CN104301241A - SOA dynamic load distribution method and system - Google Patents
SOA dynamic load distribution method and system Download PDFInfo
- Publication number
- CN104301241A CN104301241A CN201410249317.9A CN201410249317A CN104301241A CN 104301241 A CN104301241 A CN 104301241A CN 201410249317 A CN201410249317 A CN 201410249317A CN 104301241 A CN104301241 A CN 104301241A
- Authority
- CN
- China
- Prior art keywords
- service
- node
- service node
- operation flow
- soa
- 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
Landscapes
- Multi Processors (AREA)
Abstract
The invention relates to an SOA dynamic load distribution method and system. The method is characterized by when a service call request of a service flow arrives, extracting service ability information of each service node; calculating the utilization rate of each service node; and carrying out distribution to enable the utilization rate of each service node obtained after distribution to tend to be balanced, and thus the problem that with an existing distribution method, load dynamic balance and distribution node overhead cannot be taken into consideration is solved.
Description
Technical field
The present invention relates to a kind of service-oriented dynamic load distributing method.
Background technology
The main operation modes having become resource-sharing, task cooperative etc. is served under SOA environment.It is mutual that the performance of SOA application depends between the composite behaviour of collaboration services and collaboration services.
If Fig. 1 is the load management measure of service invocation request being carried out to each service node distributing.When a request arrives, first need to carry out decision-making management through distribution node according to certain rule, then select some service nodes to carry out service call.Two sides participating in distribution decision-making are in this mode the service node set of request and the rear end arrived respectively.The method realizing front end distribution mode mainly comprises: state algorithm, dynamic algorithm and based on request content algorithm.
State algorithm comprises:
Wheel robin (Round-Roubin, RR), thinks that back-end services node has identical priority, adopts the mode of serving in turn to provide service.When a request arrives, alternately is selected backend nodes by distribution node.
Weighted round robin method (Weighted Round-Roubin, WRR), service node ability be weighted, node weights are larger, provide the probability of service larger.
Random distribution method (Random Dispatch, RD), Stochastic choice service node.
Address hash method (Address Hash, AH), is divided into source address and destination address hash, finds corresponding service node according to request to the address of request server as hashed key.
Dynamic algorithm comprises:
Smallest connection method (Least-Connection, LC), estimates load with the linking number on service node, is the service node that service request selects current linking number minimum.
Minimum load method (Least-Loaded, LL), distribution node cycle or collect the load state of service node is in real time the service node that the service request arrived selects present load minimum.
Based on request content algorithm: service request, according to the difference request of Web service resource, is distributed in different Web services by these class methods.Distribution algorithms according to request content is combined comparatively tight with corresponding back-end services node, different application request all has different demands to CPU, internal memory, bandwidth, the method be applied to provide function comparatively complicated and diversified Web Application Server time effect better.
For the decision-making both sides in distribution mode, request is generally workflow, operation flow, the application system requiring resource etc., the change of fault or single service can have influence on multiple operation flow, increase new application or operation flow can make existing service transship, make current business flow process hydraulic performance decline, even break down.Because the service ability of service node is not identical, cause some node in service node set to be in overload, and have some nodes to be in light condition simultaneously, the performance that final impact is overall.Therefore, how to ensure the balancing dynamic load between each service node, improving entire system efficiency of service, is problem in the urgent need to address.
Above-mentioned state algorithm often only depends on static information, although distribution decision delay is lower, can not make a response in time along with the change of operation flow service invocation request, therefore, for the situation of operation flow dynamic change, load balance difficulty between each service node.
Based on the algorithm of dynamic load, distribution effect depends on distribution node and collects the comprehensive of current serving Node load information and real-time, but the processing expenditure of distribution node is larger.And needing to gather solicited message based on the algorithm of request content, Analysis Service request particular content, distribution node expense is larger.
Summary of the invention
The object of this invention is to provide a kind of SOA dynamic load distributing method and system, the problem of homeostasis of load and distribution node expense cannot be taken into account in order to solve existing distribution method.
For achieving the above object, the solution of the present invention comprises:
A kind of SOA dynamic load distributing method, step is as follows:
1) when the service invocation request of operation flow arrives, the service capability information of each service node is extracted;
2) utilance of each service node is calculated;
3) distribute, the first constraints of distribution is the least squares optimization of each service node utilance.
