CN101567855B - Distributed packet switching system and distributed packet switching method - Google Patents
Distributed packet switching system and distributed packet switching method Download PDFInfo
- Publication number
- CN101567855B CN101567855B CN2009101476724A CN200910147672A CN101567855B CN 101567855 B CN101567855 B CN 101567855B CN 2009101476724 A CN2009101476724 A CN 2009101476724A CN 200910147672 A CN200910147672 A CN 200910147672A CN 101567855 B CN101567855 B CN 101567855B
- Authority
- CN
- China
- Prior art keywords
- packet
- message
- module
- cable card
- card board
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention discloses a distributed packet switching system and a distributed packet switching method. A source line fastener plate buffers the uplink data packet according to the target cable fastener plate and slices according to length. The source line fastener plate polls each buffer memory queen and transmits the message slice to the corresponding target cable fastener plate by the packet switching chip. The bandwidth is dynamically distributed by the target cable fastener plate. The target cable fastener plate rearranges the received message slice. The invention is simple, can effectively improve the QoS ability of the distributed packet switching system, reduce the data flow chatter and enhance the homogeneity of the system flow.
Description
Technical field
The present invention relates to data communication field, relate in particular to a kind of distributed packet switching system and a kind of distributed package switching method.
Background technology
In order to adapt to the bandwidth demand that network increases sharply, the essence variation has taken place in the structure of intermediate layer route exchange device, progressively carry out the transition to distributed structure/architecture by traditional based on bus and central processing unit framework, improve data signaling rate thereby can pass through switching fabric (switch fabric).
The switching technology of main flow comprises cell switching and packet switch at present; Wherein, cell switching is owing to the chip price costliness and be positioned the high-end basic network equipment of carrier-class, therefore mainly is applied to ATM (Asychronous Transfer Mode, asynchronous transfer mode) network and high-end IP network equipment; Packet technology then obtains extensive use in the basic network equipment based on ethernet technology.
Fig. 1 shows the schematic diagram of a distributed packet switching system, this distributed packet switching system comprises a plurality of cable card boards and a packet switching chip (also claiming the Ethernet packet switching chip), each input port of packet switching chip receives a packet at most simultaneously, each output interface sends a packet at most simultaneously, is that unit carries out dispatching message based on the message length statistics, with the packet;
Above-mentioned basic principle of dispatching based on the message length statistics is: receiver side receives the various state informations (for example quene state, bag spatial cache etc.) of transmitter side, distributes bandwidth for each transmitter side dynamically, and transmitter side sends message according to this bandwidth; By above-mentioned distributed Flow Control based on message length, guaranteed the equity dispatching between each cable card board, avoid congested and packet loss takes place in the Ethernet packet switching chip, taken into account the bandwidth availability ratio of Ethernet packet switching chip simultaneously;
But, above-mentioned distributed packet technology is that unit is dispatched with the packet, because length of data package has very big randomness, there is the big defective of scheduling granularity in therefore distributed packet switching system, this will cause the shake of data flow to increase, and reduce the uniformity of flow system flow;
In addition, also there is " the isometric bag of short bag " problem in the above-mentioned distributed packet technology that with the packet is unit is dispatched; For example, the data packet length that cable card board A sends to cable card board B is 1514 bytes, the data packet length that sends to cable card board C is 64 bytes, because the input port that packet switching chip is connected with cable card board A can only receive a packet simultaneously, therefore send in the process of 1514 byte data bags the 64 byte data bags that need the send to cable card board C wait state of just having to be in to cable card board B at cable card board A.
Summary of the invention
Embodiments of the invention aim to provide can effectively be reduced the data flow shake, improves the flow system flow uniformity and overcome the distributed packet-switched scheme that packet is waited for, thereby improve QoS (Quality-of-Service, the service quality) ability of network equipment system.
For achieving the above object, embodiments of the invention provide a kind of distributed packet switching system, comprise packet switching chip and a plurality of cable card board; Up exchange adaptation unit and descending exchange adaptation unit that each cable card board is respectively equipped with a bag processing unit and connects described bag processing unit;
The up exchange adaptation unit of source cable card board receives the upstream data bag from the bag processing unit of this cable card board, according to corresponding purpose cable card board buffer memory and cut into slices in addition respectively by certain-length, and, each buffer queue of poll is sent to the purpose cable card board with the packet slice that satisfies bandwidth requirement by described packet switching chip;
The descending exchange adaptation unit of purpose cable card board is the up exchange adaptation unit dynamic bandwidth allocation of source cable card board, packet slice that reorganization receives from packet switching chip and the bag processing unit that sends to this cable card board.
For achieving the above object, embodiments of the invention also provide a kind of distributed package switching method, are applicable to the distributed packet switching system that comprises packet switching chip and a plurality of cable card boards; May further comprise the steps:
Step S1: the source cable card board with the upstream data bag according to purpose cable card board buffer memory and cut into slices in addition respectively by certain-length;
Step S2: source each buffer queue of cable card board poll is sent to corresponding target cable fastener plate with the packet slice that satisfies bandwidth requirement by described packet switching chip; Wherein, described bandwidth is by described purpose cable card board dynamic assignment;
Step S3: the purpose cable card board is recombinated to the packet slice that receives.
As shown from the above technical solution, embodiments of the invention adopt the parcel Flow Control to replace big bag Flow Control by the section and the reorganization of packet, have following beneficial effect:
1, can effectively reduce the shake of data flow, improve the uniformity of flow system flow;
2, avoided " the isometric bag of short bag " problem;
3, simple, effectively improve the QoS ability of distributed packet switching system.
With reference to the accompanying drawing description of a preferred embodiment, above-mentioned and other purpose of the present invention, feature and advantage will be more obvious by following.
