System and method for transferring data packets through a communication system

Field of the invention

0001The invention relates to a communication system, ports in a synchronous communication system and a method for communication in a synchronous communication system, the synchronous communication system formed as a ring network of two or more ports coupled in daisy chain fashion to one an-other to allow communication.

Description of the related art

0002A communication system is generally known as a system that permits communication between nodes interconnected by a transmission line. Each node can transmit information and receive information across the transmission line. The communication system of interconnected nodes can be organized in various topologies, such as bus, ring, star, or tree topology or a combination thereof.

0003A bus topology network is generally regarded as linear. Transmissions from one node propagate along the transmission line and are received by all other nodes connected to that bus. A ring topology network, however, generally consists of a series of nodes connected to one another by unidirectional transmission links to form a single, closed loop. Examples of a ring network are described in IEEE 802.5 and Fiber Distributed Data Interface (FDDI).

0004The transmission line between nodes can be either wired or wireless, for example, copper wire, fiber optic, or wireless transmission medium for the chosen transmission line, respectively.

0005A communication system, for real-time applications or for transferring synchronous streaming data must have a low latency and a low transmission overhead.

0006Ethernet and IEEE 802.03 specify a particular protocol in which packets of data can be sent between computing systems. Ethernet can sense multiple access collisions and can arbitrate which source device will gain mastership over the transmission line. Ethernet operates at the lowest levels of the OSI reference model, normally reserved for the data link and physical link layers. The Ethernet protocol specifies a particular frame format of a preamble, followed by a destination address and a source address and then the data payload. The data is generally encoded in a 4B/5B or 8B/10B encoding structure prior to the data being sent across a coax or twisted pair transmission line. On detection of a collision, a jam signal is transmitted to inform other nodes that a collision has occurred. A hub or a repeater will forward the jam signal on all ports, thereby informing all other nodes about the collision and forcing them to wait until the next transmission. The purpose of this jam signal is to extend a collision significantly, so that all other nodes on the network cease transmitting. Jamming is also used when dealing with congestion. It is an attempt to eliminate frame loss within a node by applying "back pressure" to other nodes consuming the node's buffer capacity. One way of accomplishing this is for a node to issue an Ethernet jam signal when buffers fill beyond a threshold level. Using the Ethernet jam signal makes a network rather indeterministic, as the forced delay for retransmission is a minimum fixed delay permitting all other nodes to cease transmission plus a certain random delay time. Furthermore a single slow node can slow down the whole network.

0007<nplcit id="ncit0001" npl-type="s"><text>A. Tanenbaum, "Computer Networks", 2003, pages 333 - 336</text></nplcit> discloses a frame according to the IEEE802.1Q standard, where additional fields are inserted into a frame for signaling VLAN information for switching.

0008<nplcit id="ncit0002" npl-type="b"><text>Stallings, W.: "Handbook of Computer Communications- LAN", 31 December 2003, page 160-161</text></nplcit> discloses IEEE802.5 frames having an access control bit which contains the priority and reservation bits which are used in the priority mechanism and the monitor bit used in the ring maintenance mechanism.

0009<patcit id="pcit0001" dnum="US6170022B"><text>US 6,170,022</text></patcit> discloses a mechanism for handling network overloads by sending messages to individual network nodes according to the generated node.

0010<patcit id="pcit0002" dnum="US20030156542A"><text>US 2003/0156542</text></patcit> discloses a token ring in which a frame is sent from an end-point (140) to a switch and ultimately to an endpoint (120) including a congestion indication.

Summary of the invention

0011The problem to be solved by the invention is to improve communication performance on communication systems as described above. Especially latency and overhead should be significantly reduced over the prior art to a minimum.

0012Solutions of the problem are described in the independent claims1,2 and 8.

0013For synchronization of the data stream a preamble, prior to the start identifier may be sent. The start identifier itself may comprise one start byte according to be Ethernet standard, but any other size may be appropriate.

0014The destination address field comprises a unique address for identifying the receiver of the frame. Alternatively the address field may contain a plurality of addresses, a multicast or a broadcast address. The destination address field may have a length of 6 bytes according to the Ethernet standard.

0015The preemptive acknowledge field is used to give arbitrary information about the receiver buffer to the transmitter. It may comprise only a single bit, alternatively it may comprise one byte or any number of bits. The preemptive acknowledge field is also referred as PACK field. It's function will be explained in detail below.

0016The data field is commonly referred to as the payload of the frame, whereas the preceding fields, i.e. start identifier, destination address and the preemptive acknowledge byte are also referred to as header. The data field may have a fixed or a variable length. There may also be a length identifier in the header. In the Ethernet standard such an identifier is implemented having two bytes in length and specifying the length of the data field in bytes.. Furthermore according to be Ethernet standard the data field may have a size from 38 to 1500 bytes. Of course any other size may be choose and if appropriate. Optionally there may be a plurality of data fields.

