CN108683602A - Data center network load balancing method - Google Patents
Data center network load balancing method Download PDFInfo
- Publication number
- CN108683602A CN108683602A CN201810772379.6A CN201810772379A CN108683602A CN 108683602 A CN108683602 A CN 108683602A CN 201810772379 A CN201810772379 A CN 201810772379A CN 108683602 A CN108683602 A CN 108683602A
- Authority
- CN
- China
- Prior art keywords
- path
- congestion
- stream
- load
- state table
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/12—Avoiding congestion; Recovering from congestion
- H04L47/125—Avoiding congestion; Recovering from congestion by balancing the load, e.g. traffic engineering
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides a method for balancing network load of a data center, which evaluates path states (mainly path congestion and faults) through comprehensive perception and performs targeted load balancing based on the path states, namely, the path congestion and fault information obtained through perception is utilized, and timely and cautious routing/rerouting is performed by taking a packet as the minimum switchable granularity under the guidance of the path congestion and the fault information and the flow state information, so that the method can be used for solving the challenges brought to the load balancing by uncertain factors such as flow dynamic transmission, topological asymmetry, equipment faults and the like in the network of the data center. The invention also provides a load balancing system deployed at the host terminal of the data center and based on the method, and a data center network system composed of the host terminals deployed with the system.
Description
Technical field
The present invention relates to the flow load balance technical fields in data center network;In particular it relates in a kind of data
The method of heart Network Load Balance and the SiteServer LBS based on this method and a kind of data centre network system.
Background technology
Data center network (Datacenter network) refers to the network being applied in data center;Data center
Network is connected a certain number of servers by network equipments such as interchanger, routers, and composition is with high bandwidth, Gao Ke
By the server network of property and load balancing, the services such as calculating and storage can be externally provided.
To meet the needs of extensive, high scalability, high robustness, data center network generally use Clos networks,
It is a kind of more trees (Multi-rooted trees), such as fat tree topology structure (Fat tree), leaf ridge topological structure
(Leaf-spine).As shown in Figure 1, above-mentioned structure can be more per a pair of of host terminal (End host) offer in network
Paths;Therefore, it in above-mentioned network, realizes stringent load balancing, is extremely necessary.
However, in fact, in data center network transmission process, there is miscellaneous uncertain factors, for example flow
Measure dynamic change (Traffic dynamics), topology asymmetric (Topology asymmetry) and equipment fault
(Failures) etc..Wherein, in data center network transmission process, flow is dynamic, therefore congestion possibly is present at net
From anywhere in network (Fig. 2);Secondly, chain caused by the introducing of heterogeneous device (Heterogenous devices) and failure
Road disconnects (Link cut), will produce topological asymmetric (Fig. 3);In addition, exchange fault also can be through in data center network
(Fig. 4) often occurs, typical exchange fault includes data forwarding black hole (Packet blackholes) and randomness packet loss
(Random packet drops, Fig. 5).
ECMP is a kind of typical load balancing scheme in data center network.It is achieved in that, by each difference
Stream (Flow) equiprobably select different equative routes to be route according to the difference of five-tuple Hash.What ECMP was used
It is the equivalent route strategy of randomness, under the situations such as a part of path experiences congestion, asymmetric topological network, the routing
Strategy just seems blindly inefficient.
Therefore, it is necessary to design a kind of effective and appropriate flow load balance scheme, to meet data center network
Demand.Under the influence of above-mentioned many uncertain factors, ideal load balancing scheme should effectively perceive above-mentioned do not know
Factor perceives congestion, that is, failure on path, and according to perceiving as a result, reply is made, as much as possible by flow load
It is properly assigned to parallel path, with load balancing.
The ECMP addressed as it before in some existing data center network load balancing schemes and other schemes are such as
RPS, DRB, Presto have ignored perception path congestion, and the performance under dissymmetric network is also weak;CONGA、
HULA, DRILL are then to realize perception path congestion under the support of specific hardware (interchanger with correlation function);And
Only it is the CLOVE-ECN that perception path congestion function is realized at Intrusion Detection based on host end, the path visibility (Visibility) obtained
It is limited.Worse situation is that most of existing flow equalization schemes cannot perceive path failure, therefore it is black to solve forwarding
Hole and randomness packet loss are not just known where to begin yet.
