CN103716208B - Support network management, system, interchanger and the network of elephant stream - Google Patents

Support network management, system, interchanger and the network of elephant stream Download PDF

Info

Publication number
CN103716208B
CN103716208B CN201310752282.6A CN201310752282A CN103716208B CN 103716208 B CN103716208 B CN 103716208B CN 201310752282 A CN201310752282 A CN 201310752282A CN 103716208 B CN103716208 B CN 103716208B
Authority
CN
China
Prior art keywords
path
network
elephant
weight
network switch
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
Application number
CN201310752282.6A
Other languages
Chinese (zh)
Other versions
CN103716208A (en
Inventor
寿国础
李洁
胡怡红
郭志刚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing University of Posts and Telecommunications
Original Assignee
Beijing University of Posts and Telecommunications
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing University of Posts and Telecommunications filed Critical Beijing University of Posts and Telecommunications
Priority to CN201310752282.6A priority Critical patent/CN103716208B/en
Publication of CN103716208A publication Critical patent/CN103716208A/en
Application granted granted Critical
Publication of CN103716208B publication Critical patent/CN103716208B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Abstract

The invention discloses a kind of network management for supporting elephant to flow, NMS, the network switch and communication network, wherein support the network management of elephant stream including:When the elephant in detecting network is flowed, the source address and destination address of elephant stream are extracted, obtain the multiple paths between source address and destination address, the weight of the respective path for supporting elephant to flow is calculated using the offered load and network topology structure in the path;NMS generates the forward table of the network switch, and send to the network switch forward table, the forward table includes the weight and the operational order of the network switch of the respective path for supporting elephant stream, the forward table is performed for the network switch, the flow that elephant is flowed is distributed according to the weight in path.The communication network that there is elephant stream can be made to keep flow equalization using the above method and system and device, network performance and resource utilization is effectively improved.