The least squares optimization of described each service node utilance be with
be minimised as target; Wherein ρ
ithe average service rate of service node i,
for
represent the average of each serving node services rate.
Also comprise the second constraints: make the Residual service time sum summation on specific service node maximum; Residual service time is the difference of maximum service response time and estimated average service time on each service node, represent this serving node services ability and operation flow to require service time between difference, difference shows that more greatly the service ability of this service node to operation flow is stronger.
According to the estimated value of the intensity that the service invocation request of operation flow arrives, the time interval of adjustment service load distribution; The time interval size of service load distribution and the arrival intensity of service invocation request are inverse ratio.
A kind of SOA dynamic load distribution systems, comprises as lower module:
1) when the service invocation request of operation flow arrives, the module of the service capability information of each service node is extracted;
2) module of the utilance of each service node is calculated;
3) carry out the module of distributing, the first constraints of distribution is the least squares optimization of each service node utilance.
The least squares optimization of described each service node utilance be with
be minimised as target; Wherein ρ
ithe average service rate of service node i,
for
represent the average of each serving node services rate.
Also comprise the second constraints: make the Residual service time sum summation on specific service node maximum; Residual service time is the difference of maximum service response time and estimated average service time on each service node, represent this serving node services ability and operation flow to require service time between difference, difference shows that more greatly the service ability of this service node to operation flow is stronger.
According to the estimated value of the intensity that the service invocation request of operation flow arrives, the time interval of adjustment service load distribution; The time interval size of service load distribution and the arrival intensity of service invocation request are inverse ratio.
The present invention proposes to balance as target with the utilance of service node, by setting up multiple-objection optimization object module and solving each service node load capacity, carries out the method for service load distribution, makes the load of each service node reach balance.
Further, turn to target so that the utilance balanced service node of service node residue service ability summation is maximum, make system have the ability of stronger anti-peak flow, strengthen the stability of a system.Service invocation request according to known operation flow arrives the real time information such as the service response time constraint of intensity and operation flow, carries out service load distribution, solve the problem of the inapplicable operation flow dynamic change of static distribution algorithms by service broker.
Further, the present invention utilizes the time interval of the load information dynamic conditioning delivery of services of acquisition, to improve efficiency of service and to control service broker's calculating strength.
Accompanying drawing explanation
Fig. 1 is that dissemination system forms schematic diagram;
Fig. 2 is the distribution flow figure of the embodiment of the present invention.
Embodiment
Below in conjunction with accompanying drawing, the present invention will be further described in detail.
Basic scheme of the present invention is as follows:
A kind of SOA dynamic load distributing method, when the service invocation request of operation flow arrives, extracts the service capability information of each service node; Calculate the utilance of each service node; Distribute, after making distribution, the utilance of each service node tends to balance.
Above thinking can balance each Duty-circle.Concrete, the parameters such as the average service rate that can extract with the service capability information from each service node characterize.
For example: utilance tend to balance with
be minimised as target; Wherein ρ
ithe average service rate of service node i,
for
represent the average of each serving node services rate.
Distribution method is completed by service broker, and service broker refers to service broker.In order to improve the stability of above-mentioned basic skills further, except the constraints that above-mentioned utilance balances, other constraintss can also be increased, as a kind of concrete execution mode of Fig. 2, add the constraints that patient time summation on service node is maximum, also add the dynamic conditioning to the distribution time interval.Specifically comprise step 1-step 5:
Step 1 system information inputs
The service invocation request of operation flow is distributed to each service node by service broker, and service broker will complete following work after receiving the service invocation request of each operation flow:
1. the arrival intensity of the service request of each operation flow is gathered;
2. the service response time constraint of each operation flow is extracted;
3. the service capability information of each service node is extracted.
Step 2 service load assigns matrix to generate
Service broker maximizes target according to the Residual service time sum summation in load balance constraints and service node and generates service load service load allocation matrix.The line number of service load service load allocation matrix represents operation flow, and row number represent service node.
In order to the load reaching each service node tends to balance and the Residual service time sum maximization target on service node, the present invention defines two constraintss:
First constraints, makes the utilance of each service node tend to balance, to promote overall service ability; Particularly, service node utilance tends to balance and is expressed as: the least squares optimization of each service node utilance.
