CN100388183C - Server load equalization method for implementing weighted minimum linked allocation - Google Patents
Server load equalization method for implementing weighted minimum linked allocation Download PDFInfo
- Publication number
- CN100388183C CN100388183C CNB031473083A CN03147308A CN100388183C CN 100388183 C CN100388183 C CN 100388183C CN B031473083 A CNB031473083 A CN B031473083A CN 03147308 A CN03147308 A CN 03147308A CN 100388183 C CN100388183 C CN 100388183C
- Authority
- CN
- China
- Prior art keywords
- server
- item
- weighted value
- mantissa
- state
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Information Transfer Between Computers (AREA)
Abstract
The present invention discloses a server load dispatching method for equalizing weighted minimum connected number. The present invention has the following technical scheme: the state representing the current connection information of each server in a server group is determined beforehand; the state is represented through a data structure which comprises at least one data item; the data item comprises data items which represent the load condition of current services of each server; the load condition of the current services is approximately represented through first scheduled data items of at least one data item of the state; when the load is regularly dispatched, all the first scheduled data items in the status which approximately represent the load status of the current service are traversed; the server corresponding to the minimum first scheduled data item is selected to access; meanwhile, the data item is modified. By using the method of the present invention, operation amount of load equalizing dispatching can be reduced, and dispatching speed can be accelerated. The present invention has simple method and fast dispatching speed, and is particularly suitable for fixed point processors of message switching equipment in a network.
Description
Technical field
The present invention relates to the load-balancing technique of multiserver, be specifically related to a kind of minimum server load balancing method of distributing that connects of weighting that realizes.
Background technology
Along with developing rapidly of Internet technology, the Internet user sharply increases.Simultaneously the ICP satisfies the higher level requirement of Internet user to internet content and form, utilizes various technology and means, and multimedia technology especially makes the easier attraction of the visual effect user of website.The increase of the increase of user capture quantity and user's single reference flow is had higher requirement to the ability that server bears concurrent visit.Because the load capacity of separate unit server is always limited, in the face of the user capture quantity and the flowing of access that increase day by day, central processing unit of separate unit server (CPU) and I/O (I/O) can become bottleneck very soon.And this problem is not just soluble by the performance that improves server hardware merely.Common solution is to adopt the group of server that is made of multiple servers and group of server is implemented load-balancing technique.In load-balancing technique, relatively more crucial technology is exactly to dispatch the server implementation in the group of server according to predetermined strategy by different load balancing schemes.In many tactful dispatching techniques, the minimum linking number dispatching method of weighting is modal method.The minimum linking number scheduling of weighting can be by performance (weighted value by this server embodies) and current transmission control protocol (TCP) linking number according to server, as calculate the ratio of TCP linking number and weighted value, select the method for best server to realize.But in fixed-point processor, division arithmetic generally will realize that with the plus-minus instruction its maximum shortcoming is exactly that speed is slow by software approach.And a large amount of message processors is a fixed-point processor in the network, and it is a lot of to realize that by fixed point division load balancing calculating can cause performance to descend, so realize that on fixed-point processor the minimum scheduling scheme that connects of weighting is crucial.
For addressing the above problem, at present common a kind of method is to utilize the timed task of backstage timer to calculate the lightest server of load in the group of server as the server that is scheduled, and the scheduler task on foreground is directly distributed to this server with new access request and handled.Specifically, in a timer refresh time interval, the current TCP linking number of each server in the timed task statistical server group, and according to the weighted value of each server, do the fixed point division computing, calculate the TCP linking number of each server and the ratio of weighted value, the server of choosing the ratio minimum by comparison operation is the lightest server of load then, as the server that is scheduled in this group of server.And the scheduler task on foreground is directly used this result of calculation, promptly in timer refresh time interval, this group of server selects this server as the server that is scheduled all the time, processing is assigned to all new access requests of this group of server in this timer refresh time interval, comprise from establishing a connection to disconnecting all tasks that connect.This method is only calculated the once connection state of each server in each timer refresh time interval, rather than all calculates when increasing newly or disconnect each connection, so the result of calculation of group of server can not in time refresh.Under the situation that the timer refresh time is big at interval or the user is more, flow is bigger, may cause the load of at interval interior certain station server of a timer refresh time sharply to increase, even surpass this server load capacity and still by this group of server as the server that is scheduled, thereby cause a large amount of new access requests to wait as long for even be dropped because of overtime.Though this problem can improve at interval by reducing the timer refresh time, but the speed of carrying out division arithmetic owing to fixed-point processor is slower, poor-performing, dwindling of timer refresh time interval is limited, thereby can not tackle the problem at its root.
