot for retransmission. Alternatively, after confirming the probability of collision again, ni and np may each be given a random number to decide whether or not to use the slot for retransmission.

Carefully listening to the control channel to confirm the collision detection of another node requires a lot of control channel processing. In a system that does not want to perform such processing, if the destination node says that a collision has occurred, do not use that slot, or do not worry about other nodes and send a random number and retransmit it. You may decide whether or not to use it.

When a random number is assigned, if the number of collided nodes is known, the random number to be used for retransmission with a probability of 1 / the number of collided nodes may be mapped. . However, if a random number is given without knowing how other nodes are, the ratio of 1 / decided to be used for retransmission can be determined by taking into account the statistical probability that other nodes may not be common empty slots. It may be slightly larger than 2.

In such a system, it must be known whether a collision has occurred. Therefore, it is necessary to add an error detection code and check whether the S / N is deteriorated by the beats of the optical carriers.

Up to this point, the control channel has been assumed to use one wavelength-multiplexed light wave in TDM.
Separate lines may be used, or subcarrier multiplexing may be used. Alternatively, not only one wave of the wavelength-multiplexed light but also several waves may be used .

Up to this point, the control channel has been described in a system in which the transmission timing (slot) in the frame is determined for each node. As shown in FIG. 8A, the control channel frame has not only the time for each node to transmit, but also the time for the NWC to transmit to each node. NWC
In some systems, the information sent to multiple nodes may be complicated and the information sent to each node may be complicated, but each node receives only a part of the slots of the NWC sent to the own node. A system such as that shown in FIG. At this time, the node desiring to transmit the connectionless data makes a declaration in the reception slot of the destination node as shown in FIG. 8C. A system in which ack is also performed in the receiving slot of the other party is also conceivable.

In this case, the declaration content includes not the destination node number but the transmission node number. In such a case, the declaration and ack may collide with the declarations and ack of other nodes, and the collision may not be received at all, but the probability is low if the system does not transmit connectionless data frequently, When the receiving slot of the own node is determined, it is clear which of the control channel information is related to the own node, so that the processing amount of control information can be small.

Further, in the system using the transmission from the NWC for clock extraction, as shown in FIG.
It is not preferable that the transmission of the data is interrupted and the information whose clock phase is out of phase is entered. In such a case, as shown in FIG. 9, if the control channel frame is partially divided for connectionless data declaration and ack, the declaration and ack can be transmitted and received without significantly increasing the control information processing amount of the node. You can At this time, in the part divided for connectionless data, ALOHA, slotted-A
It is preferable to use a protocol such as LOHA.

In the description so far, each protocol has a transmission performance (success rate, success rate,
Delays etc) are different. It suffices if the network is one in which the load status does not change so much, but if not, the load status may change and transmission performance may drop. A system that is expected to do this costs a little, but it may be possible to use several protocols depending on the load situation. Further, even if the load is the same, the performance varies depending on the protocol, so that the connectionless data may be selectively used depending on the required performance.

In the system up to this point, we have considered a topology centering on a star coupler,
The frame synchronization can be taken in the same manner as the ring, as shown in FIG.
A ring or tree as in (b) can also be implemented.

[0094]

As described above in detail, according to the present invention, it is possible to transmit connectionless data with low delay in a connection-based centralized control type optical wavelength division multiplexing network without establishing a connection. In addition, the network cost can be optimized by selecting the optimum protocol in consideration of the use of the network, the load status, and the like.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a first invention of the present application.

FIG. 2 is a time chart until ack is returned.

FIG. 3 is an example of a system for measuring the time to the star coupler.

FIG. 4 is a diagram for explaining a tendency of a slot setting algorithm.

FIG. 5 is an explanatory diagram of a second invention of the present application.

FIG. 6 is an explanatory diagram of a third invention of the present application.

FIG. 7 is a diagram illustrating a case of variable transmission wavelength.

FIG. 8 is an example of a time division form of a control channel.

FIG. 9 is an example of a time division form of a control channel.

FIG. 10 is a diagram showing a form of a network which is a premise of the present application.

FIG. 11 is a diagram showing a wavelength / time division mode of a network which is a premise of the present application.

FIG. 12 is a diagram showing a form of a network to which the protocol of the present application is applicable.

[Explanation of symbols]

1 ... Optical transmitter 2 ... Optical receiver 3 ... Counter 4 ... Register 5 ... Shift register 6 ... Multiple input or 7 ... exor 8 ... Multi-input and 9 ... Network controller 10 ... General node 11 ... Optical fiber 12 ... Star Coupler 13 ... Transmitter for control channel 14 ... Transmitter for data channel 15 ... Receiver for data channel 16 ... Control channel receiver 17 ... WDM coupler

Claims (4)

(57) [Claims] 1. A plurality of nodes comprising a data transmitter for transmitting information data using a specific wavelength, a data receiver for receiving the information data transmitted using a different wavelength, and a means for accessing a control channel. Each wavelength is composed of a frame composed of a plurality of time-divided slots, which are connected to a star coupler, and each band management node uses each slot based on a request made from the control channel by each node using the access means. In the optical wavelength division multiplexing network that performs reservation processing and uses the reserved slot until cancelled, when each node tries to transmit the connectionless information data, each node uses the control channel. , The information related to the connectionless information data and the destination node are transmitted, When the destination node receives the information data transmitted from each of the slots, the ACK information is transmitted to the corresponding slot using the control channel. The node that transmitted the connectionless information data is transmitted to the node
The information data to be retransmitted when no information is returned is returned to the ACK information.
The slot in which the information data returned was sent
Connectionless data transmission method which is characterized in that in re-transmission. 2. The slot used by each node for transmitting the connectionless information data is preceded by an algorithm used by the bandwidth management node in setting a slot.
The destination node is a slot with a high probability of being an empty slot.
2. The connectionless data transmission method according to claim 1, wherein the connectionless data transmission method is determined as a time period . 3. A plurality of nodes having a data transmitter for transmitting information data using a specific wavelength, a data receiver for receiving the information data transmitted using a different wavelength, and a means for accessing a control channel. Each wavelength is composed of a frame composed of a plurality of time-divided slots, which are connected to a star coupler, and each band management node uses each slot based on a request made from the control channel by each node using the access means. In the optical wavelength division multiplexing network that performs reservation processing and uses the reserved slot until cancelled, when each node tries to transmit the connectionless information data, each node uses the control channel. , The information related to the connectionless information data and the destination node are transmitted, After transmitting the information data using a Tsu bets, connectionless data transmission method and transmitting one or more times so as not to transmit the same information data at the same slot. 4. A plurality of nodes comprising a data transmitter for transmitting information data using a specific wavelength, a data receiver for receiving the information data transmitted using a different wavelength, and a means for accessing a control channel. Each wavelength is composed of a frame composed of a plurality of time-divided slots, which are connected to a star coupler, and each band management node uses each slot based on a request made from the control channel by each node using the access means. In the optical wavelength division multiplexing network that performs reservation processing of the reserved slot and uses the reserved slot until it is canceled, when each node attempts to transmit the connectionless information data, the bandwidth management is performed using the control channel. After receiving the information of the slot that can be transmitted from the node, Connection-less, characterized in that information related to the connectionless information data and information related to the destination node are transmitted using a channel, and the information data is transmitted using a slot corresponding to the received slot information. Data transmission method.