JP3545653B2 - Packet switch - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a packet switch for switching packets and a packet switching method.
[0002]
[Prior art]
In recent years, as the use of the Internet has spread, the growth of Internet traffic has been remarkable.
[0003]
This traffic is switched by a router, a layer 3 switch, or the like (hereinafter, a device that switches a packet, such as a router or a layer 3 switch, is generally referred to as a packet switch). For this reason, there is an increasing demand for packet switches to expand throughput. In response to this requirement, a packet switch having a multi-stage connection configuration capable of expanding the scale by connecting unit switches in multiple stages is known as one solution. For example, FIG. 9 shows a 16 × 16 packet switch configured in a three-stage cross network using a 4 × 4 unit switch 1001.
[0004]
Internet traffic is characterized by best effort, which does not reserve bandwidth on the route. Due to this best-effort property, a plurality of packets collide inside the packet switch. This phenomenon has been well known in the past, and a packet that loses due to contention in the collision occurs. The lost packet is discarded by the switch having no buffer. On the other hand, in a switch having a buffer, if there is still room for the buffer amount, such a packet is temporarily stored and then aimed at the output port of the destination. However, if there is not enough room, the packet is also discarded.
[0005]
Since such discarding is not desirable, the flow control is applied between the unit switches to suppress the input so that discarding does not occur inside the switch, or the retransmission control is performed until the discarded packet reaches the destination output port. Methods of working internally have been considered so far.
[0006]
However, the method using the flow control is not suitable for a large-scale switch configured by connecting the unit switches in multiple stages as illustrated in FIG. 9 due to the size of the control delay and the amount of control information. . Therefore, a large-scale switch often requires a retransmission control function. In this retransmission control, when a packet is discarded, the packet is retransmitted from the input port again. This retransmission guarantees that a packet arriving at the input port of the switch will always reach the destination output port.
[0007]
Normally, at the output port of a switch, packets belonging to the same flow must observe the order at the input port of this flow. However, due to the retransmission as described above, the order may arrive at the destination output port in reverse order. For example, a subsequent packet is sent to a destination output port without knowing that a previous packet of the same flow has been discarded, and the subsequent packet arrives first. In another example, a plurality of paths (multipath) can be set for an arbitrary pair of an input port and an output port inside a switch. In this case, if the routes do not have the same delay time, the order of the packets may be reversed. In these cases, a function of transmitting the order at the input port to the destination output port as a sequence number by embedding the sequence number in the header of the packet or the like, and performing a correct rearrangement is required. Such a sequence number can be used not only for retransmission but also for processing packet duplication that may occur at an output port due to erroneous delivery or the like.
[0008]
Under these circumstances, studies are being made on a method of reducing the occurrence probability of the collision as much as possible, reducing discarding and retransmission, and improving the transfer efficiency of the switch.
[0009]
For example, a unit switch in which a switch port has a plurality of physical communication paths is known. Since such a port can transfer packets by the number of the communication paths, the probability of collision can be reduced. Such an idea can be found, for example, in the document "The Performance of Multistage Interconnection Networks for Multiprocessors" (Clyde. P. Kruskal, IEEE Trans. Comp. 1983) and the document "Design of a Broadcast Packet Network" (Jonathan. S. Turner, IEEE INFOCOM'86, pp. 667-675, 1986).
[0010]
It is assumed that the unit switch 1003 shown in FIG. 10 physically has 32 links (communication paths) 1005 of 155 Mbps. As shown in FIG. 10, if four links are grouped into one port 1007, a switch having eight ports can be configured. At this time, up to four packets can be simultaneously transferred to any output port without collision. Also, by this grouping, a switch having four ports can be formed when eight links are combined, and a switch having two ports can be formed when sixteen links are combined. In these cases, up to eight or up to sixteen packets can be transferred simultaneously without collision.
[0011]
However, the method used by this unit switch has the following problems in terms of flexibility. That is, the number of communication paths that one port can have is limited at most to the number of links. That is, the number of communication paths cannot exceed the number of physically prepared links.
[0012]
[Problems to be solved by the invention]
As described above, the problem that the communication path cannot be flexibly set in the existing method of setting a plurality of physical communication paths in one port to reduce the probability of collision inside the switch. There is a point.
[0013]
The present invention has been made in consideration of the above circumstances, and enables a flexible setting of a communication path without being physically restricted, thereby further avoiding a collision inside a switch. It is an object of the present invention to provide a packet switch and a packet switching method which facilitate the switching and further improve the throughput of the switch.
[0014]
[Means for Solving the Problems]
The present invention (Claim 1) is a packet switch for exchanging packets between a plurality of incoming physical links and a plurality of outgoing physical links. A plurality of receiving means for receiving a packet in a state of data subdivided into predetermined units shorter than the packet length and receiving the subdivided data transferred in a cyclic order; and Switching means for transferring the received fragmented data to a destination for a packet to which the fragmented data belongs; and connecting the fragmented data transferred by the switching means to a connected output. And a plurality of transmitting means for transmitting to the physical link on the side.
Preferably, the number of packets transferred from one incoming physical link at the same opportunity is set to n, and the exchange means sets an upper limit of n packets to the same transmission means at the same opportunity. It may be transferred as
Preferably, when more than n packets are to be transferred to the same transmitting means at the same opportunity, the switching means may discard more than n packets. .
[0015]
According to the present invention, one physical port of a switch can be divided into a plurality of logical communication paths, and a plurality of packets can be transmitted and received in parallel using the plurality of logical communication paths. As a result, the probability of collision of packets inside the switch can be reduced, and the throughput of the switch can be improved.
[0016]
The present invention (claim 4) has a plurality of physical link groups that bundle a plurality of incoming physical links and a plurality of physical link groups that bundle a plurality of outgoing physical links. A packet switch for exchanging packets between a link group and a plurality of outgoing physical link groups, wherein a plurality of packets are subdivided into predetermined units shorter than the packet length from a connected incoming physical link. Receiving in the state of the fragmented data and where the fragmented data has been transferred in a cyclic order, a plurality of receiving means, and the fragmented data received in the receiving means, Switching means for transferring a packet to which the data belongs to one selected from a plurality of physical links belonging to an outgoing physical link group to be transferred; The data the subdivided which are transferred by means transmits to the connected outlet side of the physical link, characterized in that a plurality of transmission means.
Preferably, the number of packets transferred from one incoming physical link at the same opportunity is n, the number of physical links belonging to one physical link group is m, and the switching means is the same. At the opportunity, n × m packets may be transferred to the same transmitting unit with the upper limit.
Preferably, when more than n × m packets are to be transferred to the same transmitting means at the same opportunity, the switching means discards more than n × m packets. It may be.
[0017]
According to the present invention, a plurality of physical ports of a switch are bundled and treated as one thick port, and the thick port is divided into a plurality of logical communication paths, and the plurality of logical communication paths are divided. Can be used to transmit and receive a plurality of packets in parallel. As a result, while preventing or suppressing the extension of the packet transfer cycle, the collision probability of the packet inside the switch can be reduced, and the throughput of the switch can be improved.
[0018]
The present invention (claim 7) provides a switch network in which the packet switch according to claim 1 or 2 is connected in multiple stages as a unit switch, and input buffer means connected to a preceding stage of the switch network for each of the incoming physical links. Output buffer means connected to each output physical link at a subsequent stage of the switch network; and before transferring a packet, the input buffer means from which the packet is transferred becomes a transfer destination of the packet. Means for transmitting a request to the output buffer means via the switch network; and, when the request arrives at the output buffer means, a path through which the request has passed through the switch network for transferring the packet. Securing a response to the request arriving at the output buffer means serving as the transfer destination by the transfer source Characterized by comprising a means for communicating to become the input buffer means.
Preferably, a plurality of requests corresponding to a plurality of packets sent from each input buffer means may be sent in an order corresponding to the cyclic order in a state subdivided into the predetermined units. .
[0019]
According to the present invention, in a switch of a request / response control method, one physical port of a switch or one thick port obtained by bundling a plurality of physical ports is divided into a plurality of logical communication paths, A plurality of packets can be transmitted and received in parallel using the plurality of logical communication paths. As a result, it is possible to reduce the collision probability of requests inside the switch (this increases the number of packets that can be transmitted at the same time) and improve the throughput of the switch.
[0020]
The present invention (claim 9) provides a switch network in which the packet switch according to claim 1 or 2 is connected in multiple stages as a unit switch, and input buffer means connected to a preceding stage of the switch network for each of the incoming physical links. And a packet switching method in a packet switch comprising output buffer means connected to each of the outgoing physical links at a subsequent stage of the switch network, wherein the packet is a transfer source before the packet is transferred. A request is transmitted from the input buffer means to the output buffer means to which the packet is transferred via the switch network, and when the request reaches the output buffer means, the request passes through the switch network. Route for the transfer of the packet and arrives at the output buffer means as the transfer destination. A response to the request was characterized by transmitting to said input buffer means to be the transfer source.