Summary of the invention
The objective of the invention is to overcome the above-mentioned shortcoming of prior art, a kind of load equilibration scheduling method that has improved is provided.Utilize the present invention, can not carry out above-mentioned division arithmetic, and carry out before partial arithmetic is distributed to balance dispatching, and make speeding up of load balance scheduling, the distribution of load simultaneously is also comparatively balanced.
The invention provides a kind of realization weighting minimum joint number dispatch server load balancing method, may further comprise the steps:
Pre-determine the state of the current link information of each server in the expression group of server, described state is represented by the data structure that comprises at least one data item, described data item comprises the data item of the load state of the current service of representing described each server, and the load state of described current service is by first several the expressions approx in described at least one data item;
When new service inserts,
In the described state of the described load state of representing current service approx of described each server of traversal first several;
More described head number,
If there are minimum first several, select first several the corresponding server of this minimum to insert;
If there is no first several of minimum, select arbitrary server to insert;
Revise at least one data item of the state of the current link information of server of representing described selection.
Preferably, this method also comprises the step that pre-determines the weighted value of each server in the described group of server.
Alternatively, this step may further comprise the steps:
According to the performance of described each server, determine the original weighted value of described each server;
Calculate the weighted value of described each server according to following formula:
Weighted value=INT (original weighted value/dilution of precision).
Wherein, employed weighted value when weighted value is represented the load dispatch computing, INT represents rounding operation, original weighted value is the original weighted value of above-mentioned each server, dilution of precision is according to being not more than the maximum original weighted value of described each server, the remainder that obtains after described original weighted value is calculated is as far as possible little, and the less principle of employed weighted value numerical value is selected when making described load dispatch computing.
Preferably, the step that in the step of state of the described current link information that pre-determines described each server of expression, also comprises mantissa's item of determining the described current link information of expression, wherein, described mantissa item is represented the degree of approximation of load state of the current service of a described head number expression.Described modification represents that the step of at least one data item of state of the current link information of server of described selection comprises:
Described mantissa item is increased the step-length less than described weighted value;
Whether more described mantissa item more than or equal to described weighted value,
If described mantissa item is more than or equal to described weighted value,
With the increase by 1 of described head number,
Described mantissa item is deducted described weighted value;
And
Alternatively, when disconnecting a connection, described access server comprises step in the step of at least one data item of the state of the current link information of server of the described selection of modification expression:
Whether more described mantissa item less than described step-length,
If described mantissa item is not less than described step-length,
Described mantissa item is deducted described step-length;
If described mantissa item is less than described step-length,
Whether more described head number greater than 0,
If described head number is not more than 0,
Return error message, end process;
If described head number is greater than 0,
Described head number is deducted 1,
Described mantissa item is increased a described weighted value, and then deduct described step-length.
Preferably, this method comprises the step that pre-determines each server traversal order in the group of server and also comprises the start address of picked at random traversal and according to the step of described order traversal in the step of the described head number of described traversal.
Alternatively, at described more described head number, when selecting also to comprise first several that have a plurality of minimums in the step of access server, select the step of any the pairing server access among first several the pairing servers of the minimum that traversal has access at first or described a plurality of minimums first several.
Preferably, the described step of state that pre-determines the current link information of described each server of expression comprises the step of data structure of the state of the current link information of determining described each server of storage before the system initialization, with access server described in the system operational process previous be connected disconnect or insert in determine the step of state of the current link information of described access server.