Second constraints, makes the Residual service time sum summation on specific service node maximum, makes the ability of service node reply error also stronger.Wherein Residual service time is the difference of maximum service response time and estimated average service time on each service node, represent this serving node services ability and operation flow to require service time between difference, difference shows that more greatly the service ability of this service node to operation flow is stronger.
To sum up, services selection optimization problem can be summarized as a Nonlinear Multiobjective optimization problem, and relevant mathematical method can be applied solve.
Step 3 service load is distributed
Service broker is distributed to according to the service load service load allocation matrix generated in step 2 on each service node all service invocation request.Be described as follows with instantiation:
The parameter of the operation flow of system input is as shown in table 1, and wherein λ represents that the service request of operation flow arrives intensity, and respt represents the maximum service response time requirement of operation flow.
Table 1 operation flow parameter
5 service nodes with different service ability of setting are as shown in table 2, and wherein, μ represents the service intensity of node, and n represents the concurrent services quantity of service node.
Table 2 service node parameter
Under constraints, utilize genetic algorithm instrument to solve the service load service load allocation matrix A that obtains of constraint multiple target be:
This matrix notation i-th operation flow has α
ijservice request be assigned to a jth service node.
Step 4 service invocation request responds
Service node carries out processing and result being returned operation flow by service broker after receiving the service invocation request of service broker's distribution immediately.
Step 5 distributes time interval dynamic conditioning
The estimated value of the intensity that service broker arrives according to the service invocation request of operation flow, in the time interval of dynamic conditioning service load distribution, effectively controls the expense of service load distribution.The time interval size of service load distribution and the arrival intensity of service invocation request claim inverse ratio.When load is excessive, when constraints cannot meet, partial service node there will be overload, and the service request of queue tail abandons by service broker, and to the relevant information out of service of operation flow feedback.
Preparation process of the present invention comprises:
Service broker gathers the service ability parameter of service node, comprises the concurrent services number n of service node i
iwith service speed μ
i; These information of accurate acquisition accurately calculation services load can assign matrix, and the service node utilance in system is balanced.
Step of dealing with problems arising from an accident of the present invention comprises:
When the service invocation request of operation flow sharply increases, when the intensity of load that service broker estimates exceeds the service ability of service node, service broker no longer will accept service invocation request, and send request message out of service to each operation flow.
The randomness that between the multiserver that the present invention is directed to distributed deployment in SOA environment, performance difference and operation flow arrive causes service node laod unbalance problem, propose utilize service broker to the service request of operation flow meet service response time QoS constraint under with service node load balance for target, service is distributed, the utilance of each service node is similar to and reaches balance; Simultaneously to maximize service node patient time for optimization aim, to improve the ability of service system opposing peak flow, to improve the stability of service system.
Be presented above a kind of concrete execution mode, but the present invention is not limited to described execution mode.Thinking of the present invention is above-mentioned basic scheme, and for those of ordinary skill in the art, according to instruction of the present invention, designing the model of various distortion, formula, parameter does not need to spend creative work.The change carried out execution mode without departing from the principles and spirit of the present invention, amendment, replacement and modification still fall within the scope of protection of the present invention.
Claims (8)
1. a SOA dynamic load distributing method, is characterized in that, step is as follows:
1) when the service invocation request of operation flow arrives, the service capability information of each service node is extracted;
2) utilance of each service node is calculated;
3) distribute, the first constraints of distribution is the least squares optimization of each service node utilance.
2. a kind of SOA dynamic load distributing method according to claim 1, is characterized in that, the least squares optimization of described each service node utilance be with
be minimised as target; Wherein ρ
ithe average service rate of service node i,
for
represent the average of each serving node services rate.
3. a kind of SOA dynamic load distributing method according to claim 1, is characterized in that, also comprise the second constraints: make the Residual service time sum summation on specific service node maximum; Residual service time is the difference of maximum service response time and estimated average service time on each service node, represent this serving node services ability and operation flow to require service time between difference, difference shows that more greatly the service ability of this service node to operation flow is stronger.
4. a kind of SOA dynamic load distributing method according to claim 1 or 2 or 3, is characterized in that, according to the estimated value of the intensity that the service invocation request of operation flow arrives, and the time interval of adjustment service load distribution; The time interval size of service load distribution and the arrival intensity of service invocation request are inverse ratio.
5. a SOA dynamic load distribution systems, is characterized in that, comprises as lower module:
1) when the service invocation request of operation flow arrives, the module of the service capability information of each service node is extracted;
2) module of the utilance of each service node is calculated;
3) carry out the module of distributing, the first constraints of distribution is the least squares optimization of each service node utilance.
6. a kind of SOA dynamic load distribution systems according to claim 5, is characterized in that, the least squares optimization of described each service node utilance be with
be minimised as target; Wherein ρ
ithe average service rate of service node i,
for
represent the average of each serving node services rate.
7. a kind of SOA dynamic load distribution systems according to claim 5, is characterized in that, also comprise the second constraints: make the Residual service time sum summation on specific service node maximum; Residual service time is the difference of maximum service response time and estimated average service time on each service node, represent this serving node services ability and operation flow to require service time between difference, difference shows that more greatly the service ability of this service node to operation flow is stronger.
8. a kind of SOA dynamic load distribution systems according to claim 5 or 6 or 7, is characterized in that, according to the estimated value of the intensity that the service invocation request of operation flow arrives, and the time interval of adjustment service load distribution; The time interval size of service load distribution and the arrival intensity of service invocation request are inverse ratio.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410249317.9A CN104301241B (en) | 2014-06-05 | 2014-06-05 | A kind of SOA dynamic load distributing methods and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410249317.9A CN104301241B (en) | 2014-06-05 | 2014-06-05 | A kind of SOA dynamic load distributing methods and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104301241A true CN104301241A (en) | 2015-01-21 |
| CN104301241B CN104301241B (en) | 2018-01-09 |
Family
ID=52320814
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410249317.9A Active CN104301241B (en) | 2014-06-05 | 2014-06-05 | A kind of SOA dynamic load distributing methods and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104301241B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104580538A (en) * | 2015-02-12 | 2015-04-29 | 山东大学 | Method for improving load balance efficiency of Nginx server |
| CN105099936A (en) * | 2015-08-31 | 2015-11-25 | 联想(北京)有限公司 | Network resource configuration method and apparatus as well as network system |
| CN105491150A (en) * | 2015-12-28 | 2016-04-13 | 中国民航信息网络股份有限公司 | Load balance processing method based on time sequence and system |
| CN107205020A (en) * | 2017-05-05 | 2017-09-26 | 国网福建省电力有限公司 | Service load balancing method and system under Service-Oriented Architecture Based |
| CN108737192A (en) * | 2018-06-01 | 2018-11-02 | 北京航空航天大学 | Network service dispositions method based on service reliability |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002091936A (en) * | 2000-09-11 | 2002-03-29 | Hitachi Ltd | Device for distributing load and method for estimating load |
| CN101989951A (en) * | 2010-10-28 | 2011-03-23 | 北京瑞汛世纪科技有限公司 | Method, system and equipment for selecting service node |
| CN102004673A (en) * | 2010-11-29 | 2011-04-06 | 中兴通讯股份有限公司 | Processing method and system of multi-core processor load balancing |
| CN102185779A (en) * | 2011-05-11 | 2011-09-14 | 田文洪 | Method and device for realizing data center resource load balance in proportion to comprehensive allocation capability |
| CN102232282A (en) * | 2010-10-29 | 2011-11-02 | 华为技术有限公司 | Method and apparatus for realizing load balance of resources in data center |
| CN103118124A (en) * | 2013-02-22 | 2013-05-22 | 桂林电子科技大学 | Cloud computing load balancing method based on layering multiple agents |
| CN103226467A (en) * | 2013-05-23 | 2013-07-31 | 中国人民解放军国防科学技术大学 | Data parallel processing method and system as well as load balancing scheduler |
| CN103259832A (en) * | 2012-12-24 | 2013-08-21 | 中国科学院沈阳自动化研究所 | Cluster resource control method for achieving dynamic load balance, fault diagnosis and failover |
-
2014
- 2014-06-05 CN CN201410249317.9A patent/CN104301241B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002091936A (en) * | 2000-09-11 | 2002-03-29 | Hitachi Ltd | Device for distributing load and method for estimating load |
| CN101989951A (en) * | 2010-10-28 | 2011-03-23 | 北京瑞汛世纪科技有限公司 | Method, system and equipment for selecting service node |
| CN102232282A (en) * | 2010-10-29 | 2011-11-02 | 华为技术有限公司 | Method and apparatus for realizing load balance of resources in data center |
| CN102004673A (en) * | 2010-11-29 | 2011-04-06 | 中兴通讯股份有限公司 | Processing method and system of multi-core processor load balancing |
| CN102185779A (en) * | 2011-05-11 | 2011-09-14 | 田文洪 | Method and device for realizing data center resource load balance in proportion to comprehensive allocation capability |
| CN103259832A (en) * | 2012-12-24 | 2013-08-21 | 中国科学院沈阳自动化研究所 | Cluster resource control method for achieving dynamic load balance, fault diagnosis and failover |
| CN103118124A (en) * | 2013-02-22 | 2013-05-22 | 桂林电子科技大学 | Cloud computing load balancing method based on layering multiple agents |
| CN103226467A (en) * | 2013-05-23 | 2013-07-31 | 中国人民解放军国防科学技术大学 | Data parallel processing method and system as well as load balancing scheduler |
Non-Patent Citations (1)
| Title |
|---|
| 于国防等: "基于分配器队列模糊控制的集群负载平衡", 《计算机工程》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104580538A (en) * | 2015-02-12 | 2015-04-29 | 山东大学 | Method for improving load balance efficiency of Nginx server |
| CN104580538B (en) * | 2015-02-12 | 2018-02-23 | 山东大学 | A kind of method of raising Nginx server load balancing efficiency |
| CN105099936A (en) * | 2015-08-31 | 2015-11-25 | 联想(北京)有限公司 | Network resource configuration method and apparatus as well as network system |
| CN105099936B (en) * | 2015-08-31 | 2018-10-12 | 联想(北京)有限公司 | A kind of network resource configuration method, equipment and network system |
| CN105491150A (en) * | 2015-12-28 | 2016-04-13 | 中国民航信息网络股份有限公司 | Load balance processing method based on time sequence and system |
| CN107205020A (en) * | 2017-05-05 | 2017-09-26 | 国网福建省电力有限公司 | Service load balancing method and system under Service-Oriented Architecture Based |
| CN108737192A (en) * | 2018-06-01 | 2018-11-02 | 北京航空航天大学 | Network service dispositions method based on service reliability |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104301241B (en) | 2018-01-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chekired et al. | Industrial IoT data scheduling based on hierarchical fog computing: A key for enabling smart factory | |
| Lu et al. | Join-idle-queue: A novel load balancing algorithm for dynamically scalable web services | |
| Abhishek et al. | Fixed and market pricing for cloud services | |
| Liu et al. | D-storm: Dynamic resource-efficient scheduling of stream processing applications | |
| Gu et al. | Fairness-aware dynamic rate control and flow scheduling for network utility maximization in network service chain | |
| CN104301241A (en) | SOA dynamic load distribution method and system | |
| CN102724103B (en) | Proxy server, hierarchical network system and distributed workload management method | |
| CN103401947A (en) | Method and device for allocating tasks to multiple servers | |
| Liu et al. | Task scheduling in fog enabled Internet of Things for smart cities | |
| CN103713956A (en) | Method for intelligent weighing load balance in cloud computing virtualized management environment | |
| CN107222531A (en) | A kind of container cloud resource dispatching method | |
| CN102110014A (en) | Method for balancing loads of virtual machine (VM) | |
| CN108965014A (en) | The service chaining backup method and system of QoS perception | |
| Wen et al. | Load balancing job assignment for cluster-based cloud computing | |
| CN105975332A (en) | Method for forming multi-agent distributed union | |
| CN105119751A (en) | Service evaluation and selection method based on environment real-time perceiving | |
| Khodar et al. | New scheduling approach for virtual machine resources in cloud computing based on genetic algorithm | |
| CN104283963A (en) | Distributed and collaborative type CDN load balancing method | |
| Kadhim et al. | Hybrid load-balancing algorithm for distributed fog computing in internet of things environment | |
| CN104009904B (en) | The virtual network construction method and system of facing cloud platform big data processing | |
| CN102711178B (en) | Group association accepting control method based on user requirement and network load balancing | |
| CN103078806B (en) | A kind of load balance scheduling algorithm based on Q value method | |
| CN109600402A (en) | Server load balancing method and device, storage medium, server, service system | |
| CN103678000A (en) | Computational grid balance task scheduling method based on reliability and cooperative game | |
| CN104317657B (en) | The method and device of balanced statistics task in Real-Time Traffic Volume statistics |
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 |