Description

Support network management, system, interchanger and the network of elephant stream
Technical field
The present invention relates to network communication field, more particularly to a kind of network management, the network management of supporting elephant to flow System, the network switch and communication network.
Background technology
Existing multi-path routing technology, such as equal cost multipath(Equal-Cost Multi path Routing, ECMP) Deng stream is divided on a plurality of available equative route at random using the static Hashing mapping techniques based on data flow.However, all Data flow, either their size(Byte number), or their duration all differ, so this is simple Strategy is not enough to prevent the generation of network hotspot.This strategy based on Hash mapping only only has mouse stream in a network(Byte The small data flow of number)Without elephant stream(The big data flow of byte number)In the case of just effectively.Particularly, two long-term existence Elephant stream same path may be mapped within a very long time, therefore caused congestion can decline network performance, And at the same time may also have the idle capacity on backup path available in network, therefore use existing Multi-path route Technology, the decline of network performance is easily caused in the case where there is elephant stream.
The content of the invention
The invention provides a kind of network management for supporting elephant to flow, NMS, the network switch and logical Communication network, calculate the corresponding road that support elephant is flowed using the offered load and network topology structure in path when elephant is flowed detecting The weight in footpath, and form corresponding forward table, then the flow that the network switch can flow elephant according to forward table is according to weight point Hair, it is ensured that the performance of network.
The invention provides a kind of network management for supporting elephant to flow, the method includes:
When the elephant in detecting network is flowed, the source address and destination address of elephant stream are extracted, obtain source address and mesh Address between multiple paths, using the path offered load and network topology structure calculate support elephant flow correspondence The weight in path;
NMS generates the forward table of the network switch, and forward table is sent to the network switch, described turn The weight and the operational order of the network switch of the respective path including supporting elephant stream are delivered, is performed for the network switch described Forward table, the flow that elephant is flowed is distributed according to the weight in path.
Preferably, the flow table includes destination address and corresponding action fields, and the action fields are respectively directed to a group table; Described group of table includes multiple action items respectively, and the action items include the operation of the weight and the network switch in path respectively Instruction.
Preferably, the method also includes:Generation LLDP messages are simultaneously periodically issued to the network switch;Receive the network switch The LLDP messages of upload, network topology structure is obtained by the LLDP messages for receiving;The statistical number of the periodic polling network switch According to obtain the load of the network switch and the load of network link.
Preferably, the offered load and network topology structure using the path calculates the corresponding road for supporting elephant stream The weight in footpath, including:
Acquisition represents the set of all available paths between source address and destination address, and the number for making the path is I;It is right Each path i(0≤i≤I-1)Perform following operation:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1); Make λkji(0≤k≤Kji-1)It is the load of link k;Make ckji(0≤k≤Kji-1)It is the capacity of link k;
Calculate the link utilization of each link:LUkjiIt is the link utilization of link k;
Calculate the weight in path:wi∈ 0,1 ..., and 9 } it is path i Weight.
The present invention also provides a kind of NMS, and the NMS includes:
Elephant flow path computing unit, in detecting network elephant flow when, extract elephant stream source address and Destination address, obtains the multiple paths between source address and destination address, using the offered load and network topology in the path Structure Calculation supports the weight of the respective path of elephant stream;
Path management unit, the forward table for generating the network switch, and forward table is sent to the network switch, institute Stating forward table includes the weight and the operational order of the network switch of the respective path for supporting elephant stream, is performed for the network switch The forward table, the flow that elephant is flowed is distributed according to the weight in path, it is preferred that being distributed to the corresponding next-hop in the path Network node.
Preferably, the NMS also includes:Network topology unit, for generating LLDP messages and periodically issuing To the network switch, and the LLDP messages that the network switch is uploaded are received, network topology is obtained by the LLDP messages for receiving Structure;Offered load monitoring unit, for the statistics of the periodic polling network switch, to obtain the load of the network switch With the load of network link.
Preferably, the elephant flow path computing unit, for offered load and network topology knot using the path Structure calculates the weight in path;
Wherein, the elephant flow path computing unit, all between source address and destination address can use is represented for obtaining The set in path, the number for making the path is I;The elephant flow path computing unit, is further used for each path i(0 ≤i≤I-1)Following operation is performed, to calculate the weight in path:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1); The load for making link k is λkji(0≤k≤Kji-1);The capacity for making link k is ckji(0≤k≤Kji-1);
Calculate the link utilization of each link:LUkjiIt is the link utilization of link k;
Calculate the weight in path:wi∈ 0,1 ..., and 9 } it is path i Weight.
The present invention also provides a kind of network switch, including:Elephant flows reporting unit, for from the data flow for receiving Choose the data flow labeled as elephant stream and selected data flow is reported into NMS;Retransmission unit, for connecing The forward table of NMS transmission is received and performs, the flow that elephant is flowed is distributed according to the weight in forward table, it is preferred that point It is dealt into the corresponding next-hop network node in path.