0017The data field may be succeeded by a trailer, which may be a checksum, for example 4 bytes in the Ethernet standard.

0018All the fields described above are necessary for the invention.with the exception of the data field. Of course a frame without any data field and therefore any payload usually makes no sense, except when it is used for signalling purposes. Of course there may be additional fields in the frame.

0019According to the invention the preemptive acknowledge is not a separate message, instead it is part of each data frame. It is placed against prior to the data field to allow any receiving port to take an action before the data field is transmitted or received. To make a frame fully compatible with the Ethernet standard, it is not possible to insert an additional preventive acknowledge field. Instead other fields may be used. Such a field must be located after the destination address and before the payload or at least at the beginning of the payload. For example the data field length identifier may be used. It could be set to a non-a defined value, giving a not allowed data field size, if the second port can not accept any more data. As an alternative one or more bytes at the beginning of the data field may be used. To this end the frame could be extended for one or more additional bytes.

0020These frames are assembled by the port. A separate framer may be included in the port. Subsequently to the assembly of the frame, by e.g. a framer, the frame is transmitted by the first port via a transmission line. A second communication port receives data from the communication line. This communication port has a frame buffer for storing frames and a decoder for disassembling the frames or at least parts thereof. When receiving data the second port may first synchronize on a preamble, if available. Then the second port receives a start byte following the preamble. Next it receives the destination address. After evaluating the destination address, the second port can now determine whether the frame has to be received or not. When the frame has to be received, the second port has to check for available buffer space. If there is enough buffer space, it signals, that it can receive the frame by sending a predetermined pattern in the preemptive acknowledge field.

0021If the frame size can vary largely, the preemptive acknowledge field may be preceded by a size identifier, which preferably identifies the size of the data field or of the whole frame. This simplifies determining whether the second port has enough buffer space.

0022The ports may as well be connected to a ring network. The major difference between a ring network and a shared transmission line is that in a ring network each node forwards a received frame to the next node in the ring, whereas in the case of a shared transmission line no forwarding is necessary, as all nodes have access to the same data on the same line. In general first and second ports should be connected in such a way, that the second port may modify a frame sent by the first port and this modified frame can be received by the first port again. This is usually not the case in switched networks and in wide area networks. Especially in the latter case received frames are not forwarded to the sender again.

0023The second port is adapted to modify the frame from the first port. This can be done in various ways. According to the invention a predetermined bit pattern is put on the network to inform the first port about the status of a second port's buffer. The status can be for example buffer full, buffer empty or the size of available buffer space.

0024In the case of a bus where each node is connected only to a switch or router, each node may forward received frames to another node, which may be the sender of the frame.

0025In the case of a ring bus each node must forward received frames to permit communication to neighbour nodes. This is at least the case in broadcast or multicast frames, which must be sent to several nodes. In these cases, the forwarded frame may simply be modified by exchanging bits in the preemptive acknowledge field.

0026In an example, the second port may be adapted for generating a new frame having a modified preemptive acknowledge field. The new frame may either have the specified minimum frame size or it may even be terminated immediately after the preemptive acknowledge field. This works in most networks, except in a ring bus, as there the received data often has to be forwarded to a neighbour node.

0027In a further embodiment of the invention the second port is adapted to modify the preemptive acknowledge field in the frame, only when its frame buffer can store a full frame.

0028In another embodiment of the invention the second port does not alter the preemptive acknowledge field, if the frame buffer is full or if it has at least reached a certain limit.

0029Alternatively the second port may be modified such that it signals an overrun of the receive buffer by modifying the preemptive acknowledge field in a different way as signalling an empty receive buffer. Furthermore it may signal the size of the receive buffer, for example by returning the number of available bytes.

0030In another embodiment of the invention the second port may be configured to signal a delay time in the preemptive acknowledge field. Preferably this delay time signals an estimated time the second port will require to free the receive buffer.

0031In the invention the first port is adapted for checking the preemptive acknowledge field for modifications by the second port. Therefore it may compare the received preemptive acknowledge field with the origin of the transmitted value or with a list or table of allowed values.

0032In a further embodiment of the invention the first port is adapted for transmitting of further frames only for the case that the preemptive acknowledge field signals that the second port frame buffer can store at least a full frame.

0033In another embodiment of the invention the first port is adapted to delay further frame transmissions for a specified time. An instruction to delay further transmissions may be encoded by the preemptive acknowledge field. The preemptive acknowledge field may contain a delay time, which may be calculated by the second port, such that the buffer is expected to be emptied after this delay time. In another embodiment the delay time may be a random time, being calculated by the first port or by the second port. This prevents sending more and more frames to the second port, causing a buffer overrun.