[0021]
Note that the present invention relating to the apparatus is also realized as an invention relating to a method, and the present invention relating to a method is also realized as an invention relating to an apparatus.
[0022]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the invention will be described with reference to the drawings.
[0023]
FIG. 1 shows an example of a unit switch according to an embodiment of the present invention.
[0024]
The unit switch 1 in FIG. 1 is connected to a plurality of incoming physical links (communication paths) by a receiving unit 11 and connected to a plurality of outgoing physical links (communication paths) by a transmitting unit 12. One physical link is used as one port to switch between an input port and an output port. Although the number of input / output ports is arbitrary, FIG. 1 illustrates a case of 2 input ports × 2 output ports.
[0025]
Here, the basic principle of the present embodiment will be described with reference to FIG.
[0026]
The feature of the present embodiment is that a method for simultaneously transferring a plurality of packets at one physical port by using an interleaving method as shown in FIG. 1 is introduced into a packet switch. . That is, a plurality of fixed-length packets are interleaved in a predetermined unit and multiplexed to one port, and the packet switch switches data subdivided into a predetermined unit of each packet in one cycle. .
[0027]
Here, the unit for performing the interleaving, that is, the predetermined unit can basically take an arbitrary value such as every bit or every byte. As one extreme example, the present invention can be implemented with a unit of interleaving being a packet unit. In this case, the number of packet divisions is one, and the term interleaving is not appropriate, but it is a simple extension of the following description, and the effectiveness of the present invention can be applied as it is. Also, the number of packets multiplexed to one port can basically take any value.
[0028]
FIG. 1 shows an example in which four fixed-length packets are interleaved and multiplexed into one port. For the first input port, the four packets A, B, C, and D are divided into, for example, data for each bit, and in this order, A1, B1, C1, D1, A2, B2, C2, D2, …, And transfer them in a cyclic manner. Similarly, for the second input port, the four packets E, F, G, and H are divided into, for example, data for each bit, and in this order, E1, F1, G1, H1, E2, F2, and G2 , H2,..., And cyclically transfer.
[0029]
Here, as an example, if the destination of the packets A, C, F, and G is the first output port and the destination of the packets B, D, E, and H is the second output port, the unit switch 1 The data A1, C1, F1, and G1 subdivided into units such as are switched to a first output port, and the subdivided data B1, D1, E1, and H1 are switched to a second output port. Subsequent cycles are performed similarly.
[0030]
Then, the interleaved packets A, F, C, and G are output from the first output port, and the interleaved packets E, B, H, and D are output from the second output port. However, the order in which the fragmented data of each packet is output in each cycle in FIG. 1 is an example, and it is also possible to output in another order.
[0031]
A plurality of packets belonging to one flow may be interleaved and arrive at one input port of the switch at the same time. In this case, by determining the output order within the cycle of each output port of the switch so that the input order within the cycle at the input port of the switch is not reversed, the transfer order is reversed by the packets of the same flow passing through the switch. Can be prevented.
[0032]
With such a method, it is possible to perform switch processing by regarding one physical port as a plurality of virtual (logical) communication paths.
[0033]
Hereinafter, such a virtual communication channel is referred to as a “partition”. In other words, the method of the present embodiment divides one physical port into a plurality of partitions and distributes the packets to the respective partitions so that a plurality of packets can be transferred simultaneously in parallel. Things.
[0034]
Next, a case where a partition is provided will be described in comparison with a case where no partition is provided, with reference to FIGS.
[0035]
The pipe-shaped figure in FIG. 2 shows one port, the rectangle shows a packet, (a) shows a case where no partition is set, and (b) shows a case where a partition of this embodiment is set. Are respectively shown. The area of the rectangle in FIG. 2 indicates the data amount of the packet, the horizontal length of the rectangle indicates the time required for transferring the packet, and the area of the rectangle in FIG. 2A and the area of the rectangle in FIG. Similarly, it is assumed that the horizontal length of the rectangle in (b) is four times the horizontal length of the rectangle in (a).
[0036]
Here, it is assumed that four packets are to be simultaneously transferred to the port shown in FIG. If a partition is not set as shown in FIG. 2A, only one packet can be transferred at the same time (at this time, the remaining three packets may be discarded).