The present invention also provides a kind of communication network, including:NMS as described above;And, signal is connected to The network switch as described above of the NMS.
The beneficial effect of the embodiment of the present invention is:By the load information according to each path of network, support elephant is calculated The weight of the respective path of stream, the traffic distribution that then interchanger flows elephant according to the weight in each path, can improve network Performance.By detect elephant flow and for elephant stream carry out calculate path weight treatment, and for mouse stream then use base This routing algorithm, can control resource overhead while network performance is improved.
Brief description of the drawings
Fig. 1 is a kind of flow chart of network management for supporting elephant to flow of one embodiment of the invention;
Fig. 2 is the detailed maps of forward table of the invention;
Fig. 3 is a kind of structured flowchart of communication network of one embodiment of the invention;
Fig. 4 is a kind of structured flowchart of network switch of one embodiment of the invention;
Fig. 5 is a kind of structured flowchart of NMS of one embodiment of the invention;
Fig. 6(a)It is the schematic diagram using a kind of application scenarios of equal cost multipath routing algorithm;
Fig. 6(b)It is a kind of schematic diagram of application scenarios of one embodiment of the invention;
Fig. 7 is the contrast schematic diagram with the network delay using prior art using the present invention;
Fig. 8 is the contrast schematic diagram with the network throughput using prior art using the present invention;
Fig. 9 is the contrast schematic diagram with the link utilization using prior art using the present invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with accompanying drawing to embodiment party of the present invention Formula is described in further detail.
Technology design of the invention is essentially consisted in, by the weight for calculating the respective path for supporting elephant to flow(Load is light Path weight value is higher)So that elephant stream can be distributed according to the load of network path.Because the path weight value for loading light is high, Therefore it is big to the flow for loading the elephant stream distributed in light path, so as to reach load balancing with the overall performance for improving network Technique effect.
Fig. 1 is a kind of flow chart of network management for supporting elephant to flow of one embodiment of the invention.As shown in figure 1, The method is comprised the following steps:
S100, in detecting network elephant flow when, extract elephant stream source address and destination address, obtain source address Multiple paths and destination address between, are calculated using the offered load and network topology structure in the path and support what elephant was flowed The weight of respective path.
The forward table of S200, the generation network switch, and forward table is sent to the network switch, the forward table includes The weight and the operational order of the network switch of the respective path of elephant stream are supported, the forward table are performed for the network switch, The flow that elephant is flowed is distributed according to the weight in path, it is preferred that the flow that elephant is flowed is distributed into institute according to the weight in path State the corresponding next-hop network node in path.
By calculating the weight of the respective path for supporting elephant to flow, generation includes the weight of the respective path for supporting elephant stream With the forward table of the operational order of the network switch and send to interchanger so that interchanger can flow elephant according to forward table According to the load distribution of network path.Include the operational order of the network switch due to forward table, therefore the network switch can The corresponding next-hop network node in path is determined according to operational order(Interchanger or remaining network equipment)And send data flow To the network node;And include supporting the weight of the respective path of elephant stream due to forward table, therefore the network switch can Weight distribution elephant according to path is flowed.It is preferred that the flow of the elephant stream of distribution is directly proportional to weight.
Further, the ratio for accounting for network traffics due to the mouse stream in network is very big and above-mentioned using the path The weight that offered load and network topology structure calculate each path needs certain calculation resources expense, if flowed for mouse Equally carrying out treatment using the above method of the invention, to be easily caused resource overhead excessive.Therefore the present invention does not have for mouse stream Carry out aforesaid operations.It is preferred that using equal cost multipath for mouse stream(ECMP)Mode carry out Multi-path route.
Further, the method for the elephant stream in detection network can have various, for example, carried out in network edge interchanger Detect and mark etc..
Further, the method also includes:Generation LLDP(Link Layer Discovery Protocol, link layer hair Existing agreement)Message is simultaneously periodically issued to the network switch;The LLDP messages that the network switch is uploaded are received, by the LLDP for receiving Message obtains network topology structure.Specifically, maker function will be called every a predetermined time out period, in each tune With period, network management device generation LLDP messages, and LLDP messages are sent the network switch managed to it, receive The network switch of LLDP messages is forwarded the packet to the adjacent network switch, receives what is forwarded by remaining network switch The LLDP messages are uploaded to network management device by the network switch of LLDP messages.The LLDP that network management device passes through reception Message infers link level connection, and then obtains network topology structure.It is preferred that network topology structure is stored in a neighbour Connect in table, network topology structure can be expressed as connected graph G=(V, E), wherein V is the set of node, and E is the set of directed edge.