Alternatively, the described step that pre-determines the state of the current link information of representing described each server also comprises the step of the current linking number item of determining the described current link information of expression, wherein, described current linking number item is accurately represented the total quantity of described current service; Describedly determine that described modification represents that the step of at least one data item of state of the current link information of server of described selection comprises:
Described current linking number item is increased a step-length less than the described weighted value of expression;
Described mantissa item is increased a described step-length;
Whether more described mantissa item more than or equal to described weighted value,
If described mantissa item is more than or equal to described weighted value,
Described first data item is increased by 1,
Described mantissa item is deducted described weighted value;
And
When described access server disconnects a connection, also comprise the steps:
Described current linking number item is reduced a described step-length;
Whether more described mantissa item less than described step-length,
If described mantissa item is not less than described step-length,
Described mantissa item is deducted described step-length;
If described mantissa item is less than described step-length,
Whether more described head number greater than 0,
If described head number is not more than 0,
Return error message, end process;
If described head number is greater than 0,
Described head number is deducted 1,
Described mantissa item is increased a described weighted value, and then subtract
Go described step-length.
Alternatively, the method for the invention realizes on fixed-point processor.
As can be seen from the above technical solutions, the present invention trades space for time, by increasing the present load situation that a plurality of data item write down described each server, the computing of only need making comparisons when load balance scheduling needn't be made division arithmetic, reduced the operand of scheduling, accelerated the schedule speed of load balancing, and method is simple, equilibrium degree is good, scheduling preferably realizes easily, even also can be arranged on the simple fixed-point processor of function.Each connection to new foundation or disconnection, the data item of representing described server connection state makes corresponding changes, thereby the data item that makes the described server connection state of expression can reflect the load state of server in real time, make the result of load dispatch more accurate, make the distribution of load balanced more.The picked at random of traversal start address has also reduced the short time internal burden and has distributed unbalanced possibility.The present invention can also improve the precision of load balance scheduling by choosing of dilution of precision.
Description of drawings
Fig. 1 represents the process flow diagram of the operational process of one embodiment of the present of invention;
Fig. 2 represents the synoptic diagram of the data item structure of the current link information of each server in the storage server group;
Fig. 3 represents the data structure synoptic diagram of group of server traversal order linear list;
Implement the FB(flow block) of load balance scheduling task in Fig. 4 presentation graphs 1;
Fig. 5 represents to revise when server increases a connection newly the FB(flow block) of data item shown in Figure 3;
Fig. 6 represents to revise when server disconnects a connection FB(flow block) of data item shown in Figure 1.
Embodiment
Below in conjunction with accompanying drawing and embodiment the present invention is described in further detail.
This example realizes weighting minimum joint number dispatch server load balancing method with fixed-point processor as dispatch processor.Fig. 1 represents the process flow diagram of the operational process of one embodiment of the present of invention.At first in the step 30 of process flow diagram shown in Figure 1, choose the weighted value of each server in the group of server according to the performance of server.Weighted value is big more, and the chance that obtains connection is also big more.Be generally the strong server of load capacity and choose bigger weighted value.For example a group of server is made up of 5 station servers.The load capacity of server with the maximum of this server allow TCP linking number CMAXi (i=1 ..., 5) expression, the current service total quantity of server with the current TCP linking number Ci of this server (i=1 ..., 5) represent.For simplicity, hereinafter abbreviate current TCP linking number as current linking number.The maximum of supposing 5 station servers allows the TCP linking number to be respectively 1500,2000,1000,2000,3000, and the weighted value wi that chooses for this 5 station server (i=1 ..., 5) be respectively: 0.15,0.21,0.09,0.18,0.271.For improving computational accuracy, also adopted dilution of precision p=0.03 that weighted value has been carried out refinement in this example, the method for refinement is as follows:
Wi＝INT(wi/p)(i＝1，...，5)，
Thereby obtain the weighted value 5,7,3,6,9 after the refinement.The selection principle of dilution of precision p be this factor not should greater than Max (wi) (i=1 ..., 5), and the remainder that obtains after will making wi calculate is as far as possible little, makes less being advisable of Wi numerical value of calculating gained simultaneously.This dilution of precision can not be an integer.This dilution of precision is to be issued to fixed-point processor that dispatching office uses after good calculating in advance on the controller.According to the minimum linking number scheduling of weighting principle, the next TCP connection when system moves can send to server j, and server j satisfies following requirement:
Cj/Wj＝min{Ci/Wi}(i＝1，...，5)。
System also needs to pre-determine the data structure (not being shown among Fig. 1) that comprises the state of the current link information of each server in the storage server group before operation.
Fig. 2 is the synoptic diagram of the data item 10 of the current link information of storage i station server in this example, comprising current linking number 110 (Ci) as the i station server of the 3rd predetermined data item, as first several 120 (Hi) of the current linking number of the i station server of the first predetermined data item, as the mantissa 130 (Ri) of the current linking number of the second data item i station server of being scheduled to.For simplicity, several 120 headed by the current linking number of hereinafter referred first several 120, be called for short current linking number mantissa 130 and be mantissa 130.System also needs to pre-determine the traversal order of each server in the group of server in this example before operation.Fig. 3 represents the linear list 20 of storage server group traversal order.Comprising the number of servers in the group of server 210, numerical value is 5 in this example.Store the index address of data item 10 of the current link information of 5 station servers after the number of servers 210 successively, ID1 represents the above-mentioned address 220 of the 1st real server correspondence, and IDi represents the above-mentioned address 2i0 of i station server correspondence.Because this example is the strategy that has adopted picked at random traversal start address, when the traversal real server, not each all from fixing ID1, but choose a random value, determine initial traversal address according to this random value, the 1st real server of the traversal beginning real server of ID1 correspondence not necessarily might be ID2 or ID3 corresponding server like this.Choosing of random number is the known technology of this area, as the random number that can choose 0-100 server platform in the group of server counted to add 1 start address that obtains the server traversal after the complementation.Take the advantage of picked at random traversal start address strategy to see below described.Another one needs established data to be the every increase of server or to disconnect a connection, the step-length that current linking number 110 and linking number mantissa 130 increase or reduce, generally be taken as 1, make step-length but can choose any one number that is no more than the minimum weight value of group of server.
Enter init state behind the system boot, step 40 promptly shown in Figure 1 comprises the current linking number 110 of data item that will storage 5 station servers current link information, first several 120, mantissa's 130 zero clearings, receives the weighted value or the like of each server of self-controller.When first user asked to insert, system judged whether new access request by step 50 shown in Figure 1, enters step 60, i.e. load balance scheduling.
Fig. 4 shows the process flow diagram of load balance scheduling task.In step 610, the fixed-point processor of operation dispatching at first utilizes known technology to determine a random number, as the random number that can choose 0-100 server platform in the group of server is counted to add 1 after the complementation and obtain a random number, supposes that the random number that obtains is 2.According to step 620, processor travels through the linking number first several 120 of each server since the 2nd station server.Judge whether to exist minimum first several 120 according to step 630, then select minimum first several 120 corresponding server to insert if exist, otherwise an optional station server inserts.Because at present all server linking numbers first several 120 are 0, can think not exist minimum first severally 120, can insert according to the optional station server of known technology.Another kind for the treatment of method is, it is identical minimum first several 120 to think that 5 station servers have, and that select to have access at first is that the ID2 corresponding server inserts.Persons skilled in the art understand that two kinds for the treatment of methods do not have substantial difference to the realization of inventing.Adopt a kind of disposal route in back in this example, the 2nd station server with minimum linking number first several 120 of selecting at first visit is as the server that is scheduled, and the request of will inserting is forwarded to the 2nd station server.Processor is revised the current link information of the 2nd station server simultaneously.
Fig. 5 represents the FB(flow block) of the newly-increased data item when revising the described increase of Fig. 3 when connecting and connecting of server.According to step 810, the current linking number 110 of the 2nd station server at first increases step-length 1.According to step 820, judge whether current linking number 110 surpasses the maximum number of connections of this server, if then provide error message, end process; Otherwise forward the step 830 that mantissa 130 adds step-length 1 to.Because present current linking number 110 numerical value are 1, much smaller than the maximum permission linking number of the 2nd station server, so enter step 830, promptly mantissa 130 adds step-length 1, and numerical value becomes 1.According to step 840, judge mantissa 130 whether more than or equal to the weighted value of this server, if, then forwarding step 850 to and carry out first several 120 and add step-length 1, mantissa 130 deducts the task of weighted value; Otherwise end process.By noted earlier, the weighted value of the 2nd station server is 7, and mantissa 130 is less than the weighted value of the 2nd station server, so jump to step 860 end process, the flow process of revising data item finishes.This balance dispatching task termination.Persons skilled in the art are understood, even do not adopt the strategy of picked at random traversal address, as adopt fixing traversal start address, perhaps travel through the method that start address is moved after successively, also can realize the present invention.Adopt the method for picked at random traversal address, be more conducive to load balancing at short notice.
For being without loss of generality, be that example illustrates scheduling process of the present invention with an intermediateness more below.Along with the increase and the end of user capture amount, the current linking number 110 of each server and head several 120 thereof and mantissa 130 are in continuous variation.If the current linking number 110 of certain moment 5 station server is respectively 17,25,13,23,37, each linking number first several 120 is followed successively by 3,3,4,3,4, and each linking number mantissa 130 is followed successively by 2,4,1,5,1 (first several 120 and mantissa 130 of linking number determines according to following step).When new services request inserts,, enter load balance scheduling step 60 according to the judgement of step 50 among Fig. 1.Processor is at first determined the start address of server traversal according to abovementioned steps, suppose to travel through since the 3rd station server, after finishing, traversal have the server of minimum linking number first several 120 to have 3 successively: the 4th, the 1st, the 2nd by traversal order, the 4th station server of selecting at first to have access to minimum linking number first several 120 is as the server that is scheduled, the request of will inserting is forwarded to the 4th station server, provides access service by the 4th station server.Processor is revised the current link information of expression the 4th station server simultaneously.The current linking number 110 of the 4th station server at first adds 1, and this moment, this linking number was 24, and less than the maximum number of connections of the 4th station server, so enter step 830, promptly mantissa 130 adds 1, and numerical value becomes 6.According to the judgement of step 840, by noted earlier, the weighted value of the 4th station server is 6, and mantissa 130 equals adding of the 4th station server, jumps to step 850, and first several 110 of the 4th station server adds 1, and numerical value becomes 4, and mantissa 130 deducts weighted value, and numerical value becomes 0.The current linking number 110 of 5 station servers is respectively 17,25,13,24,37 as a result, and each linking number first several 120 is followed successively by 3,3,4,4,4, and each linking number mantissa 130 is followed successively by 2,4,1,0,1, and the flow process of revising data item finishes.
Be example still with above-mentioned result, a connection supposing the 4th station server disconnects because of network problem, it is sure situation that Fig. 1 step 70 judged result promptly takes place, processor enters and disconnects the step 80 of revising data item when connecting, revises the flow process of data item when server promptly shown in Figure 6 disconnects a connection.At first according to step 661, the current linking number 110 of the 4th station server deducts step-length 1, and numerical value becomes 23.Whether judge current linking number mantissa 130 less than step-length 1 according to step 662,, whether judge first several 120 greater than 0, otherwise forward step 665 to, directly carry out the step that mantissa 130 deducts step-length 1 if then forward step 663 to.Because current linking number mantissa 130 numerical value are 0, the result is sure, enter step 663, judge that whether linking number first several 120 is greater than 0, if then forward step 664 to, carry out first several 120 and deduct step-length 1, mantissa 130 adds that weighted value deducts the task of step-length 1 again, otherwise provide error message, end process.Current first several 120 is 4, and judged result is sure, enters step 664, and first several 120 deduct step-length 1, and numerical value becomes 3, and mantissa 130 adds that weighted value deducts step-length 1, and numerical value becomes 5.Server disconnects the flow process of revising data item when connecting to be finished.
Above embodiment has introduced server load balancing technology how to use the present invention to realize the minimum connection distribution of weighting.Those skilled in the art will appreciate that thought of the present invention and method can also be used for other occasion and environment, as on floating point processor, realizing this invention.As long as relate to the load balance scheduling technology that adopts first number of linking number and linking number mantissa to avoid division arithmetic, all in protection scope of the present invention, concrete protection domain is seen claims.In addition, the foregoing description is the embodiment as demonstration of the present invention, is not limited to protection scope of the present invention.
Claims (10)
1. realize weighting minimum joint number dispatch server load balancing method for one kind, when new services request inserts, comprise the steps:
Pre-determine the state of the current link information of each server in the expression group of server, described state is represented by the data structure that comprises at least one data item, described data item comprises the data item of the load state of the current service of representing described each server, and the load state of described current service is by first several the expressions approx in described at least one data item;
In the described state of the described load state of representing current service approx of described each server of traversal first several;
More described head number,
If there are minimum first several, select first several the corresponding server of this minimum to insert;
If there is no first several of minimum, select arbitrary server to insert;
Revise at least one data item of the state of the current link information of server of representing described selection.
2. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 1, it is characterized in that the described step that pre-determines the state of the current link information of representing described each server also comprises the step of mantissa's item of determining the described current link information of expression, wherein, described mantissa item is represented the degree of approximation of load state of the current service of a described head number expression.
3. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 2 is characterized in that described method also comprises the step that pre-determines the weighted value of each server in the described group of server.
4. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 3 is characterized in that: described modification represents that the step of at least one data item of state of the current link information of server of described selection comprises:
Described mantissa item is increased the step-length less than described weighted value;
Whether more described mantissa item more than or equal to described weighted value,
If described mantissa item is more than or equal to described weighted value,
With the increase by 1 of described head number,
Described mantissa item is deducted described weighted value;
And
Described method also comprises the step of at least one data item of the state of revising the current link information of server of representing described selection when described access server disconnects a connection:
Whether more described mantissa item less than described step-length,
If described mantissa item is not less than described step-length,
Described mantissa item is deducted described step-length;
If described mantissa item is less than described step-length,
Whether more described head number greater than 0,
If described head number is not more than 0,
Return error message, end process;
If described head number is greater than 0,
Described head number is deducted 1,
Described mantissa item is increased a described weighted value, and then deduct described step-length.
5. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 3, it is characterized in that the described step that pre-determines the state of the current link information of representing described each server also comprises the step of the current linking number item of determining the described current link information of expression, wherein, described current linking number item is accurately represented the total quantity of described current service; Described modification represents that the step of at least one data item of state of the current link information of server of described selection comprises:
Described current linking number item is increased the step-length less than described weighted value;
Described mantissa item is increased a described step-length;
Whether more described mantissa item more than or equal to described weighted value,
If described mantissa item is more than or equal to described weighted value,
With the increase by 1 of described head number,
Described mantissa item is deducted described weighted value;
And
When described access server disconnects a connection, also comprise the steps:
Described current linking number item is deducted a described step-length;
Whether more described mantissa item less than described step-length,
If described mantissa item is not less than described step-length,
Described mantissa item is deducted described step-length;
If described mantissa item is less than described step-length,
Whether more described head number greater than 0,
If described head number is not more than 0,
Return error message, end process;
If described head number is greater than 0,
Described head number is deducted 1,
Described mantissa item is increased a described weighted value, and then deduct described step-length.
6. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 3 is characterized in that the step of the weighted value of described definite described each server comprises step:
According to the performance of described each server, determine the original weighted value of described each server;
Calculate the weighted value of described each server according to following formula;
Weighted value=INT (original weighted value/dilution of precision),
Wherein, employed weighted value when weighted value is represented the load dispatch computing, INT represents rounding operation, original weighted value is the original weighted value of above-mentioned each server, dilution of precision is according to the maximum original weighted value that is not more than described each server, the remainder that obtains after described original weighted value is calculated is as far as possible little, and the less principle of employed weighted value numerical value is selected when making described load dispatch computing.
7. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 1, it is characterized in that described method also comprises the step that pre-determines each server traversal order in the group of server, the step of the described head number of described traversal also comprises the start address of picked at random traversal, and according to the step of described order traversal.
8. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 1, it is characterized in that described more described head number, select the step of access server also to comprise: when having first several of a plurality of minimums, to select the step of any the pairing server access among first several the pairing servers of the minimum that traversal has access at first or described a plurality of minimums first several.
9. the realization weighting minimum joint number dispatch server load balancing method described in claim 1 is characterized in that described method realizes on fixed-point processor.
10. realization weighting minimum joint number dispatch server load balancing method as claimed in claim 1, it is characterized in that the described step of state that pre-determines the current link information of described each server of expression is included in the step of data structure of the state of the current link information of determining described each server of storage before the system initialization, with access server described in the system operational process previous be connected disconnect or insert in determine the step of state of the current link information of described access server.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031473083A CN100388183C (en) | 2003-07-05 | 2003-07-05 | Server load equalization method for implementing weighted minimum linked allocation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031473083A CN100388183C (en) | 2003-07-05 | 2003-07-05 | Server load equalization method for implementing weighted minimum linked allocation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1567173A CN1567173A (en) | 2005-01-19 |
| CN100388183C true CN100388183C (en) | 2008-05-14 |
Family
ID=34471893
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB031473083A Expired - Fee Related CN100388183C (en) | 2003-07-05 | 2003-07-05 | Server load equalization method for implementing weighted minimum linked allocation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100388183C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103166980A (en) * | 2011-12-08 | 2013-06-19 | 腾讯科技（深圳）有限公司 | Internet data pulling method and system |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7853953B2 (en) * | 2005-05-27 | 2010-12-14 | International Business Machines Corporation | Methods and apparatus for selective workload off-loading across multiple data centers |
| US7640023B2 (en) * | 2006-05-03 | 2009-12-29 | Cisco Technology, Inc. | System and method for server farm resource allocation |
| US9218208B2 (en) * | 2008-10-10 | 2015-12-22 | Alcatel Lucent | Method and system of traffic processor selection for broadcast/multicast service in a wireless network |
| CN103095683A (en) * | 2012-12-13 | 2013-05-08 | 北京奇虎科技有限公司 | Linux virtual server scheduling system and method |
| CN103002039A (en) * | 2012-12-13 | 2013-03-27 | 北京奇虎科技有限公司 | System and method for scheduling servers |
| CN103488733B (en) * | 2013-09-17 | 2017-10-27 | 新浪网技术(中国)有限公司 | A kind of method and device for importing data |
| CN103713956B (en) * | 2014-01-06 | 2017-01-25 | 山东大学 | Method for intelligent weighing load balance in cloud computing virtualized management environment |
| CN105763473A (en) * | 2016-04-14 | 2016-07-13 | 北京思特奇信息技术股份有限公司 | Load-balanced route selection method and system |
| CN106230962A (en) * | 2016-08-11 | 2016-12-14 | 浪潮电子信息产业股份有限公司 | A kind of grade variable algorithms based on service type load balancing |
| CN107547622B (en) * | 2017-06-28 | 2021-10-12 | 新华三技术有限公司 | Resource adjusting method and device |
| CN108737493B (en) * | 2018-04-02 | 2021-09-17 | 北京奇艺世纪科技有限公司 | Address allocation method and device and electronic equipment |
| CN108924212B (en) * | 2018-06-27 | 2021-06-29 | 郑州云海信息技术有限公司 | Load balancing method, system, device and computer readable storage medium |
| CN109600429A (en) * | 2018-11-28 | 2019-04-09 | 平安科技（深圳）有限公司 | Service shunting method, device, electronic equipment and storage medium |
| CN115277711A (en) * | 2022-08-02 | 2022-11-01 | 天翼数字生活科技有限公司 | Equipment scheduling method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1307287A (en) * | 2000-01-28 | 2001-08-08 | 国际商业机器公司 | Method and device for balancing load of image server |
| WO2002010917A1 (en) * | 2000-07-27 | 2002-02-07 | Bea Systems, Inc. | System and method for concentration and load-balancing of requests |
| WO2003050706A1 (en) * | 2001-12-12 | 2003-06-19 | Valve Corporation | Method and system for load balancing an authentication system |
-
2003
- 2003-07-05 CN CNB031473083A patent/CN100388183C/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1307287A (en) * | 2000-01-28 | 2001-08-08 | 国际商业机器公司 | Method and device for balancing load of image server |
| WO2002010917A1 (en) * | 2000-07-27 | 2002-02-07 | Bea Systems, Inc. | System and method for concentration and load-balancing of requests |
| WO2003050706A1 (en) * | 2001-12-12 | 2003-06-19 | Valve Corporation | Method and system for load balancing an authentication system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103166980A (en) * | 2011-12-08 | 2013-06-19 | 腾讯科技（深圳）有限公司 | Internet data pulling method and system |
| CN103166980B (en) * | 2011-12-08 | 2018-01-26 | 腾讯科技（深圳）有限公司 | Internet data pulls method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1567173A (en) | 2005-01-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100388183C (en) | Server load equalization method for implementing weighted minimum linked allocation | |
| CN103793272B (en) | Periodical task scheduling method and periodical task scheduling system | |
| CN109062658A (en) | Realize dispatching method, device, medium, equipment and the system of computing resource serviceization | |
| CN112416585B (en) | Deep learning-oriented GPU resource management and intelligent scheduling method | |
| CN110113387A (en) | A kind of processing method based on distributed batch processing system, apparatus and system | |
| CN108170530B (en) | Hadoop load balancing task scheduling method based on mixed element heuristic algorithm | |
| CN103188345A (en) | Distributive dynamic load management system and distributive dynamic load management method | |
| CN106233276A (en) | The coordination access control of network-accessible block storage device | |
| CN103338228A (en) | Cloud calculating load balancing scheduling algorithm based on double-weighted least-connection algorithm | |
| CN103927225A (en) | Multi-core framework Internet information processing and optimizing method | |
| CN110109756A (en) | A kind of network target range construction method, system and storage medium | |
| CN103401939A (en) | Load balancing method adopting mixing scheduling strategy | |
| Mansouri et al. | A hybrid data replication strategy with fuzzy-based deletion for heterogeneous cloud data centers | |
| CN111131486A (en) | Load adjustment method and device of execution node, server and storage medium | |
| CN117032902A (en) | Cloud task scheduling method for improving discrete particle swarm algorithm based on load | |
| CN104917839A (en) | Load balancing method for use in cloud computing environment | |
| CN100362507C (en) | A server load equalizing method | |
| CN114911613A (en) | Cross-cluster resource high-availability scheduling method and system in inter-cloud computing environment | |
| CN104809026B (en) | A kind of method that CPU computing resources are borrowed using remote node | |
| CN109062683A (en) | The method, apparatus and computer readable storage medium of host resource distribution | |
| CN106487854A (en) | Storage resource distribution method, device and system | |
| Kim et al. | Virtual machines placement for network isolation in clouds | |
| CN107402812A (en) | Cluster resource dispatching method, device, equipment and storage medium | |
| CN110430236A (en) | A kind of method and dispatching device of deployment business | |
| Jiang et al. | A Resource scheduling strategy in cloud computing based on multi-agent genetic algorithm |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080514 Termination date: 20150705 |
|
| EXPY | Termination of patent right or utility model |