In addition, existing scheme provides the mode of different reply load balancing, such as flow point section heavy-route (Flowlet
Switching), strong heavy-route (Vigorous rerouting) etc..Wherein, flow point section heavy-route is only when flow point section occurs
Ability heavy-route (as shown in Fig. 6 a, 6b, 6c), it is clear that cannot (Timely) reply in time;And strong heavy-route can then aggravate to wrap it is out of order
(Packet reordering), and cause congestion mispairing (Congestion mismatch).And CN104767826A is also carried
A kind of load-balancing method of the foundation supplied under the path perception guidance based on end, after perception path congestion, failure, accordingly
Guidance, adaptive load balancing.However, objectively saying, the precise degrees of its path perception are not discussed for the time being, are only adaptive negative
Equilibrium is carried, is substantially exactly the regulation and control to message transmission speed, this is a kind of passive reply, is only avoided that and is further exacerbated by
The load pressure of congestion path, and can not fundamentally solve the problems, such as that the congestion occurred, failure cause.
Invention content
In view of this, the present invention provides a kind of method of data center network load balancing, to cope with data center network
The challenge that the uncertain factors such as middle flow dynamics transmission, topology asymmetry and equipment fault are brought to load balancing.In addition, this hair
It is bright also provide it is a kind of being deployed in data center's host terminal, SiteServer LBS based on this method, and dispose the system
Host terminal composition data centre network system.
On the one hand, the present invention provides a kind of method of data center network load balancing.This method passes through at source host end
Safeguard that stream state table and path-state table, the stream state table have recorded the flow state information sent out from the source host, the road
Diameter state table has recorded the Congestion Level SPCC and fault condition from the reachable all paths of the source host;This method includes satisfying the need
The comprehensive perception (Comprehensive Sensing) of diameter state and targetedly load balancing:
The synthesis, which perceives, includes:By (including Congestion Level SPCC and event here to path status by transport layer information
Barrier) it is assessed, it will be in the path congestion of acquisition, fault message update to the path-state table;The transport layer letter
Breath, including transmission layer signal (such as RTT, ECN, ack signal), event are (as retransmitted (Retransmission), time-out
(Timeout) etc.).
The specific aim load balancing includes:Utilize the path congestion in the path-state table and stream state table, event
Hinder information and flow state information, is minimum adjustable granularity (the minimum switchable with packet (Packet)
Granularity routing (the Routing)/heavy-route (Rerouting) of timely and careful (Cautious)) is executed.
In conjunction with described in a first aspect, in its method,
Preferably, described to be commented in conjunction with the two including the use of RTT and ECN signals with transport layer information assessment path status
Estimate path congestion level;It is by setting RTT and ECN thresholdings, RTT, ECN value under collected each path is right therewith respectively
Than to assess the Congestion Level SPCC in the path.
Further, it is preferred that be accurate judgement path congestion level, high and low two RTT threshold values set, by path root
According to producing level be divided into congestion path (Congested path, that is, be in congestion state), good path (Good path, as
Underuse state) and grey path (Gray path, i.e., the state between said two devices, including the road that moderately loads
Diameter).
Preferably, described that path status is assessed with transport layer information, including the use of including transport layer event information (such as weight
Biography, time out event etc.) including transport layer information assess path failure;Wherein, a kind of to be preferably achieved in that, pass through setting
Time out event number threshold value, retransmission events frequency threshold under single-pathway will occur in the expired times in each path, retransmit
Value compares one by one therewith, and combines the Ack signals and its Congestion Level SPCC in each path, evaluates the fault condition in each path.
Preferably, the comprehensive perception further includes periodic, low overhead active probe (Active
Probing), to improve the visibility in path.
Preferably, it is described using path status, the guidance of stream transmission state to wrap as the timely of minimum adjustable granularity and
Careful routing/heavy-route, including:
When packet is sent out from source host end, the corresponding stream of the packet is detected, (i.e. the packet is this if the packet belongs to a new stream
First data packet when stream is sent), then according to the path-state table, selection loads minimum fault-free path and sends;
If the packet belongs to the stream of an experience time-out, path congestion or path failure, gathered around according to the path in the path-state table
The status information of the stream in plug, fault message and the stream state table, assesses the heavy-route profit and loss, is striving for maximum return
Under strategy, decide whether the path of heavy-route and heavy-route.
Further, it is preferred that the stream send status information, including the transmission speed of stream, non-transmitting portion size with
And whether stream undergoes time-out;Wherein, the mode of non-transmitting portion size is flowed in a kind of common assessment, according to stream sending part
The size divided assesses the size of its non-transmitting portion;In addition, whether the stream undergoes time-out, the stream is belonged to generally according to receiving
Two neighboring packet Ack signals time interval whether be more than prefixed time interval judge.
On the other hand, in conjunction with the load-balancing method of the first aspect, the present invention provides a kind of based on this method, portion
Affix one's name to the SiteServer LBS of the host terminal in data center network.The system includes:
Path sensing module (Sensing Module) and routing module (Routing Module);Wherein,
The path sensing module, for comprehensive perception path status (including Congestion Level SPCC and failure here);It is described
Comprehensive perception, i.e., as described in the comprehensive perception in the first aspect method;The path congestion of the comprehensive perception acquisition, event
Barrier information update is to being equally maintained in the path-state table at source host end;
The routing module, for executing to wrap as the timely and careful routing/heavy-route of minimum adjustable granularity;Its
Using the path congestion in the path-state table, fault message, in conjunction with the stream in the stream state table that source host end is safeguarded
Status information instructs the timely and careful routing/heavy-route;Utilization path congestion, fault message and the stream shape
State information is to wrap for the timely and careful routing/heavy-route of minimum adjustable granularity, i.e., as in the method for the first aspect
Using path congestion, fault message and flow state information to wrap as the timely and careful routing/heavy-route of minimum adjustable granularity
It is described.
In conjunction with the second aspect, in its described system, it is preferred that the path sensing module includes actively visiting
Submodule is surveyed, for executing the active probe in the comprehensive perception.
Another aspect, in conjunction with the second aspect, the present invention provides a kind of data centre network system, data center's net
Road system includes the server host terminal for the SiteServer LBS that several deployment second aspects are addressed.
Compared with the prior art, technical solution provided by the invention has many advantageous effects, such as:
One, it is comprehensive, path congestion and failure are effectively detected by comprehensive perception, and instruct load equal accordingly
Weighing apparatus strategy;
Two, promptness is established to wrap as routing/heavy-route of minimum adjustable granularity, can cope with data rapidly
Uncertain factor in central site network, and react in time;
Three, friendly is transmitted, same its is changed with packet for minimum adjustable granularity, by careful routing/heavy-route strategy,
It limits caused by the problems such as wrapping out of order and congestion mispairing transmits flow and adversely affects well;
Four, deployable can be realized in the commercial hardware in current data center environment, without specific hardware branch
Hold (such as CONGA) or modification network system bottom code.
Description of the drawings
It, below will be in the present invention one clearly to illustrate the embodiment of the present invention or technical solution in the prior art
Attached drawing involved in section Example or description of the prior art does simple introduction.
Fig. 1 illustrates a kind of data center network architecture in the prior art (leaf ridge topological structure) schematic diagram;
Fig. 2 discloses the defect that congestion easily occurs for the data center network in Fig. 1;
Fig. 3 discloses the origin cause of formation of the asymmetric this uncertain factor of topology in the data center network in Fig. 1;
Fig. 4 illustrates the data center network under exchange fault;
Fig. 5 illustrates the problems such as random packet loss existing in the prior art under exchange fault;
Fig. 6 a illustrate point-to-point transmission there are multiple streams while the case where when sending, and Fig. 6 b show weight ideally
Routing, and Fig. 6 c show the heavy-route under flow point section adjustment technology means;
Fig. 7 illustrates the network traffic load equilibrium system of one embodiment of the invention offer being deployed on host terminal
System;
Fig. 8 displayings are a kind of preferred implementation that Fig. 7 provides system, the i.e. network flow with active probe function
SiteServer LBS.
Specific implementation mode
With reference to the attached drawing of the embodiment of the present invention, technical solution in the embodiment of the present invention carries out clearly and completely
Description.Obviously, described embodiment is only the embodiment of a part of the invention, instead of all the embodiments.Based on this hair
Embodiment in bright, the every other implementation that those of ordinary skill in the art are obtained under the premise of without creative work
Example, shall fall within the protection scope of the present invention.
Fig. 1 is a kind of Clos network diagrams of leaf ridge topological structure, and this network structure can be each in network
Mulitpath is provided to host terminal.Therefore, using this network transmitting data in parallel, accomplish that stringent flow load balance is
It is extremely necessary.However, when data center runs, just not saying inevitably other interchanger, router etc. are set in its network system
It is standby to break down, it is singly in data transmission procedure under multipath situation, flow congestion caused by routing algorithm can not just be kept away at all
Exempt from.
The embodiment of the present invention is applied in above-mentioned data center network load balancing field, realizes data center network as far as possible
The high efficiency of transmission of information flow-rate.
A part of the embodiment of the present invention provides a kind of method of data center network load balancing, and this method passes through in source
Host side safeguards that stream state table and path-state table, the stream state table have recorded the flow state information sent out from the source host,
The path-state table has recorded the Congestion Level SPCC and fault condition from the reachable all paths of the source host;Wherein, it flows
The information of state table and path-state table can be set as needed life cycle, each flow state information wherein in stream state table
Start to be sent to it and be sent completely as the period;And the congestion information of each path decays at any time in path-state table, i.e., one
The congestion information of denier paths does not update more than preset time (being set according to actual conditions), and Congestion Level SPCC then declines gradually
Become;
As the key of the technical program, this method includes being perceived with targetedly load to the comprehensive of path status
Weighing apparatus:
The synthesis, which perceives, includes:Path status (i.e. Congestion Level SPCC and failure) is commented by transport layer information
Estimate, the transport layer information in the path congestion of acquisition, fault message update to the path-state table here is generally included to pass
Defeated layer signal (such as RTT, ECN, ack signal), event (such as retransmitting, time-out).
The specific aim load balancing includes:Utilize the path congestion in the path-state table and stream state table, event
Hinder information and flow state information, timely and careful routing/heavy-route is executed for minimum adjustable granularity with packet.
In the above-described embodiment, section Example provides a kind of preferred realization method, i.e., is commented with transport layer information
When estimating path status, assessment path congestion level is combined including the use of both RTT and ECN signals.In some related art schemes
In, RTT is usually brought directly assessment path congestion level end to end.But although RTT has information value
(Informative), but in the case of no advanced NIC hardware supporteds, RTT when accurately measuring be difficult to realize.Cause
This, a high RTT value not necessarily indicates path congestion (for example the delay of host terminal network stack can also increase RTT values),
But small RTT values can then explicitly indicate that path is underused, this is more to have information value for us.ECN usually quilts
For capturing the congestion of single-hop (hop), it supports to realize by commercial interchanger, is widely used in congestion avoidance algorithm.But
ECN signals can only reflect the congestion situation that most congestion one is jumped in path.And under high-load network, congestion may be accumulated from multi-hop
Get up.Therefore, a low ECN value is also not necessarily indicated that congestion does not occur, especially the ECN undersamplings the case where
Under, it is particularly evident.
Further, some in above-described embodiment also provide a kind of more preferably realization method, that is, set high and low two
RTT threshold values, as shown in the table, its producing level, is divided into congestion path, grey path, good path by inference:
In any of the above-described embodiment, section Example provides a kind of preferred realization method, i.e., believes with transport layer
It when breath assessment path status, is retransmitted including the use of transport layer event information, time-out and transmission layer signal Ack assess path failure;
By setting time out event number threshold value, retransmission events frequency threshold under single-pathway, the time-out in each path will occur in
Number, re-transmission value compare one by one therewith, meet preset condition (time-out time number, retransmission events frequency be more than threshold value and
To receive Ack packets) and exclude path congestion when, then it is assumed that the path failure.
In any of the above-described embodiment, section Example provides a kind of preferred realization method, i.e., described comprehensive sense
Know to further include periodic, low overhead active probe.Detection in some existing schemes uses complete trails detection, but true
On, we need not move through detection All Paths to grasp its Congestion Level SPCC, but an only small portion of the random probing in them
Point, the path for improving load balancing is similarly helped in this way, and expense can be greatly decreased.Therefore, when active probe, can to
Several (such as two) paths of machine selection are detected.In addition, in the above-described embodiments, in order to improve detection stability and
Increase the possibility for finding and not utilizing path, can also extraly detect the shape observed before this while random routing detects
The best path of state.
In any of the above-described embodiment, section Example provides a kind of preferred realization method, i.e., described utilizes road
Diameter state, the guidance of stream transmission state are to wrap as timely and careful routing/heavy-route of minimum adjustable granularity, detailed process
It is as follows:
When data packet is sent out from source host end, the corresponding stream of the packet is detected;
If the packet belongs to a new stream (i.e. the packet is first data packet when the stream is sent), according to the road
Diameter state table, selection load minimum fault-free path and send;
If the packet belongs to the stream of an experience time-out, path congestion or path failure, according in the path-state table
Path congestion, in fault message and the stream state table stream status information, assess the heavy-route profit and loss, striving for most
Under the strategy of big income, decide whether the path of heavy-route and heavy-route.
Herein, it is worth noting that, in reply data center network congestion existing and disclosed by the invention, failure etc.
It is proposed by the present invention that granularity can adjust to be minimum with packet in probabilistic scheme, it is adjusted relative to flow point section, it is clear that be a kind of
More timely with positive counte-rplan;And when congestion occurs, establish the careful heavy-route (Cautious on Profit Assessment
Rerouting), compared with strong heavy-route, also there is unrivaled advantage.This is because when congestion occurs for stream, it is hastily heavy
Routing, be adjusted to mistake path (such as more congestion or failure path on), cannot not only solve the problems, such as, cause out instead
Pin increases and income reduces, and is more likely to that problem is caused further to deteriorate.Even grasping path status (such as CN104767826A
The appraisal procedure of offer), it is not the highest selection of income, such as a stream residue entirely to be re-routed to the lower path of load also
When part is smaller, no matter how wise heavy-route, income is also likely to be negative.This phenomenon is to smaller stream heavy-route
When it is particularly evident.
Further, some in above-described embodiment also provide a kind of more preferably realization method, i.e. the stream is sent
Status information, including the transmission speed of stream, non-transmitting portion size and the overtime transmission speed thresholding of stream;Wherein, a kind of common
Assessment flow the mode of non-transmitting portion size, according to the size of stream transmitting portion assess the big of its non-transmitting portion
It is small;It therefore, can the roads Bu Zuochong for smaller stream, the smaller stream of transmitting portion or transmission speed are more than the stream of preset value
By handling.In addition, whether the stream undergoes time-out, generally according to receive the two neighboring packet Ack signals for belonging to the stream when
Between be spaced whether be more than prefixed time interval judge.
Fig. 7 provides a kind of data center network SiteServer LBS for another part embodiment of the present invention.As shown,
The system includes:
Path sensing module and routing module;Wherein,
The path sensing module, for comprehensive perception path status (including Congestion Level SPCC and failure here);It is described
Comprehensive perception, i.e., as described in the comprehensive perception in the first aspect method, major function is perceiving congestion and perception event
Barrier;The path congestion of the comprehensive perception acquisition, fault message are inserted into the path-state table for being equally maintained in source host end;
The routing module, for executing to wrap as the timely and careful routing/heavy-route of minimum adjustable granularity;Its
Using the path congestion in the path-state table, fault message, in conjunction with the stream in the stream state table that source host end is safeguarded
Status information instructs the timely and careful routing/heavy-route;Utilization path congestion, fault message and the stream shape
State information is to wrap for the timely and careful routing/heavy-route of minimum adjustable granularity, i.e., as in the method for the first aspect
Using path congestion, fault message and flow state information to wrap as the timely and careful routing/heavy-route of minimum adjustable granularity
It is described.
Fig. 8 is some preferred implementations provided in above-described embodiment, includes actively in the sensing module of path
Detection, for executing active probe.
In addition, the another section Example of the present invention provides a kind of data centre network system, the data center network system
System includes the server host terminal for the SiteServer LBS that several any of the above-described embodiments of deployment are addressed.
The foregoing is merely the specific implementation modes of the present invention, but scope of protection of the present invention is not limited thereto.
Claims (10)
1. a kind of data center network load-balancing method, which is characterized in that
Safeguard that stream state table and path-state table, the stream state table record the stream mode sent out from the source host at source host end
Information, Congestion Level SPCC and fault condition of the path-state table record from the reachable all paths of the source host;
The method includes being perceived and targetedly load balancing to the comprehensive of path status;
The comprehensive perception includes assessing path status by transport layer information, the state includes path congestion, failure;
By the path congestion of acquisition, fault message update to in the path-state table;The transport layer information includes transport layer
Signal, time;
The specific aim load balancing includes being believed using path congestion, the failure in the path-state table and stream state table
Breath and flow state information execute timely and careful routing/heavy-route with packet for minimum adjustable granularity.
2. load-balancing method according to claim 1, which is characterized in that
Described assesses path status with transport layer information, including is combined RTT and ECN signal evaluation path congestion levels;
By setting RTT and ECN thresholdings, respectively in contrast by RTT, ECN value in each path, to assess the Congestion Level SPCC in the path.
3. load-balancing method according to claim 2, which is characterized in that
High and low RTT thresholdings are set, path is divided into congestion path, good path and grey path.
4. load-balancing method according to claim 1, which is characterized in that
Described assesses path status with transport layer information, and path failure is assessed including the use of transport layer information;The transmission
Layer information includes transport layer event information;The event includes re-transmission, time out event.
5. load-balancing method according to claim 1, which is characterized in that
The described comprehensive perception further includes periodic, low opening small active probe;The active probe, including select at random
Several ways diameter is selected to be detected.
6. load-balancing method according to claim 5, which is characterized in that
The active probe, further include additionally detected while randomly choosing path detection observe before this it is in the best state
Path.
7. load-balancing method according to claim 1, which is characterized in that
It is described using path status, the guidance of stream transmission state with wrap the timely and careful routing for minimum adjustable granularity/
Heavy-route, including:When packet is sent out from source host end, the corresponding stream of the packet is detected, if the packet belongs to a new stream, basis
The path-state table, selection load minimum fault-free path and send;If the packet belongs to experience time-out, a path congestion
Or the stream of path failure is somebody's turn to do then according to the path congestion in the path-state table, fault message in conjunction in the stream state table
The status information of stream assesses the heavy-route profit and loss, decides whether the path of heavy-route and heavy-route.
8. load-balancing method according to claim 7, which is characterized in that
Whether the flow state information, including the transmission speed of stream, non-transmitting portion size and stream undergo time-out.
9. a kind of data center network SiteServer LBS, which is characterized in that
The system includes path sensing module and routing module;Wherein,
The path sensing module, for comprehensive perception path status, the state includes path congestion, failure, described
Comprehensive perception is the comprehensive perception in the load-balancing method as described in any in claim 1-8;What the comprehensive perception obtained
Path congestion, fault message are inserted into the path-state table for being maintained in source host end;
The routing module, for executing to wrap as the timely and careful routing/heavy-route of minimum adjustable granularity;The road
By module using the path congestion in the path-state table, fault message, in conjunction with being maintained in the stream state table at source host end
Flow state information, instruct the timely and careful routing/heavy-route;The utilization path congestion, fault message and
Flow state information is to wrap for the timely and careful routing/heavy-route of minimum adjustable granularity, i.e., such as any institute in claim 1-8
State utilization path congestion, fault message and flow state information in load-balancing method with wrap as the adjustable granularity of minimum in time and
Careful routing/heavy-route.
10. a kind of data centre network system, which is characterized in that
Data center's network system, including the server host of SiteServer LBS described in several deployment claims 9 are whole
End.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810772379.6A CN108683602B (en) | 2018-07-13 | 2018-07-13 | Data center network load balancing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810772379.6A CN108683602B (en) | 2018-07-13 | 2018-07-13 | Data center network load balancing method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108683602A true CN108683602A (en) | 2018-10-19 |
CN108683602B CN108683602B (en) | 2022-05-13 |
Family
ID=63813956
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810772379.6A Active CN108683602B (en) | 2018-07-13 | 2018-07-13 | Data center network load balancing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108683602B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110351187A (en) * | 2019-08-02 | 2019-10-18 | 中南大学 | Data center network Road diameter switches the adaptive load-balancing method of granularity |
CN112448982A (en) * | 2019-08-31 | 2021-03-05 | 深圳致星科技有限公司 | Method and system for sharing network resources among multi-task training cluster tasks |
CN113364689A (en) * | 2020-03-06 | 2021-09-07 | 瞻博网络公司 | Detecting faulty wiring in a leaf-spine topology of a plurality of network devices |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050147031A1 (en) * | 2004-01-07 | 2005-07-07 | Lucent Technologies Inc. | Method for fast network re-optimization |
CN101494614A (en) * | 2009-03-09 | 2009-07-29 | 华为技术有限公司 | Method, apparatus and system for rapid rerouting |
CN102088413A (en) * | 2011-03-02 | 2011-06-08 | 华为技术有限公司 | Network flow shunting method, network node and network system |
CN104767826A (en) * | 2015-04-23 | 2015-07-08 | 南京大学 | Distributed data center load balancing method based on fault tolerance of side |
CN105634823A (en) * | 2016-01-18 | 2016-06-01 | 河南科技大学 | Method for recovering fault of data center network based on multi-routing configuration |
CN107959633A (en) * | 2017-11-18 | 2018-04-24 | 浙江工商大学 | A kind of load balance method based on price mechanism in industry real-time network |
CN108199970A (en) * | 2017-12-15 | 2018-06-22 | 浙江大学 | Packet path reconstructing method in a kind of software defined network |
CN108243111A (en) * | 2016-12-27 | 2018-07-03 | 华为技术有限公司 | The method and apparatus for determining transmission path |
-
2018
- 2018-07-13 CN CN201810772379.6A patent/CN108683602B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050147031A1 (en) * | 2004-01-07 | 2005-07-07 | Lucent Technologies Inc. | Method for fast network re-optimization |
CN101494614A (en) * | 2009-03-09 | 2009-07-29 | 华为技术有限公司 | Method, apparatus and system for rapid rerouting |
CN102088413A (en) * | 2011-03-02 | 2011-06-08 | 华为技术有限公司 | Network flow shunting method, network node and network system |
CN104767826A (en) * | 2015-04-23 | 2015-07-08 | 南京大学 | Distributed data center load balancing method based on fault tolerance of side |
CN105634823A (en) * | 2016-01-18 | 2016-06-01 | 河南科技大学 | Method for recovering fault of data center network based on multi-routing configuration |
CN108243111A (en) * | 2016-12-27 | 2018-07-03 | 华为技术有限公司 | The method and apparatus for determining transmission path |
CN107959633A (en) * | 2017-11-18 | 2018-04-24 | 浙江工商大学 | A kind of load balance method based on price mechanism in industry real-time network |
CN108199970A (en) * | 2017-12-15 | 2018-06-22 | 浙江大学 | Packet path reconstructing method in a kind of software defined network |
Non-Patent Citations (1)
Title |
---|
HONG ZHANG等: "Resilient Datacenter Load Balancing in the Wild", 《THE CONFERENCE OF THE ACM SPECIAL INTEREST GROUP ON DATA COMMUNICATION》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110351187A (en) * | 2019-08-02 | 2019-10-18 | 中南大学 | Data center network Road diameter switches the adaptive load-balancing method of granularity |
CN110351187B (en) * | 2019-08-02 | 2021-03-12 | 中南大学 | Load balancing method for path switching granularity self-adaption in data center network |
CN112448982A (en) * | 2019-08-31 | 2021-03-05 | 深圳致星科技有限公司 | Method and system for sharing network resources among multi-task training cluster tasks |
CN113364689A (en) * | 2020-03-06 | 2021-09-07 | 瞻博网络公司 | Detecting faulty wiring in a leaf-spine topology of a plurality of network devices |
US11716276B2 (en) | 2020-03-06 | 2023-08-01 | Juniper Networks, Inc. | Detecting miswirings in a spine and leaf topology of network devices |
Also Published As
Publication number | Publication date |
---|---|
CN108683602B (en) | 2022-05-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Refaei et al. | A reputation-based mechanism for isolating selfish nodes in ad hoc networks | |
Andersen et al. | Resilient overlay networks | |
Yi et al. | On the role of routing in named data networking | |
Zhang et al. | A Transport Layer Approach for Improving End-to-End Performance and Robustness Using Redundant Paths. | |
US7336613B2 (en) | Method and apparatus for the assessment and optimization of network traffic | |
CA2424680C (en) | Method and apparatus for the assessment and optimization of network traffic | |
CN102123088B (en) | Set up the method and apparatus of traffic engineering tunnel | |
US7773536B2 (en) | Method and apparatus for the assessment and optimization of network traffic | |
Kanagevlu et al. | SDN controlled local re-routing to reduce congestion in cloud data center | |
CN101159669A (en) | Service flow switching method and apparatus | |
CN108683602A (en) | Data center network load balancing method | |
Jen et al. | Towards a New Internet Routing Architecture: Arguments for Separating Edges from Transit Core. | |
Mérindol et al. | Improving load balancing with multipath routing | |
Kanagavelu et al. | OpenFlow based control for re-routing with differentiated flows in Data Center Networks | |
Wang et al. | A study of Internet packet reordering | |
Kvalbein et al. | Multipath load-adaptive routing: putting the emphasis on robustness and simplicity | |
Mandalari et al. | Measuring ECN++: good news for++, bad news for ECN over mobile | |
Hansen et al. | Fast proactive recovery from concurrent failures | |
Deng et al. | Comparison of RIP, OSPF and EIGRP Routing Protocols based on OPNET | |
Keralapura et al. | Can coexisting overlays inadvertently step on each other? | |
Proto et al. | A framework for packet-droppers mitigation in OLSR wireless community networks | |
Karaoglu et al. | On the scalability of path exploration using opportunistic path-vector routing | |
CN109981407B (en) | Anti-attack message method and device | |
Fujinoki | Multi-path BGP (MBGP): A solution for improving network bandwidth utilization and defense against link failures in inter-domain routing | |
Tanaka | Effects of length and number of paths on simultaneous multi-path communication |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
CB02 | Change of applicant information |
Address after: 518000 a2405, building 9, zone 2, Shenzhen Bay science and technology ecological park, 3609 Baishi Road, community, Yuehai street, Nanshan District, Shenzhen City, Guangdong Province Applicant after: Shenzhen Zhixing Technology Co., Ltd Address before: 100088 Haidian District 57 North Central Ring Road 44, 57 cottage No. 63. Applicant before: BEIJING HANHAI CLUSTAR TECHNOLOGY Co.,Ltd. |
|
CB02 | Change of applicant information | ||
GR01 | Patent grant | ||
GR01 | Patent grant |