[0037]
On the other hand, when four partitions are set as shown in FIG. 2B, by distributing packets to these four partitions, four packets can be transferred simultaneously. The state in which four elongated rectangles are stacked in FIG. 2B indicates a state in which four packets are interleaved, that is, a state in which four partitions are set (this notation will be described later). The same applies to each figure).
[0038]
Thus, the ability to transfer simultaneously arriving packets up to “the number of partitions” times (in this example, four times) the case where partitions are set, compared to the case where partitions are not set. Yes, the probability of disposal can be reduced.
[0039]
FIGS. 3A and 3B show a switch (bufferless switch) having no buffer of 2 input ports × 2 output ports. FIG. 3A shows a case where no partition is set, and FIG. 3B shows a partition of this embodiment. Each case is shown. In the case of (b), the number of partitions is set to four.
[0040]
In the case of a switch without partitions as shown in FIG. 3A, packets destined for ports 0 and 1 are sent from the input ports 0 and 1 to the switch in packet units. Here, four packets are input from input port 0 in the order of destination output ports 0, 1, 0, and 1, and four packets are input from input port 1 in the order of destination output ports 1, 0, 0, and 1. Assume that a packet is input. At this time, in the third packet cycle, two packets bound for port 0 collide in the switch, and the lost packet is discarded. Similarly, in the fourth packet cycle, two packets destined for port 1 collide and the lost one is discarded.
[0041]
On the other hand, in the case of a switch having partitions as shown in FIG. 3B, four packets interleaved at the same time are transferred to the switch. In FIG. 3B, it is assumed that a hatched packet indicates a port 0 and an unhatched packet indicates a port 1. These packets are switched to their respective destination output ports. Since four partitions are set in the output port, a maximum of four packets can be transferred to the same output port at the same time (in an interleaved state). As in this example, when both the port 0 and the port 1 have four packets each, all packets can be switched at the same time without any discard.
[0042]
By the way, even if the partitions are set as described above, if the number of packets destined for the same output port exceeds the number set in the partitions, all the packets cannot be transferred simultaneously. For example, in the case of FIGS. 2B and 3B, since the number of partitions set is 4, when five or more packets are concentrated on one output port, all the packets are simultaneously transmitted. Cannot be transferred. That is, packets that are not selected at the time of simultaneous transfer (packets exceeding the number of partitions set; one packet in this example) are retained in the buffer in the case of a switch having a buffer (or However, if the switch does not have a buffer as shown in FIG. 3, it will be discarded. Therefore, in order to reduce the packet drop probability, the number of partition settings should be set to a large number.
[0043]
On the other hand, when the information transfer speed of the physical port is constant, if the number of partitions set is increased, the transfer time of one packet (packet transfer cycle) is extended by interleaving. For example, as shown in FIG. 1, when the original packet transfer cycle is T and the number of partitions is set to 4, the packet transfer cycle extends four times to 4T.
[0044]
Therefore, when the information transfer speed of the physical port is constant, there is a trade-off between the packet drop probability and the transfer time of one packet.
[0045]
Here, assuming that the packet length is 100 bytes, the speed of the port is changed and the packet transfer cycle is compared as follows.
When the port speed is 10 Mbps,
About 80μs when no partition is installed,
When the number of partitions is 4, it takes about 320 μs.
If the port speed is 622 Mbps,
When no partition is introduced, it takes about 1.3 μs.
When the number of partitions is 4, it is about 5.2 μs.
[0046]
Thus, as the port speed increases, the packet transfer cycle and the corresponding increase in the fixed delay time decrease. Therefore, in general, it can be said that the partition setting number can be set to be larger as the port speed increases.
[0047]
From the above, it is desirable that the number of partitions can be set flexibly in consideration of the drop probability and the packet transfer cycle. If the setting can be made only fixedly, the drop probability cannot be reduced, or the packet transfer cycle is too long, which adversely affects the quality of the application. In this regard, according to the method of the present embodiment, the number of partitions can be freely set by the interleaving method, so that it is possible to obtain a desired discard probability and a packet transfer cycle.
[0048]
As described above, the basic configuration of the present embodiment has been described using the 2 × 2 unit switch as an example, but various variations to which the above basic configuration is applied will be described below.
[0049]
First, FIG. 1 shows an example in which the interleaving method is applied to physical ports. However, as shown in FIG. 4, a plurality of ports are physically combined into one bundle and one thick port (super port and It is also possible to set a plurality of partitions by applying the interleaving method to the superport.