Description of drawings
Fig. 1 is the schematic diagram of distributed packet switching system in the prior art;
Fig. 2 is the structured flowchart of distributed packet switching system provided by the present invention;
Fig. 3 is the working state schematic representation between source cable card board and the purpose cable card board in the distributed packet switching system provided by the present invention;
Fig. 4 is in the distributed packet switching system provided by the present invention, the block diagram of up exchange adaptation unit one embodiment;
Fig. 4 a is the corresponding relation schematic diagram of the first packet buffer module and message length memory module in the up exchange adaptation unit;
Fig. 4 b is the operation principle schematic diagram of packet slice module one embodiment in the up exchange adaptation unit;
Fig. 5 is in the distributed packet switching system provided by the present invention, the block diagram of descending exchange adaptation unit one embodiment;
Fig. 6 is the flow chart of distributed package switching method one embodiment provided by the invention;
Fig. 7 a is the cut into slices flow chart of an embodiment of packet;
Fig. 7 b is the cut into slices flow chart of another embodiment of packet.
Embodiment
To describe specific embodiments of the invention in detail below.Should be noted that the embodiments described herein only is used to illustrate, be not limited to the present invention.
Main design of the present invention is to adopt elongated parcel to replace elongated big bag to carry out distributed packet switch, solves the shake, the uniformity and the packet that exist in the available data packet switch and waits for problem.To introduce distributed packet switching system provided by the present invention and distributed package switching method in detail below.
Below distributed packet switching system is described.
Please in conjunction with Fig. 2, distributed packet switching system 200 comprises packet switching chip and a plurality of cable card board;
Packet switching chip is also referred to as the Ethernet packet switching chip, and its each input port receives a packet (i.e. message is generally the IP message of ethernet frame format) at most simultaneously, and each output interface sends a packet at most simultaneously; Generally, data packet messages length based on industry ethernet is 64～1514 bytes, promptly the data packet messages length that exchanges by packet switching chip is 64～1514 bytes, has caused the even and packet wait problem of data flow shake, mass flow discrepancy that exists in the prior art greatly owing to granularity;
The present invention is based on existing packet switching chip, overcome the problems referred to above by improvement to the cable card board structure, concrete, up exchange adaptation unit 220 and descending exchange adaptation unit 230 that each cable card board is respectively equipped with a bag processing unit 210 and connects bag processing unit 210; Wherein,
This bag processing unit 210 is continued to use prior art for the processing of upstream data bag and downlink data packet, comprise: for the upstream data bag carry out message reception, table look-up, routing, Business Processing, in message, add operation such as transmission information and encapsulation between plate, then message is sent to up exchange adaptation unit 220; And, receiving the downlink data packet that descending exchange adaptation unit 230 sends, transmission information between peel plate is carried out the Business Processing and the message encapsulation of down direction, and is finished the transmission of message to outgoing interface; Do not repeat them here;
For up exchange adaptation unit 220 and descending exchange adaptation unit 230, though each cable card board all has said structure, but what relate in real work is that the forwarding of packet between two cable card boards handled, promptly for a packet, it is to send to a purpose cable card board from a source cable card board by packet switching chip; For convenience of explanation, please in conjunction with Fig. 3, shown that packet in the distributed switching system mails to the working state schematic representation of another purpose cable card board from a source cable card board; Those skilled in the art can understand, in actual use, the source cable card board can send different packets to a plurality of purpose cable card boards, and the purpose cable card board also can receive the different pieces of information bag that the multiple source cable card board sends, but specific to the forwarding of each packet, its operation principle all as shown in Figure 3:
The up exchange adaptation unit 320 of source cable card board receives the upstream data bag from the bag processing unit 310 of this cable card board, according to corresponding purpose cable card board buffer memory and cut into slices in addition respectively by certain-length, and, each buffer queue of poll is sent to the purpose cable card board with the packet slice that satisfies bandwidth requirement by packet switching chip;
The descending exchange adaptation unit 330 of purpose cable card board is up exchange adaptation unit 320 dynamic bandwidth allocations of source cable card board, packet slice that reorganization receives from packet switching chip and the bag processing unit 310 that sends to this cable card board;
As can be seen, up exchange adaptation unit 320 and descending exchange adaptation unit 330 cooperate the Flow Control scheduling of having finished packet slice, its basic principle is: descending exchange adaptation unit 330 receives the various state informations (as quene state, bag spatial cache etc.) of the up exchange adaptation unit 320 of other cable card boards, dynamic is that each up exchange adaptation unit 320 distributes bandwidth, and each up exchange adaptation unit 320 sends message according to this bandwidth; By the Flow Control scheduling of packet slice, can guarantee that packet switching chip does not exist congested and the packet loss phenomenon, and guarantee the equity dispatching between each cable card board, take into account the bandwidth availability ratio of packet switching chip simultaneously; Because being transmitting terminal, receiving terminal distribute the technology of bandwidth in Chinese patent application CN200710176023.8, CN200710179965.1 and CN200710193772.1, to carry out detailed discussion, so do not repeat them here;
Simultaneously, up exchange adaptation unit 320 is finished buffer memory, the sectioning of message, and descending exchange adaptation unit 330 is finished the reorganization operation of packet slice, i.e. section starts from up exchange adaptation unit 320, the descending finally exchange adaptation unit 330 of recombinating has the following advantages:
(1) owing to adopt the parcel exchange of certain-length to replace the whole packet switch of original 64～1514 byte data bags, the exchange granularity reduces, thereby can effectively reduce the uniformity that flow system flow is shaken and improved to data flow; Simultaneously, be necessary to point out, determined by the packet switching chip self character, it can support elongated packet, and this just means that also it supports elongated packet slice equally, therefore, when the not enough set certain-length of the last message residue length of section, need not to carry out byte and fill, therefore can not bring extra overhead;
(2) adopt poll to send, promptly working as a source cable card board has a big bag will send to purpose cable card board A simultaneously, when a parcel will send to purpose cable card board B, because two packets are all by slicing treatment, therefore under the situation that bandwidth all meets the demands, the packet slice that is two packets is alternately sending, and does not have the phenomenon of the isometric bag of short bag;
(3) bag section and framing are finished by up exchange adaptation unit and descending exchange adaptation unit, can not increase bag processing unit extra work; It may be noted that, in the distributed packet switching system of the present invention, up exchange adaptation unit on the cable card board and descending exchange adaptation unit can adopt independently software realization, also can adopt independently hardware mode realization, perhaps adopt the mode of independently software and hardware combination to realize, to guarantee can not reduce system's bag transfer capability owing to bringing bag processing unit extra work; Wherein independently hardware realizes it being best execution mode, such as adopting field programmable gate array to realize.