Further, the method also includes:The statistics of the periodic polling network switch, to obtain the network switch Load and the load of network link.Specifically, the statistics of all of interchanger can be regularly inquired about, consolidates and store, These data, can be used in the load and the load of link of the interchanger of calculating communication network, and then for elephant flow path Calculate.It is preferred that from network on all of interchanger, based on the statistical information that each table, each stream and each port are collected, During internal memory is stored in the form of the snapshot object., using a time series number as mark, it is in each time for each snapshot Increase 1 behind interval such that it is able to recognize the network condition of special time.In order to save memory space, 2 only nearest snapshot quilts It is maintained in internal memory.
In the present embodiment, the offered load and network topology structure using the path is calculated supports the right of elephant stream The weight in path is answered, using following weighting routing algorithms, is specifically included:
Acquisition represents the set of all available paths between source address and destination address, makes the number in the path for I, tool Body ground, can be detected from interchanger be marked as elephant stream data flow in data flow source address and destination address, And then obtain available path using various Network Topology Algorithms.Further, to each path i(0≤i≤I-1)Perform with Lower operation:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1); Make λkji(0≤k≤Kji-1)It is the load of link k;Make ckji(0≤k≤Kji-1)It is the capacity of link k.
It is possible to pass through below equation(1)The link utilization of each link is calculated, wherein, LUkjiIt is the link of link k Utilization rate.
Can be by below equation(2)The weight in path is calculated, wherein, wi∈ 0,1 ..., and 9 } it is the weight of path i.
In the present embodiment, the assignment of traffic ratio of the respective path flowed from the support elephant of some interchanger outflow, with The weight in the path is directly proportional.For example, having four, power from the path of the one elephant stream of support of some interchanger outflow 4,6,7,3 are respectively again, then the flow for flowing elephant is distributed according to 20%, 30%, 35%, 15% ratio.
In the present embodiment, weight path load high is light, the low path load weight of weight, due to the stream that the elephant distributed is flowed Amount is directly proportional to weight, so not distributing the flow of elephant stream to the path if it there is the path that weight is 0, remaining path is pressed According to the flow of the ratio distribution elephant stream being directly proportional to weight.For example, it is assumed that in the presence of four paths, weight is respectively 0,2,3,5, The flow that then elephant is flowed is distributed according to 0,20%, 30%, 50% ratio, certainly, weight can also be directly excluded in the present embodiment It is 0 path(Refering to the Exclude the overutilized path in false code)So that the road of load too high is not utilized The flow of footpath forwarding elephant stream, so that the performance in the path of load too high is improved faster.
It is because the path weight value for loading light is high therefore big to the flow for loading the elephant stream distributed in light path, so as to reach To load balancing, the overall performance of network is improved.
In order to weighting routing algorithm of the invention is better described, the false code of weighting routing algorithm is provided, false code is such as Shown in lower.
Fig. 2 is the detailed view of forward table of the invention.The forward table includes flow table(Flow Table)With a group table (Group Table).
The flow table(Flow Table)Including destination address(IP Dst)With corresponding action fields(Instructions), The action fields are respectively directed to a group table, specifically, by a group table id(Group ID)Point to a group table.
Described group of table(Group Table)Include multiple action items respectively(Action Buckets), the action item Mesh includes the weight in path respectively(Weight)With the operational order of the network switch(Action).
More specifically, the present embodiment using cooperatively by flow table and group table, to support Multi-path route.Multi-path route It is to be realized by the way that the action fields in flow table are pointed into a group table.Each group table is made up of a set group set of actions, each Group set of actions includes a set of action that will be applied onto in the data flow that the match is successful.Each set has a weight field, it Define the share of the flow as handled by the flow handled by the set accounts for whole group of table.The stream of the elephant stream that each is detected Action fields in table can all point to a group table.Group table includes set of actions, and each set is related to a paths, and elephant stream can Location is arrived at via these paths.
Fig. 3 is a kind of structured flowchart of communication network of one embodiment of the invention.Communication network includes NMS 3rd, multiple network switch 2, only symbolically show a network switch 2 among Fig. 3, and Practical Project is not limited.
Fig. 4 is a kind of structured flowchart of network switch of one embodiment of the invention.The network switch 2, including elephant stream Reporting unit 21 and retransmission unit 22.
Elephant flows reporting unit 21, for choosing data flow labeled as elephant stream from the data flow for receiving and by institute The data flow of selection reports to NMS 3.Specifically by selecting elephant stream packets(Marked Packets) And selected elephant stream packets are reported into NMS 3, so that NMS 3 is calculated according to packet The weight in path.In the present embodiment, unmarked is the packet of elephant stream(Unmarked Packets as depicted)And mark It is designated as elephant stream packets(Marked Packets as depicted)It is the network switch 2 from the network terminal or remaining network The received data bag of interchanger 2.
Retransmission unit 22, for receiving and perform NMS 3 transmission forward table, by elephant flow flow according to Weight in forward table(Weight)Distribution, it is preferred that being distributed to the corresponding next-hop network node in the path(For example it is next Network switch or the network terminal or core network device of jump etc.).It is preferred that using equal cost multipath for mouse stream (ECMP)Mode carry out Multi-path route.
Fig. 5 is a kind of structured flowchart of NMS of one embodiment of the invention.NMS 3 includes elephant Flow path computing unit 31 and path management unit 32.