[0050]
In FIG. 4, the number of partitions is set to 8 in units of a superport in a 4 × 4 switch in which two ports are bundled (superported) to accommodate twice the normal speed. A case is shown (the number of partitions = 4 for each port). Port 0 and port 1 are collectively referred to as superport A, and port 2 and port 3 are collectively referred to as superport B. The superports A and B, which combine two ports, can interface at twice the speed of a normal port.
[0051]
In FIG. 4, it is assumed that a hatched packet indicates an egress superport A and an unhatched packet indicates an egress superport B. That is, five packets destined for the egress superport A and three packets destined for the egress superport B are input to the ingress superport A, and the egress superport B is input to the egress superport B. And three packets destined for the egress superport B are input.
[0052]
In this case, there are a total of eight packets destined for the egress superport A and a total of eight packets destined for the egress superport B, so that all 16 packets can be transferred.
[0053]
Note that a packet destined for the incoming super port A may be transferred to the output port 0 or the output port 1. Therefore, in the case of the example of FIG. 4, eight packets destined to the inbound superport A may be distributed to the output port 0 and the output port 1 by four on a predetermined basis. Similarly, a packet destined for the incoming super port B may be forwarded to the output port 2 or forwarded to the output port 3. Therefore, in the case of the example of FIG. 4, eight packets destined to the inbound superport B may be distributed to the output port 2 and the output port 3 by four on a predetermined basis.
[0054]
Of course, if packets destined for the same superport exceed the number of partitions set for the superport, all of those packets cannot be transferred simultaneously as described above.
[0055]
In this way, by introducing partitions in units of superports, it is possible to virtually increase the number of partitions to be set and to prevent or suppress the extension of the packet transfer cycle.
[0056]
Next, FIG. 5 shows an example of a unit switch adopting the interleaving method of 4 partitions as shown in FIG. 1 in each port of the 8 × 8 switch.
[0057]
As a result of the interleaving method, this unit switch 31 may be considered to have the same number (32) of ports as the unit switch shown in FIG. By using such unit switches adopting the interleaving method of the present embodiment and forming a multi-port switch with a multi-stage connection configuration such as a delta network, a cross network, or a banyan network, the number of wires between the unit switches can be reduced. There is an implementation advantage. Specifically, the number of wirings is one-sixth of the number of partitions (one-fourth in this example) as compared with the case where the unit switches as shown in FIG. 9 are used.
[0058]
Next, FIG. 6 shows an example of a packet switch in which unit switches employing the interleaving method of the present embodiment are connected in multiple stages.
[0059]
FIG. 6 shows an example of a 64 × 64 switch in which a switch network 45 is configured by a three-stage cross network connection using an 8 × 8 unit switch 41 adopting the interleaving method as in FIG. 1 and FIG. In FIG. 6, the description of the connection between the unit switches is partially omitted.) The number of partitions set in each unit switch 41 is the same, and this set number is determined in consideration of the balance between the discard probability and the packet transfer cycle as described above.
[0060]
In this packet switch, an input buffer 43 is provided for each input port, and an output buffer 47 is provided for each output port. The input / output buffers 43 and 47 perform protocol conversion and protocol processing. For example, when a variable-length packet of the IP protocol is input, the input buffer 43 segments the variable-length packet of the IP protocol into fixed-length packets unique to the switch, and interleaves the plurality of fixed-length packets. The output buffer 47 performs the reverse assembly and outputs the packet as a variable length packet of the IP protocol. The input / output buffers 43 and 47 may also have a request / response control function, as described later.
[0061]
Although an example in which control is performed on a packet has been described above, an example in which an interleave method is applied to a switch using a request-response control method will be described below.
[0062]
FIG. 7 schematically shows a case where a request / response control method is introduced into a switch having a multi-stage connection configuration as shown in FIG.
[0063]
In FIG. 7, 53 indicates an input buffer, 55 indicates a switch network, and 57 indicates an output buffer. Note that unit switches inside the switch network 55 are omitted.
[0064]
In the request-response control method, the following basic procedure is performed before transferring a packet.
The source input buffer sends a "request" corresponding to the packet to the destination output buffer using the switching network.
-The output buffer receiving the "request" returns a "response" to the "request" to the input buffer of the transmission source.
-The input buffer receiving this "response" transfers the corresponding packet.
[0065]
In the case of the present embodiment, the details are as follows.
[0066]
In the switch of the present embodiment to which the interleave method is applied, partitions can be set not only for packets but also for requests and responses. That is, in the present embodiment, interleaving is also applied to the request, and the request is transferred by the same partition as the partition to which the corresponding packet is to be transferred.