Though above the distributed packet switching system of the present invention is introduced on the whole, but in actual applications, those skilled in the art can realize the function of above-mentioned up exchange adaptation unit and descending exchange adaptation unit in different ways, respectively its preferred embodiment is described below, to obtain preferable technique effect.
Fig. 4 shows up exchange adaptation unit 420 1 embodiment, comprise uplink message receiving processing module 421, message length statistical module 422, uplink message sending module 426 and Flow Control module 427, also comprise and be respectively message length memory module 423, the first packet buffer module 424 and the packet slice module 425 that each other cable card board (1～n does not comprise the cable card board at up exchange adaptation unit 420 places) is provided with; Below each part of up exchange adaptation unit 420 is described respectively:
The upstream data bag that uplink message receiving processing module 421 receives from the bag processing unit according to the purpose cable card board information of carrying in the upstream data bag, sends it to the corresponding first packet buffer module 424; Wherein, uplink message receiving processing module 421 needs the corresponding relation of default first packet buffer module 424 and purpose cable card board, this corresponding relation can be set directly in the uplink message receiving processing module 421, also can be as shown in Figure 4, be arranged on independently in the buffer memory correspondence table; The buffer memory correspondence table is an example to be arranged on independently, uplink message receiving processing module 421 is after receiving the upstream data bag, can store this packet into certain first packet buffer module 424 according to purpose cable card board information that encapsulates in the packet and the set information of buffer memory correspondence table;
Message length statistical module 422 is connected with described uplink message receiving processing module 421, adds up the message length information of described upstream data bag, and stores corresponding message length memory cell 423 into;
The message length memory module 423 and the first packet buffer module 424 are one to one, the same message with 424 buffer memorys of the corresponding first packet buffer module of its each message length information of storing is one to one, shown in Fig. 4 a, the first packet buffer module 424 of corresponding n cable card board stores m IP message, then the just corresponding length information of preserving this m IP message of the message length memory module 423 of corresponding n cable card board; Certainly, the packet number that can store of 424 of the first packet buffer modules is relevant with the buffer memory capacity specification of cable card board;
Though the present invention does not limit the mode of storage in the message length memory module 423 and the first packet buffer module 424, can determine that different implementations may realize on the effect difference to some extent voluntarily by those skilled in the art; Such as, if continuous carry out packet storage and length information storage successively, just must additional different message fields divide to indicate, thereby assurance to the correct addressing with message obtained of message length information, this will bring some extra processing expenditure to system;
Therefore, as preferred embodiment, message length memory module 423 can be the preservation that unit carries out message length information with four bytes; With situation shown in Fig. 4 a is example, and preceding four bytes are used to preserve the length information of IP message 1 in the message length memory module 423 of n cable card board, and the 5th～8 byte is used to preserve the length information of IP message 2, and the like, until IP message m;
Similarly, the first packet buffer module 424 can be divided into the cache sub-module of fixed byte, and each cache sub-module is used for data packet voice of buffer memory; Because data packet length is generally 64～1514 bytes, therefore the cache sub-module of 2048 bytes can be set, and it can satisfy the packet storage demand fully, certainly, those skilled in the art also can adopt the cache sub-module of other specifications, as long as can satisfy the packet memory requirement; With situation shown in Fig. 4 a is example, and in the first packet buffer module 424 of n cable card board, preceding 2048 bytes are used to preserve IP message 1, the 2049～4096 byte and are used to preserve IP message 2, and the like, until IP message m;
Message length memory module 423 and/or the first packet buffer module 424 are carried out byte divide by above-mentioned, length information that it is preserved and packet can natural correspondences, and the identification that need not to carry out special sign can realize the message addressing.
Continuation is described other parts in the up exchange adaptation unit 420:
The packet slice module 425 of corresponding cable card board 1～n is connected with the first packet buffer module 424 and the message length memory module 423 of homologous lines clamp respectively, and the message length information of extraction upstream data bag is also cut into slices according to certain-length;
The present invention is not limited the specific implementation of packet slice module 425; Preferable, shown in Fig. 4 b, packet slice module 425 can comprise:
Packet pointer 4251 is pointed to current upstream data packet address;
Message length pointer 4252 points to the address of storing current data packet message length information in the message length memory cell;
Packet slice pointer 4253 points to current packet slice address;
Packet slice transmitting-receiving submodule 4254, be used for sending to uplink message sending module 426 according to the message data of above-mentioned pointer extracting certain-length, and with the address behind the described certain-length of described current packet slice address offset as new current packet slice address, carry out described packet slice pointer 4253 redirects; If current message is sent completely, carry out the redirect of packet pointer 4251 and message length pointer 4252 simultaneously, in general, be according to the order of packet in the first packet buffer module 424, order jumps to next data packet messages;
Further, packet slice module 425 can also comprise message priority recognin module (not shown), it is connected with the first packet buffer module 424 and packet slice transmitting-receiving submodule 4254, be used for discerning the priority that is buffered in the first packet buffer module, 424 message data bags, after the current data packet voice is sent completely, makes packet pointer 4251 and message length pointer 4252 point to packet and cut into slices with higher priority; If have the packet of two or more higher priorities, then generally carry out the packet section by the buffer memory order;
As preferred embodiment, if message length memory module 423 is that unit carries out the preservation of message length information and the first packet buffer module 424 when being split into the cache sub-module of 2048 bytes with four bytes, the operation principle of the packet slice module 425 of corresponding n cable card board is shown in Fig. 4 b:
Those skilled in the art can understand, and above-mentioned preferred embodiment is provided with in packet slice module 425 under the situation of message priority recognin module, and packet pointer 4251 and message length pointer 4252 might not be the order redirects; Still cutting apart message length memory module 423 and cut apart the first packet buffer module 424 with 2048 bytes with 4 bytes is example, the address of next IP message is current message address+x*2048, the address of next message length buffer memory is current address+x*4, x by the number of data packets at interval; As can be seen, no matter whether carry out packet slice, divide by message length memory module 423 and/or the first packet buffer module 424 being carried out byte according to priority, can both the addressing of more easy realization message;
Further, for convenience of descending exchange adaptation unit framing, to being provided with framing information labels submodule 4255 in the packet slice module 425, be used for adding the framing information labels at each packet slice, at least identifying the position of packet slice in data packet messages, promptly is first section, middle slice or last section of packet; In addition, can also sort, such as 1, the second packet slice sign 2 of first packet slice sign for each section ... and the like, and/or, can also provide other information that helps framing, such as check digit etc.;
It is important to point out that the framing information labels can be used as independently label to be added message to and transmit between sign and the section message, also can be used as the part that message is transmitted sign, promptly utilize message to transmit not have in the sign field realization of using; In common Ethernet packet switching chip is used, message is transmitted sign can't take effective band width in physical, 10Gbps Ethernet switching chip as Broadcom Corp is carried out message switching by the HiGig header, the HiGig header does not take effective band width in physical, when therefore the framing information labels is transmitted identify a part of as message, can not bring any extra overhead; When framing information labels during as separate label, the overhead that brings is also very little, such as each packet slice of hypothesis is 128 bytes, if each packet slice needs the framing information labels expense of 2 bytes, expense only accounts for 2/128=1.5% so, because the information of framing information labels is limited, need not take too much byte, so the expense under the worst case also only is 3.1%;
Further again, in order to tackle the abnormal conditions that may occur, can also be provided with the first abnormality processing submodule 4256 in the packet slice module 425, be used for receiving that the purpose cable card board resets or during abnormal signal, suspend described packet slice transmitting-receiving submodule 4254, and abandon the residue packet slice of current corresponding upstream data bag; Certainly, if this cable card board resets or state machine is unusual, because cable card board itself can carry out initialization, comprise first whole on the cable card board packet buffer modules 424, whole message length memory module 423 and all pointer zero clearings in the packet slice module 425, therefore need not the first abnormality processing submodule 4256 and operated separately; The initialization mechanism of cable card board itself is comparatively ripe, does not repeat them here;
Continuation is described other parts in the up exchange adaptation unit 420:
Uplink message sending module 426 is connected with described Flow Control module 427 and each packet slice module 425 of poll, if distribute bandwidth to satisfy slice length requirement in the packet slice module 425 of institute's poll, then send to packet switching chip by extracting a packet slice in the described packet slice module 425;
Those skilled in the art as can be seen, what uplink message sending module 426 carried out is polling dispatching, so the packet slice module 425 of each cable card board correspondence can both obtain the dispatcher meeting in a poll, does not have the problem of the isometric bag of short bag; And, because the port identity of packet switching chip, the port that promptly is connected with a certain cable card board sends a packet simultaneously at the most, therefore when this packet slice does not also have transmission not finish, can not cut into slices and send, so the reorganization of the section of up direction and down direction be all quite simple other message;
Further, can be to each the first packet buffer module, 424 prioritization, uplink message sending module 426 can be according to the corresponding packet slice module 425 of priority order poll like this;
Further, Flow Control module 427 can realize that promptly Flow Control module 427 comprises update of token bucket submodule 4271 and is respectively the token bucket 4272 that each other cable card board is provided with by the mode of token bucket; This token bucket updating submodule 4271 is connected with each token bucket 4272, is used to receive the bandwidth of each purpose cable card board dynamic assignment, periodically adjusts the token number in each token bucket; Like this, uplink message sending module 426 is by carrying out the token assessment in the corresponding token bucket 4272, just can determine distribute bandwidth whether to satisfy the packet slice length requirement of institute's poll, when meeting the demands for assessment, packet slice module 425 provides a message after the section, sends to packet switching chip by uplink message sending module 426; Because Chinese patent application CN200810112356.9 with regard to how carrying out the token assessment is described in detail, so repeat no more in the present invention.