Elephant flow path computing unit 31, when the elephant in detecting network is flowed, extracts the source address and mesh of elephant stream Address, the multiple paths between source address and destination address are obtained, using the offered load and network topology knot in the path Structure calculates the weight of the respective path for supporting elephant stream.
Path management unit 32, the forward table for generating the network switch, and forward table is sent to the network switch, The forward table includes the weight and the operational order of the network switch of the respective path for supporting elephant stream, is held for the network switch The row forward table, the flow that elephant is flowed is distributed according to the weight in path, it is preferred that it is corresponding next to be distributed to the path Jump network node.
The present invention calculates the weight in path by elephant flow path computing unit 31, and the generation of path management unit 32 includes road The forward table of the weight in footpath and the operational order of the network switch is simultaneously sent to the network switch 2 so that the network switch 2 can Elephant is flowed the load distribution according to network path according to forward table.Include the operational order of the network switch due to forward table, Therefore the network switch 2 can determine the corresponding next-hop network node in path according to operational order(Interchanger or remaining network Equipment)And send to the network node data flow;And include the weight in path, therefore the network switch 2 due to forward table Can be flowed according to weight distribution elephant, the data volume of distribution is directly proportional to weight.
Further, the ratio for accounting for network traffics due to the mouse stream in network is very big and above-mentioned using the path The weight that offered load and network topology structure calculate each path needs certain calculation resources expense, if flowed for mouse It is same above-mentioned treatment to be carried out by elephant flow path computing unit 31 to be easily caused resource overhead excessive.Therefore the present invention is for mouse Stream is not carried out the operation of above-mentioned calculating weight by elephant flow path computing unit 31, it is preferred that being used for mouse stream of equal value Multipath(ECMP)Mode carry out Multi-path route.
The NMS 3 of the present embodiment is not provided with the device for detecting the stream of the elephant in network.This area skill Art personnel are it will be appreciated that In remaining embodiment NMS 3 can set up the device for detecting the stream of the elephant in network.
Further, NMS 3 also includes network topology unit 33 and offered load monitoring unit 34.
Network topology unit 33, network topology unit, for generating LLDP messages(Link Layer Discovery Protocol, Link Layer Discovery Protocol)And the network switch is periodically issued to, and receive the LLDP reports that the network switch is uploaded Text, network topology structure is obtained by the LLDP messages for receiving.Specifically, every a predetermined time out period network topology list Unit 33 will call maker function, during each is called, the generation LLDP messages of network topology unit 33, and by LLDP messages The network switch 2 managed to it is sent, LLDP messages are sent to the adjacent network switch 2, received by the network switch 2 The LLDP messages are uploaded to network topology unit 33 by the network switch 2 of the LLDP messages forwarded by remaining network switch 2. Network topology unit 33 infers link level connection by the LLDP messages for receiving, and then obtains network topology structure.Preferably Ground, network topology structure is stored in an adjacency list, and network topology structure can be expressed as connected graph G=(V, E), wherein V It is the set of node, E is the set of directed edge.
Offered load monitoring unit 34, for the statistics of the periodic polling network switch, to obtain the network switch Load and network link load.Specifically, offered load monitoring unit 34 can regularly be inquired about, consolidate and store all The network switch 2 statistics, these data, can be used in calculate communication network the network switch 2 load and chain The load on road, and then for the calculating of elephant flow path.It is preferred that from network on all of network switch 2, based on each The statistical information that table, each stream and each port are collected, in being stored in internal memory in the form of the snapshot object.Each snapshot is with one Used as mark, it increases 1 to individual time series number after each time interval such that it is able to recognize the network condition of special time.For Saving memory space, only nearest 2 snapshots are maintained in internal memory.
In the present embodiment, elephant flow path computing unit 31 is calculated using the offered load and network topology structure in path The weight of the respective path of elephant stream is supported, including:
Acquisition represents the set of all available paths between source address and destination address, makes the number in the path for I, tool Body ground, can be from the packet for being marked as elephant stream(Marked Packets as shown in Figure 3)In data flow source ground Location and destination address, and then obtain available path using various Network Topology Algorithms.Further, to each path i(0≤i ≤I-1)Following operation is performed, to calculate the weight in path:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1); Make λkji(0≤k≤Kji-1)It is the load of link k;Make ckji(0≤k≤Kji-1)It is the capacity of link k.
It is possible to pass through below equation(1)The link utilization of each link is calculated, wherein, LUkjiIt is the link of link k Utilization rate.
Can be by below equation(2)The weight in path is calculated, wherein, wi∈ 0,1 ..., and 9 } it is the weight of path i.
In the present embodiment, by the calculating of elephant flow path computing unit 31, the support from the outflow of some interchanger is big As the assignment of traffic ratio of the respective path of stream, the weight to the path is directly proportional.For example, flowed out from some interchanger Support that the path of elephant stream has four, weight is respectively 4,6,7,3, then the flow for flowing elephant is according to 20%, 30%, 35%th, 15% ratio distribution.
It is because the path weight value for loading light is high therefore big to the flow for loading the elephant stream distributed in light path, so as to reach To load balancing, the overall performance of network is improved.