[0067]
At this time, the effect of the interleaving method reduces the probability of collision between requests in the switch network as compared with the conventional case, and as a result, the probability of the request reaching the output buffer becomes higher than before.
[0068]
When the request arrives from the input buffer to the output buffer, a route actually passed by the request is reserved in the switch network for a packet to be transferred by the input port and the partition. In the present embodiment, it is assumed that a set of an input port and a partition is held at the entrance and the exit of each unit switch as a route through which the request has passed.
[0069]
The response is returned to the input port of the transmission source using a switch network (for example, following the route through which the request has been made). For example, if one set of input port and output port is implemented as one module, the input port that sent the request via the switch network from the input port implemented in the same module as the output port that received the request A response may be returned via a route to an output port implemented in the same module as. It is assumed that this response route is secured at the same time when the request reaches the destination output port. The response may be returned using a separate line without using the switch network.
[0070]
When the input buffer receives the response, the corresponding packet is transferred according to the path configured by the pair of the port and the partition (passed by the request) held in the switch network. As a result, the packet safely reaches the output buffer without colliding with another packet on the way.
[0071]
Note that the request-response control method is applied to a multi-stage switch composed of bufferless unit switches, and the time required for the basic procedure can be easily predicted because the unit switch which is a component does not have a buffer.
[0072]
Before the request reaches the destination output port, the request may lose contention due to a collision that occurred on the way, and may be discarded in the switch network. However, even if it is discarded, the time required for the basic procedure of the request / response can be easily detected because it can be predicted as described above. When the discard is detected, the request is retransmitted to the destination output port at the next opportunity (retransmission is repeated until a response is obtained).
[0073]
As described above, by sequentially performing the procedure of request / response, the packet is transmitted after the path of the packet is secured. Therefore, if the response is received, it is guaranteed that the response is always delivered to the destination. In other words, in the request-response control method, a route is set for each packet.
[0074]
Further, the number of requests that can be transmitted and the number of packets that can be transmitted are accurately managed in the input buffer for each flow, and packets enqueued in the input buffer are controlled so as to be surely delivered to the output buffer of the destination. The number of requests that can be sent is increased by one when a packet is enqueued in the input buffer and when discard of the request is detected. Conversely, when a request is sent out, it is decremented by one. On the other hand, the number of packets that can be transmitted is increased by one when a response is received, and reduced by one when a packet is transmitted. In this way, the number of packets enqueued in the input buffer of the transmission source and the number of packets delivered to the output port of the destination can always be equal.
[0075]
In addition, since only the number is managed, such as the number of packets that can be transmitted, and the packet is transmitted after receiving the response, it is possible to transfer the packet to the destination output port without fail in order. This eliminates the need for packet rearrangement control at the output port.
[0076]
Here, adopting the interleaving method of the present invention in a switch of the request / response control method has the following advantages.
In other words, when a plurality of partitions are set, the same number of requests can be sent at the same time, so that the probability that requests collide inside the switch network and be discarded decreases. As the request collision probability decreases, the number of packets that can be transmitted simultaneously increases. These packets are never discarded because the route is secured. When the number of packets that can be transmitted simultaneously in the input buffer increases, the memory area occupied by these packets increases because they can be released together with the transmission.
As described above, when the present invention that allows a port to appear as a plurality of communication paths (partitions) is combined with a request / response control type switch, the efficiency of the switch can be further increased.
[0077]
Next, in the above-described basic procedure, information indicating a priority is added to the request, and when the requests collide, control is performed to determine which packet should be discarded based on the priority given to the request. More could be done.
[0078]
As a procedure for determining the priority, for example, as shown in FIG. 7, the congestion information of the output port (buffer) is added to the response, and the response is returned to the input port. The priority given to the request corresponding to the packet is obtained based on the congestion information.
[0079]
As the congestion information, for example, a predetermined measurement value related to the congestion situation, specifically, the number or number of packets arriving in a unit time, the queue length of the packet queue (the number of packets or the number of bytes), discarded from the packet queue For example, the number of packets or the number of bytes), a value indicating a congestion level, or bit data indicating whether or not a congestion state exists.
[0080]
Regarding the setting of the priority, as a basic idea, for example, the priority is set lower as the congestion information indicates a higher congestion degree, and the priority is set higher as the congestion information indicates a lower congestion degree.
[0081]
As for the selection of requests in the unit switch, the basic idea is that, for example, requests having the highest priority are selected in order from among conflicting requests.
[0082]
Now, when the request arrives at the destination output port as described above, this output port determines whether or not it is in a congestion state according to the buffer status. As shown in FIG. 7, the congestion information determined based on this determination is included in the response. This response is returned to the input port of the transmission source using, for example, a switch network as described above.
[0083]
Upon receiving the response, the input port of the transmission source transmits the packet as described above. The packet itself always arrives at the destination output port, for example, via the route secured by the previous request. On the other hand, the input port of the transmission source extracts the congestion information of the output port from the response, and determines the priority of the next packet destined for this output port based on this information. The next packet attempts to secure a route by the request given this priority. If the reservation is successful, the request is delivered to the destination output port by following the same route as the request.
[0084]
If such a control method is used, for example, the characteristics of packets destined for other output ports, especially the delay characteristics are not deteriorated by the influence of the output port that is in a congested state, so that traffic having a high best-effort property is accommodated. It is considered to be particularly effective in the case of a packet switch.
[0085]
Here, some examples of the intra-switch packet transfer control based on the priority will be described.
[0086]
First, the most typical control, that is, the priority control performed according to the congestion state will be described.
For example, a request for a packet destined for a congested destination (congested port) has a low priority. Since a high-priority request is prioritized even if it collides inside the switch network, it is the same as a high-priority request having no low-priority request to a congested port. In other words, it is the same as there is no low-priority packet destined for a congested port for a high-priority packet.
[0087]
This has the effect that the flow of packets destined for non-congested ports is significantly less disturbed by the flow of packets destined for congested ports. Even for packets destined for a congested port, since the priority is only changed and the transfer speed is not suppressed, it is necessary to continue trying to transfer to the output port using the gap of the transfer of high priority packets Can be. Therefore, packets can be continuously transferred to a congested port as long as the transfer capability of the switch network is maintained.
[0088]
Next, priority control for retransmission will be described.
[0089]
In a packet switch in which the input buffer retransmits a request discarded due to a collision inside the switch network, if the same request is discarded continuously, the entire packet waiting to be transferred after the packet corresponding to the request Flow may be worsened.
[0090]
As a method of solving this, when a request is discarded inside the switch network, the priority of the retransmission request may be made higher than the priority of the original request as shown in FIG. . Such control has an advantage that retransmission requests are less likely to be discarded again. If the retransmission request is discarded again, the priority of the retransmission request for the retransmission request may be made higher. As a result, it is possible to preferentially transfer unfortunate requests that have been retransmitted many times.
[0091]
By the way, in the method of the present embodiment using the interleaving method, it is possible to use a feature that various control signals given to a plurality of interleaved packets exist at substantially the same position in time. is there.
[0092]
For example, a signal (information) for achieving packet synchronization can be shared between all partitions. By doing so, it is possible to reduce the signal for packet synchronization to 1 / “the number of partitions” as compared to the case where a signal is prepared for each partition.
[0093]
Of course, sharing of certain information among all partitions as described above can be applied to a code (for example, a parity code) for detecting a transmission path error. That is, instead of attaching a parity code to each packet, a parity code can be assigned in units of a plurality of interleaved packets. By doing so, it is possible to reduce the overhead due to the parity code.
[0094]
FIG. 8 shows an example of a format exchanged between unit switches when the number of partitions set in the unit switch is 4 and this request / response control method is used. In the example of FIG. 8, the synchronization signal, the request, the response, the packet body, and the parity code can be transferred simultaneously. FIG. 8 shows that, at first glance, the synchronization signal and the parity code appear to be prepared for each packet (partition). However, as described above, since they are located at the same phase, they are connected together. By sharing the information among the four partitions, the amount of information to be used can be reduced (of course, a synchronization signal and a parity code can be prepared for each packet (partition)).
[0095]
The format shown in FIG. 8 is an example that can be adopted in the embodiments as shown in FIGS. 3B and 4, and is a format in which the position of a synchronization signal or a parity code is replaced, and other formats. Even in a format in which information is added or a format in which a part is replaced with other information, it is of course possible to share the data among all partitions.
[0096]
The present invention is not limited to the above-described embodiment, and can be implemented with various modifications within the technical scope thereof.
[0097]
【The invention's effect】
According to the present invention, since a plurality of packets are interleaved for each physical link and subjected to switch processing, collisions inside the switch can be more flexibly avoided, and the throughput of the switch can be improved.
[Brief description of the drawings]
FIG. 1 is a diagram showing a configuration example of a packet switch according to an embodiment of the present invention.
FIG. 2 is a diagram for explaining a case where there is a partition as compared with a case where there is no partition;
FIG. 3 is a diagram for explaining a packet switch having partitions in comparison with a packet switch having no partitions;
FIG. 4 is a diagram showing an example of a packet switch in which a partition is set in a super port.
FIG. 5 is a diagram showing an example of a unit switch in which a partition can be set;
FIG. 6 is a diagram showing an example of a multi-stage connection switch using unit switches for which partitions can be set;
FIG. 7 is a diagram showing an example of a multi-stage connection switch using unit switches that can be set to partitions to which a request-response control method is applied.
FIG. 8 is a diagram showing an example of an internal data format of a packet switch to which a partition setting and a request / response control method are applied;
FIG. 9 illustrates an example of a conventional switch.
FIG. 10 is a diagram for explaining another example of a conventional switch.
[Explanation of symbols]
1, 21, 31, 41 ... unit switch
11 receiving unit
12 ... Transmission unit
43, 53 ... input buffer
45, 55 ... Switch network
47, 57 ... output buffer

Claims (4)

A packet switch for exchanging packets between a plurality of incoming physical links and a plurality of outgoing physical links,
Receives a state in which a plurality of packets are fragmented into predetermined units shorter than the packet length from the connected incoming physical link, and where the fragmented data is transferred in a cyclic order. A plurality of receiving means;
Exchange means for transferring the fragmented data received by the receiving means to a transfer destination for a packet to which the fragmented data belongs;
A plurality of transmission means for transmitting the fragmented data transferred by the exchange means to the connected physical link on the outgoing side,
When the number of packets transferred from one physical link on the ingress side at the same opportunity is set to n, when transferring up to n packets to the same transmitting means at the same opportunity as the upper limit, a buffer for temporarily storing more than n packets. A packet switch for exchanging packets between a plurality of incoming physical links and a plurality of outgoing physical links,
Receives a state in which a plurality of packets are fragmented into predetermined units shorter than the packet length from the connected incoming physical link, and where the fragmented data is transferred in a cyclic order. A plurality of receiving means;
Exchange means for transferring the fragmented data received by the receiving means to a transfer destination for a packet to which the fragmented data belongs;
A plurality of transmitting means, which transmits the fragmented data transferred by the switching means to a connected physical link on the outgoing side,
Assuming that the number of packets transferred from one incoming physical link at the same opportunity is n, the switching means sends the same transmission means at the same opportunity to the same transmission means with n packets as an upper limit. A packet switch for discarding a packet exceeding n in forwarding. It has multiple physical link groups that bundle multiple physical links on the ingress side and multiple physical link groups that bundle multiple physical links on the outgoing side. A packet switch for exchanging packets with a physical link group,
Receives a state in which a plurality of packets are fragmented into predetermined units shorter than the packet length from the connected incoming physical link, and where the fragmented data is transferred in a cyclic order. A plurality of receiving means;
A switching unit for transferring the fragmented data received by the receiving unit to one selected from a plurality of physical links belonging to an outgoing physical link group to which a packet to which the data belongs is to be transferred; When,
A plurality of transmission means for transmitting the fragmented data transferred by the exchange means to the connected physical link on the outgoing side,
If the number of packets transferred from one incoming physical link at the same opportunity is n, and the number of physical links belonging to one physical link group is m, the same transmission means is used at the same opportunity. And a buffer for temporarily storing packets exceeding the number of n when transferring data with an upper limit of n × m packets. It has multiple physical link groups that bundle multiple physical links on the ingress side and multiple physical link groups that bundle multiple physical links on the outgoing side. A packet switch for exchanging packets with a physical link group,
Receives a state in which a plurality of packets are fragmented into predetermined units shorter than the packet length from the connected incoming physical link, and where the fragmented data is transferred in a cyclic order. A plurality of receiving means;
A switching unit for transferring the fragmented data received by the receiving unit to one selected from a plurality of physical links belonging to an outgoing physical link group to which a packet to which the data belongs is to be transferred; When,
A plurality of transmitting means, which transmits the fragmented data transferred by the switching means to a connected physical link on the outgoing side,
If the number of packets transferred from one incoming physical link at the same opportunity is n, and the number of physical links belonging to one physical link group is m, the switching unit will be able to use the same opportunity. A packet switch for discarding more than n × m packets when transferring up to n × m packets to the same transmission means.

Images