Next, Fig. 5 shows the embodiment of a descending exchange adaptation unit 530, comprise downlink message receiving processing module 531 and downlink message sending module 534, also comprise and be respectively the message framing module 532 and the second packet buffer module 533 that each other cable card board (1～n does not comprise the cable card board at descending exchange adaptation unit 530 places) is provided with;
The packet slice that downlink message receiving processing module 531 receives from packet switching chip is transmitted sign according to the message that packet slice carries, and sends it to corresponding message framing module 532; Similar with uplink message receiving processing module 421, the corresponding relation that needs default message framing module 532 and source cable card board, this corresponding relation can be set directly in the downlink message receiving processing module 531, also can be as shown in Figure 5, be arranged on independently in the buffer memory correspondence table;
Each second packet buffer module 533 of downlink message sending module 534 polling dispatchings sends to the bag processing unit with current message in institute's poll second packet buffer module 533;
In above-mentioned descending exchange adaptation unit 530, though packet slice in order successively framing be easy to realize, but existence in view of abnormal conditions, therefore preferred embodiment is that the packet slice that up exchange adaptation unit sends carries the framing information labels, and descending exchange adaptation unit 530 is recombinated according to framing information labels in the packet slice;
Further, processing for the reply abnormal conditions, can be provided with the second abnormality processing submodule (not shown) in the message framing module 532, be used for the message framing is monitored and packet discard message in the following cases: surpass the framing that Preset Time is not finished a packet, abandon whole packet slices that received of described packet; The framing message is imperfect, abandons whole packet slices that received of this packet; And/or, the source cable card board reset or state machine unusual, abandon the whole packet slices that received of not finishing the framing packet that receive from this source cable card board; Certainly, if this cable card board resets or state machine is unusual, can directly carry out initialization by cable card board so;
Generally speaking, owing to carried out distributed Flow Control, exchange can not take place and abandon in packet slice in packet switching chip, therefore the incomplete situation of framing message much is because the unusual of source cable card board causes, be that common situation is when not receiving last packet slice of last data bag, just receive the packet slice of another one packet, at this moment, need abandon whole packet slices that received of described last data bag;
But, although possibility is very low, also can not get rid of the message disappearance that electronic jamming or other factors cause, comprise first packet slice of not receiving a certain packet or middle certain or some packet slice or the like, in this case, abandon whole packet slices that received of this packet.
Below, distributed package switching method provided by the invention to be described, this distributed package switching method is applicable to the distributed packet switching system that comprises packet switching chip and a plurality of cable card boards.
As shown in Figure 6, distributed package switching method provided by the invention may further comprise the steps:
Step S1: the source cable card board with the upstream data bag according to purpose cable card board buffer memory and cut into slices in addition respectively by certain-length;
That is, the source cable card board sends it to corresponding buffer queue according to the purpose cable card board information of carrying in the upstream data bag, adds up the message length information of described upstream data bag simultaneously; And the source cable card board extracts the message length information of described upstream data bag and cuts into slices according to certain-length;
For the ease of follow-up framing, can also may further comprise the steps among the step S1: in each packet slice, add the framing information labels, identify the position of packet slice in data packet messages at least;
Further, can select the higher message of priority to cut into slices earlier calls for follow-up poll;
Further again, because the present invention is that the packet slice with 64～1514 byte lengths is that elongated parcel (is 64～128 bytes as length) is dispatched, for the ease of those skilled in the art's better application technical scheme provided by the present invention, provide two kinds of preferable section schemes below:
Shown in Fig. 7 a, a kind of section scheme may further comprise the steps:
Step S101: the length of identification upstream data bag, if the message length N of packet is more than or equal to 256 bytes, execution in step S102; If the message length N of packet is less than 256 bytes but greater than 128 bytes, execution in step S103; If the message length N of packet is smaller or equal to 128 bytes, execution in step S104;
Step S102: according to the section of 128 bytes, if residue message length N ' re-executes step S102 more than or equal to 256 bytes; If residue message length N ' is less than 256 bytes but greater than 128 bytes, execution in step S103; If the message length N ' of remaining data bag is smaller or equal to 128 bytes, execution in step S104;
Step S103:, be the unit section then with N/2 or N '/2 if message length N or N ' are even number; If message length N or N ' are odd number, then section reaches (N+1)/2 or (N '-1)/2 and (N '+1)/2 for (N-1)/2; Certainly, for N, can last section be (N-1)/2, back one section is (N+1)/2, and perhaps, last section is (N+1)/2, and back one section is (N-1)/2; N ' in like manner;
Step S104: do not cut into slices;
As can be seen, carry out packet slice according to scheme shown in Fig. 7 a, there is not the situation less than 64 bytes in the length of section between 64～128 bytes;
Shown in Fig. 7 b, another kind of section scheme may further comprise the steps:
Step S111: the length of identification upstream data bag, if the message length N of packet is more than or equal to 192 bytes, execution in step S112; If the message length N of packet is less than 192 bytes but greater than 128 bytes, execution in step S113; If the message length N of packet is smaller or equal to 128 bytes, execution in step S114;
Step S112: according to the section of 128 bytes, if residue message length N ' re-executes step S112 more than or equal to 192 bytes; If residue message length N ' is less than 192 bytes but greater than 128 bytes, execution in step S113; If the message length N ' of remaining data bag is smaller or equal to 128 bytes, execution in step S114;
Step S113: section is 64 bytes and N-64 byte or N '-64 byte; Can select last section is 64 bytes, and can certainly select back one section is 64 bytes;
Step S114: do not cut into slices;
Equally as can be seen, carry out packet slice according to scheme shown in Fig. 7 b, the length of section also is between 64～128 bytes, does not have the situation less than 64 bytes;
Certainly, the section scheme that above-mentioned Fig. 7 a and Fig. 7 b are provided is also unrestricted, in actual applications, can the section scheme be adjusted or change according to concrete network condition;
It is to be noted, though the present invention does not limit section opportunity,, be not to carry out follow-up poll again after all section is finished with message to send as preferred embodiment, but when being polled to a certain buffer queue, just to the transmission of once cutting into slices of selected packet in this buffer queue.
Step S2: source each buffer queue of cable card board poll is sent to corresponding target cable fastener plate with the packet slice that satisfies bandwidth requirement by described packet switching chip; Wherein, described bandwidth is by described purpose cable card board dynamic assignment;
Wherein, the source cable card board can determine whether packet slice satisfies bandwidth requirement by the mode of token assessment, and is specific as follows:
The source cable card board receives the bandwidth of each purpose cable card board dynamic assignment, periodically adjusts the token number in the corresponding purpose cable card board token bucket;
The source cable card board carries out the token assessment in the corresponding token bucket, if assessment meets the demands, then bandwidth can satisfy the packet slice length requirement of institute's poll;
Further, in order to tackle the abnormal conditions that may occur, the source cable card board also carries out abnormal monitoring, monitors that this cable card board resets or state machine when unusual, carries out the cable card board initialization, i.e. whole spatial caches of initialization cable card board; Monitor that the purpose cable card board resets or during abnormal signal, suspend the transmission of packet slice, and abandon the residue message of current upstream data bag.
Step S3: the purpose cable card board is recombinated to the packet slice that receives;
In step S3, the purpose that need be in the abnormal conditions that reply may occur is equally monitored the message framing:
If surpass the framing that Preset Time is not finished a packet, abandon whole packet slices that received of described packet;
If the packet slice of a certain packet that receives is imperfect, then abandon whole packet slices that received of this packet, comprise: receive the packet slice of another one packet when not receiving last packet slice of last data bag, abandon whole packet slices that received of described last data bag; Do not receive first packet slice or the middle packet slice of a certain packet, abandon whole packet slices that received of described packet;
If a source cable card board resets or state machine is unusual, abandon the whole packet slices that received of not finishing the framing packet that receive from this source cable card board;
If this cable card board resets or state machine is unusual, then carry out the initialization of this cable card board.
Purpose cable card board dynamic bandwidth allocation technology and token assessment technology related among above-mentioned steps S1～S3 have obtained openly not repeat them here in existing patent application; The specific implementation of each step sees also the corresponding module of the distributed packet switching system of the present invention;
By above-mentioned steps S1～S3 as can be seen, distributed package switching method provided by the present invention is based on message length packet slice is carried out the Flow Control scheduling, therefore still have the not congested and dropping packets not of the packet switching chip of assurance, and the advantage that guarantees equity dispatching between each cable card board; Simultaneously,, reduced the Flow Control granularity, therefore effectively reduced the shake of data flow because original IP packet exchange section is elongated little packet switch, the uniformity of raising flow system flow, and the switching delay of packet will diminish; And each packet slice poll sends, and has avoided " the isometric bag of short bag " phenomenon.
Generally speaking, distributed packet switching system provided by the invention and distributed package switching method need not existing packet switching chip is changed, and realize simply relatively, and cost is also relatively low, and can promote the QoS ability of network equipment system effectively.
Though described the present invention with reference to several exemplary embodiments, should be appreciated that used term is explanation and exemplary and nonrestrictive term.The spirit or the essence that do not break away from invention because the present invention can specifically implement in a variety of forms, so be to be understood that, the foregoing description is not limited to any aforesaid details, and should be in the spirit and scope that claim limited of enclosing explain widely, therefore fall into whole variations in claim or its equivalent scope and remodeling and all should be the claim of enclosing and contain.
Claims (17)
1. a distributed packet switching system comprises packet switching chip and a plurality of cable card board; It is characterized in that up exchange adaptation unit and descending exchange adaptation unit that each cable card board is respectively equipped with a bag processing unit and connects described bag processing unit;
The up exchange adaptation unit of source cable card board receives the upstream data bag from the bag processing unit of this cable card board, according to corresponding purpose cable card board buffer memory and cut into slices in addition respectively by certain-length, and, each buffer queue of poll is sent to the purpose cable card board with the packet slice that satisfies bandwidth requirement by described packet switching chip;
The descending exchange adaptation unit of purpose cable card board is the up exchange adaptation unit dynamic bandwidth allocation of source cable card board, packet slice that reorganization receives from packet switching chip and the bag processing unit that sends to this cable card board.
2. distributed packet switching system according to claim 1, it is characterized in that, described up exchange adaptation unit comprises uplink message receiving processing module, message length statistical module, uplink message sending module and Flow Control module, also comprises being respectively message length memory module, the first packet buffer module and the packet slice module that each other cable card board is provided with;
Described uplink message receiving processing module receives the upstream data bag from the bag processing unit, according to the purpose cable card board information of carrying in the described upstream data bag, sends it to the corresponding first packet buffer module;
Described message length statistical module is connected with described uplink message receiving processing module, adds up the message length information of described upstream data bag, and stores corresponding message length memory cell into;
Each packet slice module is connected with the first packet buffer module and the message length memory module of homologous lines clamp respectively, extracts the message length information of upstream data bag and cuts into slices according to certain-length;
Described Flow Control module is used to receive the bandwidth of each purpose cable card board dynamic assignment;
Described uplink message sending module is connected with described Flow Control module and each packet slice module of poll, if distribute bandwidth to satisfy slice length requirement in the packet slice module of institute's poll, then send to described packet switching chip by extracting a packet slice in the described packet slice module.
3. distributed packet switching system according to claim 2 is characterized in that, described Flow Control module comprises the update of token bucket submodule and is respectively the token bucket that each other cable card board is provided with;
Described update of token bucket submodule is connected with each token bucket, is used to receive the bandwidth of each purpose cable card board dynamic assignment, periodically adjusts the token number in each token bucket;
Described uplink message sending module is by carrying out the token assessment in the corresponding token bucket, determine distribute bandwidth whether to satisfy the packet slice length requirement of institute's poll.
4. distributed packet switching system according to claim 2 is characterized in that, is provided with in the described packet slice module:
Packet pointer is pointed to current upstream data packet address;
The message length pointer points to the address of storing current data packet message length information in the message length memory cell;
The packet slice pointer points to current packet slice address;
Packet slice transmitting-receiving submodule, be used for sending to described uplink message sending module according to the message data of above-mentioned pointer extracting certain-length, and with the address behind the described certain-length of described current packet slice address offset as new current packet slice address, carry out the redirect of described packet slice pointer.
5. distributed packet switching system according to claim 4, it is characterized in that, described packet slice module also is provided with message priority recognin module, be used for discerning the priority that is buffered in the first packet buffer module data packet messages, the data packet messages that makes packet pointer and message length pointed have higher priority is cut into slices.
6. distributed packet switching system according to claim 4 is characterized in that, the described first packet buffer module segmentation is the cache sub-module of first fixed byte, and each cache sub-module is used for data packet voice of buffer memory; And/or described message length memory cell is cut apart according to second fixed byte, and the space after each is cut apart is used to preserve the length information of a data packet voice; Then after current upstream data bag is sent completely, described current upstream data packet address is offset described first fixed byte as new current upstream data packet address, carry out described packet pointer redirect, and/or, with storing the address of described second fixed byte of address offset of current data packet message length information, carry out the redirect of described message length pointer as new storage current data packet message length information.
7. according to the arbitrary described distributed packet switching system of claim 4-6, it is characterized in that, also be provided with framing information labels submodule in the described packet slice module, be used for adding the framing information labels, identify the position of packet slice in data packet messages at least at each packet slice.
8. according to the arbitrary described distributed packet switching system of claim 4-6, it is characterized in that, also be provided with the first abnormality processing submodule in the described packet slice module, be used for receiving that the purpose cable card board resets or during abnormal signal, suspend described packet slice transmitting-receiving submodule, and abandon the residue packet slice of current upstream data bag.
9. distributed packet switching system according to claim 1, it is characterized in that, described descending exchange adaptation unit comprises downlink message receiving processing module and downlink message sending module, also comprises being respectively the message framing module and the second packet buffer module that each other cable card board is provided with;
Described downlink message receiving processing module receives the packet slice from described packet switching chip, transmits sign according to the message that described packet slice carries, and sends it to corresponding message framing module;
Described message framing module is carried out the message reorganization and is sent to the corresponding second packet buffer module;
Each second packet buffer module of described downlink message sending module polling dispatching sends to the bag processing unit with current message in institute's poll second packet buffer module.
10. distributed packet switching system according to claim 9, it is characterized in that, be provided with the second abnormality processing submodule in the described message framing module, be used for the message framing is monitored and packet discard message in the following cases: surpass Preset Time and do not finish framing, the framing message is imperfect, and/or, the source cable card board reset or state machine unusual.
11. a distributed package switching method is applicable to the distributed packet switching system that comprises packet switching chip and a plurality of cable card boards; It is characterized in that, may further comprise the steps:
Step S1: the source cable card board with the upstream data bag according to purpose cable card board buffer memory and cut into slices in addition respectively by certain-length;
Step S2: source each buffer queue of cable card board poll is sent to corresponding target cable fastener plate with the packet slice that satisfies bandwidth requirement by described packet switching chip; Wherein, described bandwidth is by described purpose cable card board dynamic assignment;
Step S3: the purpose cable card board is recombinated to the packet slice that receives.
12. distributed package switching method according to claim 11 is characterized in that, and is further comprising the steps of among the described step S1: in each packet slice, add the framing information labels, identify the position of packet slice in data packet messages at least.
13., it is characterized in that described step of cutting into slices according to certain-length comprises according to claim 11 or 12 described distributed package switching methods:
Step S101: the length of identification upstream data bag, if the message length N of packet is more than or equal to 256 bytes, execution in step S102; If the message length N of packet is less than 256 bytes but greater than 128 bytes, execution in step S103; If the message length N of packet is smaller or equal to 128 bytes, execution in step S104;
Step S102: according to the section of 128 bytes, if residue message length N ' re-executes step S102 more than or equal to 256 bytes; If residue message length N ' is less than 256 bytes but greater than 128 bytes, execution in step S103; If the message length N ' of remaining data bag is smaller or equal to 128 bytes, execution in step S104;
Step S103:, be the unit section then with N/2 or N '/2 if message length N or N ' are even number; If message length N or N ' are odd number, then section reaches (N+1)/2 or (N '-1)/2 and (N '+1)/2 for (N-1)/2;
Step S104: do not cut into slices.
14., it is characterized in that described step of cutting into slices according to certain-length comprises according to claim 11 or 12 described distributed package switching methods:
Step S111: the length of identification upstream data bag, if the message length N of packet is more than or equal to 192 bytes, execution in step S112; If the message length N of packet is less than 192 bytes but greater than 128 bytes, execution in step S113; If the message length N of packet is smaller or equal to 128 bytes, execution in step S114;
Step S112: according to the section of 128 bytes, if residue message length N ' re-executes step S112 more than or equal to 192 bytes; If residue message length N ' is less than 192 bytes but greater than 128 bytes, execution in step S113; If the message length N ' of remaining data bag is smaller or equal to 128 bytes, execution in step S114;
Step S113: section is 64 bytes and N-64 byte or N '-64 byte;
Step S114: do not cut into slices.
15. distributed package switching method according to claim 11 is characterized in that, the source cable card board is also carried out following steps among the described step S2:
Receive the bandwidth of each purpose cable card board dynamic assignment, periodically adjust the token number in the corresponding purpose cable card board token bucket;
Carry out the token assessment in the corresponding token bucket, if assessment meets the demands, then bandwidth can satisfy the packet slice length requirement of institute's poll.
16., it is characterized in that described step S2 source cable card board is also carried out following steps according to claim 11 or 15 described distributed package switching methods:
Carry out abnormal monitoring;
Monitor that this cable card board resets or state machine when unusual, carry out the cable card board initialization;
Monitor that the purpose cable card board resets or during abnormal signal, suspend the transmission of packet slice, and abandon the residue message of current upstream data bag.
17. distributed package switching method according to claim 11 is characterized in that, described step S3 is further comprising the steps of:
The message framing is monitored;
If surpass the framing that Preset Time is not finished a packet, abandon whole packet slices that received of described packet;
If the packet slice of a certain packet that receives is imperfect, abandon whole packet slices that received of described packet;
If a source cable card board resets or state machine is unusual, abandon the whole packet slices that received of not finishing the framing packet that receive from this source cable card board;
If this cable card board resets or state machine is unusual, carry out the cable card board initialization.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101476724A CN101567855B (en) | 2009-06-11 | 2009-06-11 | Distributed packet switching system and distributed packet switching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009101476724A CN101567855B (en) | 2009-06-11 | 2009-06-11 | Distributed packet switching system and distributed packet switching method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101567855A CN101567855A (en) | 2009-10-28 |
| CN101567855B true CN101567855B (en) | 2011-12-21 |
Family
ID=41283811
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009101476724A Active CN101567855B (en) | 2009-06-11 | 2009-06-11 | Distributed packet switching system and distributed packet switching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101567855B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102377631B (en) * | 2010-08-06 | 2015-08-05 | 北京乾唐视联网络科技有限公司 | A kind of data transmission method based on flow control and communication system |
| CN102006226B (en) * | 2010-11-19 | 2013-01-02 | 福建星网锐捷网络有限公司 | Message cache management method and device as well as network equipment |
| CN102123080B (en) * | 2010-12-31 | 2014-02-05 | 杭州华三通信技术有限公司 | Data forwarding device and cable fastener plate and data forwarding method |
| CN102088412B (en) * | 2011-03-02 | 2014-09-03 | 华为技术有限公司 | Exchange unit chip, router and transmission method of cell information |
| CN104572498B (en) * | 2014-12-26 | 2018-04-27 | 曙光信息产业(北京)有限公司 | The buffer memory management method and device of message |
| CN104954292B (en) * | 2015-05-18 | 2018-04-20 | 烽火通信科技股份有限公司 | The system and method for data packet cutting and restructuring based on CLOS exchange networks |
| CN104869081B (en) * | 2015-05-29 | 2018-09-07 | 华为技术有限公司 | MESSAGE EXCHANGE processing method, business board and internet exchange system |
| CN105591986A (en) * | 2015-08-26 | 2016-05-18 | 杭州华三通信技术有限公司 | Method and device for processing broken message |
| CN107592248A (en) * | 2016-07-08 | 2018-01-16 | 成都夸克光电技术有限公司 | A kind of high efficiency of transmission control method based on industry ethernet irregular data |
| CN112217733B (en) * | 2019-07-09 | 2024-02-02 | 中兴通讯股份有限公司 | Message processing method and related device |
| CN114629841B (en) * | 2020-11-27 | 2023-05-16 | 华为技术有限公司 | Communication method, device and system |
| CN113518046B (en) * | 2021-03-31 | 2023-12-26 | 新华三信息安全技术有限公司 | Message forwarding method and frame type switching equipment |
-
2009
- 2009-06-11 CN CN2009101476724A patent/CN101567855B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN101567855A (en) | 2009-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101567855B (en) | Distributed packet switching system and distributed packet switching method | |
| US6654342B1 (en) | Accumulating and distributing flow control information via update messages and piggybacked flow control information in other messages in a packet switching system | |
| US5936958A (en) | ATM exchange for monitoring congestion and allocating and transmitting bandwidth-guaranteed and non-bandwidth-guaranteed connection calls | |
| US6735173B1 (en) | Method and apparatus for accumulating and distributing data items within a packet switching system | |
| US7149212B2 (en) | Apparatus, method and limited set of messages to transmit data between scheduler and a network processor | |
| US6092108A (en) | Dynamic threshold packet filtering of application processor frames | |
| CA2119205C (en) | Improvements in or relating to asynchronous transfer mode communication systems | |
| US5577035A (en) | Apparatus and method of processing bandwidth requirements in an ATM switch | |
| US7304987B1 (en) | System and method for synchronizing switch fabric backplane link management credit counters | |
| CN101668005B (en) | Data transmission accelerating engine method based on multiple access passages of transmitting end | |
| CN101299721B (en) | Method for switching message of switching network, and switching device | |
| US20080196033A1 (en) | Method and device for processing network data | |
| CN101572673B (en) | Distributed packet switching system and distributed packet switching method of expanded switching bandwidth | |
| US20010007562A1 (en) | Packet switch device and scheduling control method | |
| EP1508227B1 (en) | Buffer memory reservation | |
| US20050135356A1 (en) | Switching device utilizing requests indicating cumulative amount of data | |
| US6636510B1 (en) | Multicast methodology and apparatus for backpressure-based switching fabric | |
| CN102546098B (en) | Data transmission device, method and system | |
| US6728211B1 (en) | Method and apparatus for delaying packets being sent from a component of a packet switching system | |
| US7392355B2 (en) | Memory sharing mechanism based on priority elevation | |
| JP2000252997A (en) | Traffic shaper | |
| CN100466603C (en) | Method and apparatus for shaping transmission service stream in network | |
| CN108540221B (en) | Data sending method and device | |
| CN104303459A (en) | Method and apparatus for implementing traffic flags for large service groups | |
| US7139253B2 (en) | Packet switches |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address |
Address after: 310052 Binjiang District Changhe Road, Zhejiang, China, No. 466, No. Patentee after: Xinhua three Technology Co., Ltd. Address before: 310053 Hangzhou hi tech Industrial Development Zone, Zhejiang province science and Technology Industrial Park, No. 310 and No. six road, HUAWEI, Hangzhou production base Patentee before: Huasan Communication Technology Co., Ltd. |
|
| CP03 | Change of name, title or address |