Understand from the above mentioned, in the present embodiment, the network switch 2(For example, OpenFlow interchangers)Elephant stream report Labeled data are forwarded a packet to NMS 3 by unit 21.The elephant flow path computing unit 31 of NMS 3 After receiving packet, the source address and destination address of data flow will be extracted, then by consultation network topology unit 33, to obtain Topology information between source host and destination host, and calculate all possible shortest path according to these information.Once it is all After shortest path is computed, the statistical information provided by offered load monitoring unit 34 will be used to calculate each path Load.Then elephant flow path computing unit 31 is calculated according to the loading condition in current each path with weighting Multi-path route Method, calculates the weight of each path candidate, and the path obtained after calculating and its weight information are passed into path management unit 32, it is responsible for being finally issued to these path integrations into the stream entry of related network switch 2 on the network switch 2 by it.
Fig. 6(a)It is the schematic diagram using a kind of application scenarios of equal cost multipath routing algorithm.Fig. 6(b)It is the present invention one A kind of schematic diagram of the application scenarios of embodiment.Fig. 6(a)And Fig. 6(b)Describe an application scenarios of data center, two Server in Pod1 transmits two elephants to same destination Pod4 and flows simultaneously, transfer rate be respectively 500Mbps and 300Mbps.This communication pattern is in the data center very common, such as distributed big data treatment application will produce this Plant communication pattern.Assuming that the port 1 in the core layer network switch 1 has been processed by big batch stream, port utilization ratio has reached 80%, and the port 1 of the network switch 2 is in light condition, port utilization ratio only has 40%.If using equal cost multipath road By algorithm(ECMP), the big stream that the two speed are big, existence time is long is it may happen that Hash is collided and final all trends are negative Carry the port 1 in the network switch 1 very high.Assuming that link bandwidth is 1Gbps, required speed after two elephant stream polymerizations It is 800Mbps, it has exceeded the residual available bandwidth of the port 1 of the network switch 1(200Mbps).So will now gather around Plug, causes the service quality of the application for producing elephant stream to decline.In this example, if using scheme disclosed by the invention If, the two elephants flow the transmission rate needed for just can all reaching them.Because the port 1 of core layer switch 1 is most Extra 200Mbps can only be transmitted, this can not even meet the less elephant stream of bandwidth requirement(Required speed is 300Mbps).Therefore, optimal forwarding scheme is that two elephant streams are polymerized at the network switch AGG1 of network polymerization layer, Then by the elephant of polymerization stream according to the ratio segmentation being inversely proportional with link load be forwarded to core layer switch 1 port 1 and At the port 1 of interchanger 2, for example, the flow rate of both links transmission is respectively 200Mbps, 600Mbps.So solve Congestion problems, improve network utilization, will not produce infringement to the service quality for producing the application of elephant stream.
Fig. 7 is the contrast schematic diagram with the network delay using prior art using the present invention.
In the present embodiment, a test environment is built using OpenFlow network simulators mininet, topological structure is such as Fig. 6(a), Fig. 6(b)Shown, the network switch is all based on the OpenFlow interchangers of software, and it uses Openflow to assist View 1.3 editions, the network management of the support elephant stream of this hair embodiment especially weights routing algorithm and operates in NOX controls On device.Meanwhile, realize equal cost multipath on NOX controllers(ECMP)Routing algorithm and single path routing algorithm are used as right According to.
Further, simulation produces the flow rate mode that an elephant stream and mouse stream coexist.Each data flow is by possessing phase With TCP five-tuples(Source IP, source port, purpose IP address, destination interface, protocol type)Sequence of data packet composition.Data flow Destination address be evenly distributed in network.Flow rate mode meets Poisson distribution, obtains the test as shown in Fig. 7, Fig. 8, Fig. 9 As a result.
The transverse axis of Fig. 7 is offered load(Unit is Mbps)The longitudinal axis is time delay(Unit is microsecond), the present embodiment can lead to Crossing increases the ratio that the raw flow of elephant miscarriage accounts for all flows(For example, proportion is 95.6%-99.2%)It is whole to increase Body is loaded.With the increase of offered load, the time delay being route using single path is increased quickly because serious data flow is collided. Conversely, can efficiently control time delay using ECMP and using weighting routing algorithm and just occur in the case where load is very high The decline of performance.When elephant stream ratio is sufficiently high, also begin to increase substantially using the time delay of ECMP, and add and route using weighting The time delay of algorithm is still showed well, better than the time delay obtained using ECMP.
Fig. 8 is the contrast schematic diagram with the network throughput using prior art using the present invention.The transverse axis of Fig. 8 is network Load(Unit is Mbps)The longitudinal axis is handling capacity(Unit is Mbps), it is route using single path and has been rapidly reached bottle using ECMP Neck, because mass data coating is blocked in a network after occurring focus in network.Compared using weighting routing algorithm and used ECMP handling capacities improve about 10%.
Fig. 9 is the contrast schematic diagram with the link utilization using prior art using the present invention.The transverse axis of Fig. 9 is network Load(Unit is Mbps)The longitudinal axis is link utilization.When offered load reaches 100%, the average chain being route using single path Road utilization rate is 70%, the use of the average link utilization of ECMP is 84%, and uses the average link profit of weighting routing algorithm It is 92% with rate.
Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the scope of the present invention.It is all Any modification, equivalent substitution and improvements made within the spirit and principles in the present invention etc., are all contained in protection scope of the present invention It is interior.

