WO2014173438A1 - Method and apparatus for transmitting a data packet in a communication network in the event of buffer overflow - Google Patents

Abstract

In standard communication networks, data packets are transmitted from network node to network node. In the process, it is accepted that packets are lost during transport and need to be retransmitted. The method and the apparatus specify a strategy for how to deal with the case in which packets would need to be rejected at a network node on account of a lack of available memory. The essential inventive concept is that of not rejecting the data packet in question but rather of transmitting it back to the sending network element, and in this way effectively buffer-storing it in the network.

Description

description

Method and device for transmitting a data packet in a communication network in the event of a buffer overflow

The use of communication networks is now common in many applications. In addition to classical audio and video communication between people, another important use case is the exchange of data between machines or humans and machines.

Packet switching is one of the most frequently used methods of data transmission in computer networks. Longer messages are to split into individual data packets and either connectionless or connection-oriented übermit ^¬ telt.

In packet switching, the data packets traverse the communication network as independent and independent units and can be temporarily stored in the (switching) network elements. At a network element is understood as any device which is part of the communication network and the two or more transmission paths of a Telecommunication network ^¬ zes connects, for example, a switch, a bridge, repeaters ter, hub, router or the like.

This approach has the substantial advantage that to the delegation ^¬ transmission speed between individual links is no longer a limitation. Each must be passed network element receives the data packet and forwards it to its initial cut parts (the network element) that can be target of many Sen ^¬ applications though, so the tendency to overload situations arises. The whole thing can therefore be represented as a network ^¬ work of queues. The existence of the queues on the one hand produces delays in the delivery package or packet loss, the repetitions and thus a further Be ^¬ utilization of the transmission path have the effect. For transmitter and receiver, this process is not transparent, that you needed ^¬ gen and receive no information about the transmission gungsweg. As a rule, this information can not be obtained, since the transmission paths can also change dynamically.

State of the art

Network elements that forward data packets can be flooded by too large a volume of packets. This occurs usual cherweise always when arriving at a network element more Since ^¬ tenpakete than it can handle.

FIG. 1 illustrates by way of example what happens in a network element:

Shown is a small communication network with three Sen ^¬ countries (EG1, EG2, EG3) and a receiver (EG4), as well as several switching network elements (SW1, SW2, SW3). All incoming data packets are stored in the network element (SW1) to the proces ^¬ processing (transfer to the next network element) in a data buffer (BUF). In the example shown in the figure, the terminals EG1, EG2 and EG3 send all data to the terminal EG4 (eg sensors send measured values to a processing PLC). If more packets arrive at the network element than it can process directly, these data packets must be stored in the data buffer (BUF). Since the data buffer capacity has only a limited Speicherkapa ^¬, it can come to buffer overflow, that is, the data buffer is at some point full and can not accept any more data packets.

When the data buffer is fully occupied and now more Since ^¬ tenpakete arrive so would these can not be processed by the receiving network element, ie the network element would discard these data packets (ie delete). This leads to the loss of these data packets (DEL). This packet loss is contributed by the sender of the packet to either protocol ^¬ plane (z. B. by an automatically initiated retransmission based on a unique packet number, For example, by the TCP protocol) or on application ^¬ level z. B. treated by retransmission. As a result, the sender retransmits the packet to the receiver after a certain time, resulting in an even higher network load.

It is an object of the invention to provide a method which uses an advantageous strategy when forwarding Datenpake ^¬ th. It is another object of the invention to provide a device for carrying out the method.

DESCRIPTION OF THE INVENTION According to the invention, the object is achieved by a method having the features of patent claim 1.

The inventive method for transmitting a Datenpa ^¬ kets in a communication network, which consists of individual mutually independent network elements, of which at least some for buffering the data packets have a data buffer, from a sender to a receiver, wherein the data packet from the network element (Switch) after receiving wei ^¬ ter conductive or in the data buffer can be stored.

If at least one data packet can not be ^temporarily stored, because there is insufficient storage space in the data buffer, the received data packet is sent back to the transmitting network element by the receiving network element according to the invention.

The object is further achieved by a device having the features of patent claim 6.

The network element according to the invention for transmitting Datenpa- ketene in a communication network from a transmitter to a receiver includes receiving means, transmitting means and an Since ^¬ TEN buffer, wherein a signal received from an adjacent network element by means of the reception means after reception of data packet can not be stored in the data buffer and wherein the transmitting means of the receiving network element send the received data packet back to the sending network element.

Further embodiments are given in the dependent claims ^¬ .

In an advantageous embodiment, each additional EMP captured data packet is sent back to the respective transmitter, after a first received data packet from the ^¬ receive the network element has been sent back to the sending network element. This is carried out until the first received ^¬ ne data packet could be transmitted successfully to the next network element.

It is also advantageous to carry out a monitoring of the mid ^¬ lung try of the data packet by measuring how much time and / or how many delivery attempts be taken benö- to conduct the first data packet to the next Netzele ^¬ ment on.

If the monitoring indicates that the transmission of the data packet (P3) takes too much time and / or too many delivery attempts, may be the first data packet are deleted in another embodiment of the invention, and so an end-to-end transmission by means of the normal data transmission Proto ^¬ be stimulated. The method and the device work in a further embodiment according to the IEEE 802 Ethernet standard.

So far it is the case that switching network elements with full buffer utilization discard all further incoming data packets. Thus, the communication ^performance already provided for the packet is not taken into account. The problem affects every network element that forwards packets in a network, such as: For example, a switch or router. The problem affects all types of Ethernet networks:

wired and wireless networks as well as office and industrial networks.

One problem lies in the fact that with the dropping of packets in network nodes also already provided communication Leis ^¬ obligations eg for forwarding by previous Netzwerkge- discard rate. Under communication performance is the effort to understand that already spent by the transport of the data packet in the network and the network nodes who ^¬ had to the current time. The discarding of a packet thus automatically results in that all of the already provided communication performance must be provided by repeating the packet transmission again through the network. (More precisely: all nodes on the network that have already passed this package), the more Kommunikati ^¬ kiln output already in a packet through the network: It is true was "invested", the more "valuable" the package for the network or otherwise is said the more important it is for the network that the packet is transported to the destination without retransmission of the packet due to packet loss due to buffer overflow.

The idea is now to send back the last received data packet (the packet discarded in the prior art) to the transmitter. The term transmitter is understood to mean the immediate predecessor on the transport path of the packet, that is not necessarily the actual original transmitter of the data packet. This reduces the number of data packets to be discarded in the network due to buffer overflow to a minimum.

This approach means that the predecessor of this data ^¬ packet sent back to the switch, or possibly also seeks a different path through the communications network and transmits the data packet to another network element, wel ^¬ Ches is not overloaded. The time gained in this way makes it very likely that the package will be successfully delivered on the next attempt. The data packet is cached between the participants "on the road", as it were.

If several parcels are received by different transmitters at the same time, this procedure leads to a kind of automatic "round-robin" procedure, so that starvation or endless circulation of parcels becomes very unlikely all new packets from other predecessors are in an advantageous embodiment, these returned to the first data packet of the first predecessor to its target could be forwarded success ^¬ rich. Then, the packet of the next predecessor is to transmit it, etc.

Solves the overload problem for a long time not, so in a further embodiment, a method according to the prior art, in order to bring about a relief of the congested port. This should be reported by a manageable switch as SNMPTrap, as it may indicate a basic network malfunction.

The method according to the invention affords the following advantages:

 No communication service already provided is rejected.

An increased communication load arises only between the switch with the overloaded output port and its predecessor.

The method can only work on individual network elements

(Switches) are implemented, it need not be known and propagated in the ^Vor¬ gänger network elements. Each network element can react by itself according to the prior art. This may be, for example, a foreign device that has no knowledge of the method. This advantage is be ^¬ Sonder's important because usually devices are integrated into existing networks. It is therefore not necessary to introduce the change on all network elements.

In addition, the switch does not need to know about the infrastructure of the network. There must be no analysis or engineering of the network, either before commissioning or during operation.

There is no need to implement a separate mechanism to prevent starvation / circling of packets - implied by the procedure itself.

^{BRIEF DESCRIPTION OF THE DRAWINGS} The invention is explained below on the basis of ^exemplary embodiments. Show

 1 shows an exemplary network according to the prior art, Figure 2 shows a first example according to the subject invention and

Figure 3 shows a second example according to the subject invention.

FIG. 2 outlines the problem by way of example for a network consisting of four terminals EG1, EG2 and EG3 (eg PCs, programmable logic controller PLCs, PLC etc.) and three switches SW1, SW2, SW3. If the terminals EG1, EG2 and EG3 now each send data packets PI, P2, P3 to terminal EG4, an overflow in the buffer BUF can occur in the switch SW1, in which all packets meet one another. Switch SW1 would then discard any further incoming data packets in the prior art, as long as the buffer BUF is busy. In the figure, only one data buffer BUF and one port are shown. However, as described above, there may be multiple data buffers.