0034In a further embodiment the first port is adapted for reducing the frame rate in case the preemptive acknowledge field has not been modified. By this way it is attempted to communicate with non-responding devices.

0035A further embodiment of the invention comprises a first port being adapted for entering a sequence to handle an exception for a second port being not responsive. This sequence is entered when the preemptive acknowledge field has not been modified. For the case, the second port is responsive, it will modify the preemptive acknowledge field by either signalling a full or an empty frame buffer.

0036Another embodiment of the invention comprises a first port being configured for transmitting a frame having the size of the available buffer space of the second port.

0037A communication system according to the invention comprises at least a pair of communication ports as described above. Preferably there is a higher number of communication ports.

0038The major advantage of the invention is, that the second port can during the transmission of a frame check for a buffer overflow and accordingly signal this to the first port sending the frame. When the second port receives a frame it can evaluate the address field and check, whether it has to receive and to store the frame into a frame buffer. It then can check for available buffer space. If the buffer space is not sufficient for a frame, it can immediately signal this information about the second port's buffer status to the first port by modifying the preemptive acknowledge field. The frame is again received by the first port, evaluating the preemptive acknowledge field. If a buffer overflow (overrun) is signalled, it can stop transmitting further data of the frame immediately. Therefore in most cases the preemptive acknowledge field may be the last field of frame, as the transmission of the frame is aborted. This frees the communication system and the transmission line immediately and makes it available for other communication and other frames. Therefore the inventive ports and communication system results in a low latency of the bus system in case of buffer overflow of a receiving port.

Description of Drawings

0039In the following the invention will be described by way of example, without limitation of the general inventive concept, on examples of embodiment with reference to the drawings. <ul id="ul0001" list-style="none"><li><figref idref="f0001">Figure 1</figref> shows four nodes connected to a ring bus.</li><li><figref idref="f0001">Figure 2</figref> shows a single network node more detailed.</li><li><figref idref="f0002">Figure 3</figref> shows an exemplary frame according to the invention.</li><li><figref idref="f0002">Figure 4</figref> shows a standard Ethernet frame.</li></ul>

0040In <figref idref="f0001">figure 1</figref> a simple ring bus comprising four nodes 10, 20, 30, 40 is shown. Each node has a port with an input and an output as shown in <figref idref="f0001">figure 2</figref>. The transmitting port of each node is connected to the receiving port of the next node, therefore resulting in a closed loop. If node 10 will send a frame to node 30, it transmits each bit of the frame, one after one, on its port 12 to node 20. This node receives the bits by its port, checks the address field, determines that this node is not addressed and forwards the bits by its transmit port to node 30. Node 30 receives the bits by its port. After checking the address field it knows that it is the receiver of that frame. Therefore it checks its frame buffer for free space to buffer a frame. If there is enough buffer space, the PACK field of the frame is set to a first value and the incoming bits are stored in the frame buffer. If there is not enough buffer space the PACK field of the frame is set to a second value and the incoming bits are discarded. The bits of the frame with the modified PACK field are then forwarded by the transmit port to node 40. This node receives the bits by its port, checks the address field, determines that this node is not addressed and forwards the bits by its port to node 10. This node receives all the bits with its port and may compare the received bit sequence with the transmitted sequence. When it recognizes changed bits in the PACK field of the frame, signalling that node 30 cannot buffer the frame, it immediately stops transmitting the rest of the frame and frees the bus. For the case the PACK bits have not been modified at all, node 10 may assume that node 30 is not available at all and it may also stop transmission.

0041In <figref idref="f0001">figure 2</figref> a more detailed view of a network node is given. The network node 10 has an input 11 for receiving signals from other nodes and an output 12 for transmitting signals to other nodes.

0042In <figref idref="f0002">figure 3</figref> an exemplary frame according to the invention is shown. This frame comprises the fields Start: start byte, Destination: destination address Pre-emptive ACK (PACK): preemptive acknowledge Length: number of bytes in the data field Data: data field (payload) CRC: checksum for the frame or the data field End: end byte

0043The right column of the table specifies the number of bytes in the field. Compared to the standard Ethernet frame in <figref idref="f0002">figure 4</figref>, there is an additional preemptive acknowledge field. Optionally there may be included a source address field identifying the source of the frame after the destination address field.

0044In <figref idref="f0002">figure 4</figref> an Ethernet frame is shown. This frame comprises the fields Start: start byte, Destination: destination address Source: source address Length: number of bytes in the data field Data: data field (payload) CRC: checksum for the frame or the data field End: end byte

0045The right column of the table specifies the number of bytes in the field.