Claims (7)

1. it is a kind of support elephant flow network management, it is characterised in that the method includes:
When the elephant in detecting network is flowed, the source address and destination address of elephant stream are extracted, obtain source address and destination Multiple paths between location, the respective path for supporting that elephant is flowed is calculated using the offered load and network topology structure in the path Weight;
NMS generates the forward table of the network switch, and forward table is sent to the network switch, the forward table Weight and the operational order of the network switch including supporting the respective path of elephant stream, the forwarding is performed for the network switch Table, the flow that elephant is flowed is distributed according to the weight in path;
The offered load and network topology structure using the path calculates the weight of the respective path for supporting elephant stream, bag Include:
Acquisition represents the set of all available paths between source address and destination address, and the number for making the path is I;To each Path i (0≤i≤I-1) performs following operation:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1);Make chain The load of road k is λkji(0≤k≤Kji-1);The capacity for making link k is ckji(0≤k≤Kji-1);
Calculate the link utilization of each link:LUkjiIt is the link utilization of link k;
Calculate the weight in path:wi∈ 0,1 ..., and 9 } it is the power of path i Weight.
2. the method for claim 1, it is characterised in that the forward table includes flow table and group table;
The flow table includes destination address and corresponding action fields, and the action fields are respectively directed to a group table;
Described group of table includes multiple action items respectively, action items weight respectively including path and the network switch Operational order.
3. the method for claim 1, it is characterised in that the method also includes:
Generation LLDP messages are simultaneously periodically issued to the network switch;The LLDP messages that the network switch is uploaded are received, by receiving LLDP messages obtain network topology structure;
The statistics of the periodic polling network switch, to obtain the load of the network switch and the load of network link.
4. a kind of NMS, it is characterised in that the NMS includes:
Elephant flow path computing unit, for when the elephant in detecting network is flowed, extracting the source address and purpose of elephant stream Address, obtains the multiple paths between source address and destination address, using the offered load and network topology structure in the path Calculate the weight of the respective path for supporting elephant stream;
Path management unit, the forward table for generating the network switch, and forward table is sent to the network switch, described turn The weight and the operational order of the network switch of the respective path including supporting elephant stream are delivered, is performed for the network switch described Forward table, the flow that elephant is flowed is distributed according to the weight in path;
The elephant flow path computing unit, calculates for the offered load and network topology structure using the path and supports big As the weight of the respective path of stream;
Wherein, the elephant flow path computing unit, all available paths between source address and destination address are represented for obtaining Set, make the path number be I;
The elephant flow path computing unit, is further used for performing each path i (0≤i≤I-1) following operation, in terms of Calculate the weight in path:
The number for making its subpath for including is Ji;The number for making the link that subpath j includes is Kji(0≤j≤Ji-1);Make chain The load of road k is λkji(0≤k≤Kji-1);The capacity for making link k is ckji(0≤k≤Kji-1);
Calculate the link utilization of each link:LUkjiIt is the link utilization of link k;
Calculate the weight in path:wi∈ 0,1 ..., and 9 } it is the power of path i Weight.
5. NMS as claimed in claim 4, it is characterised in that the NMS also includes:
Network topology unit, for generating LLDP messages and being periodically issued to the network switch, and receives on the network switch The LLDP messages of biography, network topology structure is obtained by the LLDP messages for receiving;
Offered load monitoring unit, for the statistics of the periodic polling network switch, to obtain the load of the network switch With the load of network link.
6. a kind of network switch, is connected with the NMS described in claim 4 or 5, including:
Elephant flows reporting unit, and the data flow flowed labeled as elephant for the selection from the data flow for receiving simultaneously will be selected Data flow reports to NMS;
Retransmission unit, for receiving and perform NMS transmission forward table, by elephant flow flow according to forward table In weight distribution.
7. a kind of communication network, including:
NMS as described in any one of claim 4 to 5;And
Signal is connected to the network switch as claimed in claim 6 of the NMS.
CN201310752282.6A 2013-12-31 2013-12-31 Support network management, system, interchanger and the network of elephant stream Active CN103716208B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310752282.6A CN103716208B (en) 2013-12-31 2013-12-31 Support network management, system, interchanger and the network of elephant stream