In the described solution, the surplus data packet P3 is now not discarded, but sent back to the preceding in the transmitter "chain" network element SW2. The network element SW2 receives the data packet, but may not even know that it has already been sent once to the network element SW1, determines the next step on the way to the final destination EG4 and sends the data packet P3 again, retry.

In this example, there is only one possible way from transmitter EG3 to receiver EG4 via the network elements SW3, SW2, SW1. It is also conceivable that an alternative route exists and the network element after each further reception of the data packet P3 again decides which way is selected to the receiver, via SW2 or another switch (not shown in the figure).

3 shows a further embodiment for a second network also consisting of four terminals EG1-EG4 (for example PCs, Programmable Logic Controller PLCs, PLC etc.) and four switches SW1 to SW4.

If now send the terminals EG1, EG2 and EG3 each data packets PI, P2, P3 to terminal EG4, it comes in switch SW1, in which all packets meet, an overflow of the data buffer BUF.

First data packet P2 is received by EG2 and optionally buffered in Da ^¬ tenpuffer BUF.

It is now assumed that the data packet P3 via SW3 and SW2 arrives next in the network element SW1 as the next, and therefore somewhat earlier in time, the third data packet PI from the terminal EG1. Furthermore, it is assumed that SW1 can not cache the data packet and therefore according to the invention. according to the process back to the previous network element SW2 sends.

Now, if the next data packet PI is received, it provides an embodiment that the network element SW1 this data packet also sends back to the transmitter (in this case, the network ^¬ element SW4), regardless of whether the data buffer BUF is now free again and the data package could possibly even be forwarded. In this case, the retransmission of the data packet P3 is awaited, retry3, before the data packet PI is also transmitted.

Claims

Patent claims Method for transmitting a data packet (PI, P2, P3) in a communications network from a transmitter (EG1, EG2, EG3) to a receiver (EG4), wherein the communication network consists of individual, independent network elements (SW1, SW2, SW3, SW4) and at least some of the network elements have a data buffer (BUF) for buffering the data packets, and the data packet (PI, P2, P3) from the network element ( SW1, SW2, SW3) can be forwarded after receipt or stored in the data buffer, and at least one data packet (P3) cannot be buffered because there is no longer sufficient storage space in the data buffer (BUF), characterized in that the received data packet is sent back from the receiving network element (SW1) to the sending network element (SW2). Method according to patent claim 1, characterized in that After a first received data packet (P3) has been sent back from the ^receiving network element (SW1) to the sending network element (SW2), each further received ^data packet (PI) is sent back to the respective transmitter (SW4), as long as until the first received data packet could be ^successfully transmitted to the next network element (EG4). Method according to patent claim 1 or 2, characterized in that The transmission attempts of the data ^packet (P3) are monitored by monitoring how much time and/or how many delivery attempts are required to forward the first data packet (P3) to the next network element (EG4). Method according to patent claim 3, characterized in that The monitoring shows that the transmission of the data packet (P3) requires too much time and/or too many delivery ^attempts , whereby the data packet (P3) is deleted and an error message is generated. Method according to one of the previous patent claims, characterized in that the communication network works in accordance with the IEEE 802 Ethernet standard. Network element (SW1, SW2, SW3, SW4) for transmitting data packets (PI, P2, P3) in a communication network from a transmitter (EG1, EG2, EG3) to a receiver (EG4), with receiving means (EM), transmitting means (SM ) and with a data buffer (BUF), wherein a data packet (PI, P2, P3) received from a neighboring network element (SW2) by means of the receiving means cannot be stored in the data buffer (BUF) after receipt characterized in that the transmission means (SM) of the receiving network element (SW1) send the received data packet (P3) back to the sending network element (SW2, SW4). Network element according to patent claim 6, characterized in that the transmission means send every further received data packet (PI) back to the respective transmitter (SW4) after a first received data packet (P3) has been sent back from the receiving network element (SW1) to the sending network element (SW2), until the first received data ^packet could be successfully ^transmitted to the next network element (EG4). 8. Network element according to claim 6 or 7, characterized in that Monitoring means monitor the transmission attempts of the data ^packet (P3) by measuring how much time and/or how many delivery attempts are required to forward the first data packet (P3) to the next network element (EG4). 9. Network element according to claim 6 or 7, characterized in that the data packet (P3) is deleted (DEL) if the ^monitoring means determine that the transmission of the ^data packet (P3) requires too much time and/or too many delivery ^attempts and an error message is generated. 10. Network element according to one of claims 6 to 9, characterized in ^that it works according to the IEEE 802 Ethernet standard.