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310752282.6A CN103716208B (en) 2013-12-31 2013-12-31 Support network management, system, interchanger and the network of elephant stream

Publications (2)

Publication Number Publication Date
CN103716208A CN103716208A (en) 2014-04-09
CN103716208B true CN103716208B (en) 2017-06-30

Family

ID=50408813

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310752282.6A Active CN103716208B (en) 2013-12-31 2013-12-31 Support network management, system, interchanger and the network of elephant stream

Country Status (1)

Country Link
CN (1) CN103716208B (en)

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105282044B (en) * 2014-07-01 2018-08-07 上海宽带技术及应用工程研究中心 Network multipath diameter realization method and system based on OpenFlow agreements
CN104468353A (en) * 2014-12-26 2015-03-25 深圳市新格林耐特通信技术有限公司 SDN based data center network flow management method
CN104767694B (en) * 2015-04-08 2017-10-13 大连理工大学 A kind of stream compression forwarding method towards Fat Tree data center network architectures
US10075364B2 (en) * 2015-05-19 2018-09-11 Alcatel-Lucent Method and apparatus for self-tuned adaptive routing
CN106330730A (en) * 2015-06-30 2017-01-11 中兴通讯股份有限公司 Multicast control method and apparatus
EP3306873B1 (en) 2015-06-30 2019-09-11 Huawei Technologies Co., Ltd. Routing method, and relevant device and system
CN105376178B (en) * 2015-12-07 2019-03-15 北京邮电大学 A kind of method and apparatus of data transmission
CN106209669B (en) * 2016-06-30 2019-06-07 中国人民解放军国防科学技术大学 Towards SDN data center network maximum probability path stream scheduling method and device
CN106301921B (en) * 2016-08-16 2019-05-14 清华大学 Elephant flow transmission dispatching method and system based on tunnel
CN106506395A (en) * 2016-11-28 2017-03-15 迈普通信技术股份有限公司 A kind of business stream scheduling method and device
US10263883B2 (en) 2016-12-14 2019-04-16 International Business Machines Corporation Data flow configuration in hybrid system of silicon and micro-electro-mechanical-switch (MEMS) elements
CN107682258A (en) * 2017-09-27 2018-02-09 北京邮电大学 A kind of multi-path network transmission method and device based on virtualization
CN107592272B (en) * 2017-10-17 2020-10-27 盛科网络(苏州)有限公司 Method and device for load balancing based on source IP address
CN110505148A (en) * 2019-07-24 2019-11-26 烽火通信科技股份有限公司 Realize the method and system of weighting multipath
CN112311674B (en) * 2019-07-31 2022-02-25 北京华为数字技术有限公司 Message sending method, device and storage medium
CN110572332B (en) * 2019-08-30 2022-08-26 西南交通大学 Network equipment message observation data acquisition task dividing method
CN112751781A (en) * 2019-10-31 2021-05-04 阿里巴巴集团控股有限公司 Method, device and equipment for processing flow data and computer storage medium
CN112260950B (en) * 2020-10-19 2022-05-06 桂林电子科技大学 Cloud storage network multicast stream scheduling method based on service priority differentiation
CN116781624A (en) * 2022-03-08 2023-09-19 华为技术有限公司 Method for issuing routing table and related device
CN114884883B (en) * 2022-06-16 2024-01-30 深圳星云智联科技有限公司 Traffic forwarding method, device, equipment and storage medium

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1910876A (en) * 2004-01-23 2007-02-07 西门子公司 Optimisation of traffic distribution in multipath routing
CN102984064A (en) * 2012-12-28 2013-03-20 盛科网络(苏州)有限公司 Method and system for distinguishing and transmitting elephant flow

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1910876A (en) * 2004-01-23 2007-02-07 西门子公司 Optimisation of traffic distribution in multipath routing
CN102984064A (en) * 2012-12-28 2013-03-20 盛科网络(苏州)有限公司 Method and system for distinguishing and transmitting elephant flow

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
加权多路径路由在LEO卫星星座组网中的仿真应用;张进 等;《电讯技术》;20090828(第8期);全文 *

Also Published As

Publication number Publication date
CN103716208A (en) 2014-04-09

Similar Documents

Publication Publication Date Title
CN103716208B (en) Support network management, system, interchanger and the network of elephant stream
Long et al. LABERIO: Dynamic load-balanced routing in OpenFlow-enabled networks
CN102075444B (en) Network system and method for guaranteeing multi-type service quality
Liu et al. SDN based load balancing mechanism for elephant flow in data center networks
CN102263697B (en) Method and device for sharing aggregated link traffic
CN107959633A (en) A kind of load balance method based on price mechanism in industry real-time network
CN103825823B (en) Data forwarding method based on different priorities in software-defined network
CN109314666A (en) Virtual channel endpoint for congestion aware load balancing
CN106357547A (en) Software-defined network congestion control algorithm based on stream segmentation
CN109547340A (en) SDN data center network jamming control method based on heavy-route
CN106209669A (en) Towards SDN data center network maximum of probability path stream scheduling method and device
CN103023804B (en) A kind of aggregated links adaptive flow load-balancing method, device and the network equipment
Trestian et al. OFLoad: An OpenFlow-based dynamic load balancing strategy for datacenter networks
Cui et al. Difs: Distributed flow scheduling for data center networks
CN107579923A (en) The balancing link load method and SDN controllers of a kind of SDN
CN103152284B (en) A kind of router multipath exports method and the router of intelligent load equilibrium
CN103916319B (en) Link selecting method and stack equipment in LACP stacking networkings
CN104486809A (en) Wireless local area network routing method
Cui et al. Difs: Distributed flow scheduling for adaptive routing in hierarchical data center networks
CN106713182A (en) Method and device for processing flow table
CN108337179A (en) Link flow control method and device
CN102739518A (en) Flow load sharing method and equipment
CN110351286A (en) Link flood attack detects response mechanism in a kind of software defined network
CN101789949B (en) Method and router equipment for realizing load sharing
EP4149066A1 (en) Communication method and apparatus

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant