CN108881432A - Cloud computing cluster load dispatching method based on GA algorithm - Google Patents
Cloud computing cluster load dispatching method based on GA algorithm Download PDFInfo
- Publication number
- CN108881432A CN108881432A CN201810622604.8A CN201810622604A CN108881432A CN 108881432 A CN108881432 A CN 108881432A CN 201810622604 A CN201810622604 A CN 201810622604A CN 108881432 A CN108881432 A CN 108881432A
- Authority
- CN
- China
- Prior art keywords
- task
- load
- algorithm
- time
- node
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
- H04L67/61—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources taking into account QoS or priority requirements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
- H04L67/1008—Server selection for load balancing based on parameters of servers, e.g. available memory or workload
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
- H04L67/62—Establishing a time schedule for servicing the requests
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
- H04L67/63—Routing a service request depending on the request content or context
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Genetics & Genomics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Hardware Design (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Cloud computing cluster load dispatching method based on GA algorithm, including：S1：Task queue is read, the operating parameter of each task is found out from " service role execution time model table "；S2：Operating parameter is encoded, operating parameter set is converted into chromosome bit string set, is produced initial population Group (g), g=0；S3：Calculate the fitness value of each individual in initial population；S4：Judge termination condition g>=Gmax (Gmax is maximum reproductive order of generation), if condition is unsatisfactory for, executes S5, otherwise goes to S8：S5：It carries out selection operation and forms next-generation population Group (g), g+=1, using the selection algorithm of simulated annealing；S6：With probability PcCarry out crossover operation；S7：With probability PmMutation operation is carried out, S3 is gone to；S8：Algorithm terminates, and exports current optimal scheduling scheme.The request task corresponding time is greatly shortened in the present invention, and GIS service quality can be also obviously improved in the case where user volume increases, and improves users' satisfaction degree.
Description
Technical field
The present invention relates to network load scheduling fields, and in particular to a kind of cloud computing cluster load dispatch based on GA algorithm
Method.
Background technique
The rapid development of cloud computing application, to its performance and service quality, more stringent requirements are proposed, multi-core processor with
And the rise of virtualization technology, it is the further development of cloud computing cluster, common load balance dispatching is common square in the prior art
Method is scheduled, and load balance scheduling algorithm is a key technology of cluster, oneself is through there is numerous dispatching algorithms in difference
Effective use has been obtained under scene.In the Web system of multi-user's high concurrent, the quality of load-balancing algorithm is directly affected
QoS is a hot spot of computer and related fields research all the time.In general, the algorithm partition is static load balancing
Algorithm and Dynamic Load-balancing Algorithm.
1. static load balancing algorithm
The algorithm mainly considers from pending task angle, it is believed that system oneself known that the operation of task is special
Property, such as the type of task, the time cost spent required for running.Then load-balancing algorithm is according to so known task
Characteristic calculates the optimal task schedule method of salary distribution.How as far as possible accurately obtain task operation characteristic, and how root
Go out optimal scheduling scheme according to these information plannings, is the key problem to be paid close attention to of the algorithm.
But static scheduling algorithm does not consider cluster operating status, and the operating status of cluster is continually changing, static state
Algorithm will appear the uneven situation of distribution unavoidably, waste resource utilization ratio.
2. Dynamic Load-balancing Algorithm
One important algorithm is to need load balancer timely to obtain from the viewpoint of the operating status of cluster
The use state of cluster Current resource such as CPU, memory, network etc. and the operating status of task.Algorithm is being scheduled it
Before, each node load situation can be evaluated according to system running state, current task generally can all be distributed to the smallest section of load
Point, to realize the scheduling of task.It is how accurate and timely to obtain cluster operating status and make load evaluation be the algorithm
Key problem.
Relative to static load balancing algorithm, dynamic load-balancing algorithm considers the situation of change of group system, and
Targetedly distribution is adjusted.The problem of load distribution, has generality, it may be considered that introduces from other field some new
Method new approaches solve the problems, such as.
Related science man has done comparison to the geometric space operation based on spatial database and has been concluded that the overwhelming majority
Spatial operation requires consumption great number CPU and a large amount of I/O, the complexity of spatial data determine spatial operation be a kind of CPU and
I/O intensity calculating task, and the difference of the complexity of different types of spatial operation is also very big.Space querying not only wants root
Ask the index that condition calculates spatial object according to Check, also the object inquired be read out and is modeled with objectification, be it is a kind of with
Task based on I/O, supplemented by CPU calculating；Spatial overlay analysis is a kind of binary operation, not only sets mass data from storage
Standby graftabl, it is also necessary to carry out complex space operation, complexity will often be higher by space querying one or more quantity
Grade.In a cloud with the different complexities such as space querying, dynamic cartography, spatial data editing, spatial analysis, network analysis
On the one hand computing cluster system, the pressure faced are that user volume is big, be on the other hand because of the greatest differences between request task
And the laod unbalance between the clustered node amplified.
Simple load-balancing algorithm such as polling method, be for the cluster that load requests differ greatly it is unsuitable, it is some
Increasingly complex Dynamic Load-balancing Algorithm can cope with such demand, but the complexity of itself and to system-computed and logical
The occupancy of news resource can introduce some new problems, affect the development process of cloud computing cluster.
Summary of the invention
In view of the deficiencies of the prior art, the purpose of the present invention is intended to provide a kind of virtualization cloud computing cluster load balance calculation
Method can also be obviously improved GIS service quality in the case where user volume increases so that the request task corresponding time is most short, improve
Users' satisfaction degree.
To achieve the above object, the present invention adopts the following technical scheme that：
Cloud computing cluster load dispatching method based on GA algorithm, including：
S1：Task queue is read, the operating parameter of each task is found out from " service role execution time model table "；
S2：Operating parameter is encoded, operating parameter set is converted into chromosome bit string set, produces initial population
Group (g), g=0；
S3：Calculate the fitness value of each individual in initial population；
S4：Judge termination condition g>=Gmax (Gmax is maximum reproductive order of generation), if condition is unsatisfactory for, executes S5, no
Then go to S8：
S5：It carries out selection operation and forms next-generation population Group (g), g+=1, using the selection algorithm of simulated annealing；
S6：With probability PcCarry out crossover operation；
S7：With probability PmMutation operation is carried out, S3 is gone to；
S8：Algorithm terminates, and exports current optimal scheduling scheme.
Further, a gene represents a task to be allocated in the chromosome bit string, and a gene representation is
(t, l, p, n), wherein t indicates mission number, and l indicates task scale, i.e. the expected of task executes the time, and p indicates that task is preferential
Grade, p is a Boolean, indicates high priority with 1 (True), can be prioritized in intra-node task execution, 0
(False) it indicates low priority, execution can be postponed, n indicates the clustered node server number that the task is assigned to, n's
Value takes random manner to generate, i.e., from the value range of n 0<n<Positive integer is randomly selected in Ns and gives n assignment, forms a dye
Colour solid；One chromosome is made of the task that one group of band distributes, the corresponding load dispatch scheme of a chromosome, a dyeing
The expression way of body be (t, l, p, n) | t>C,l>0, p=0 | 1,0<n<Ns }, wherein C is that load strategy converts threshold values (C=K*
Ns), 1 is positive integer, and Ns is clustered node quantity.
Further, the fitness value is indicated by fitness function, and the fitness function is positioned asIts
The node load amount deviation of middle reflection clustered node load balance degreeClustered node average load
Amount isNode load total amount isWherein clustered node loads
Measure (Li) it is current distribution load value (Lc) and expected (algorithm distribution) load value (Le) the sum of i.e. Li=Lc+Le, node currently bears
Load number of tasks is Tc, n-th of task is expected to execute the time as ctn, it is contemplated that loading commissions number is Te, when n-th of task is expected to execute
Between be etn。
Further, the current distribution load value (Lc) and node present load number of tasks TcIt is current to be referred to as clustered node
Load information, clustered node current load information are obtained by the way of load capacity snapshot.
Further, the mode of the load capacity snapshot refers to when executing scheduling scheme, and task distribution condition is synchronous
It being recorded in task execution real-time status snapshot table, each node also feeds back execution status of task into load balancer, and
When more new state record；And when executing fitness evaluation, task real-time status snapshot table can be generated, it is then real-time to task
State snapshot table is calculated and is analyzed, and obtains the value of present load amount.
Further, the task real-time status snapshot table record task ID, task priority, it is expected execute the time,
Time started and state.
Further, described expected when executing the time by average execution in " GIS service task execution time model table "
Between obtain, Jian time of beginning is time for distribute to node of task, and state value includes " 0 " and " 1 ", and " 0 " expression task execution fails,
" 1 " expression task is allocated successfully, if task execution success, deletes the scheduler task from real-time task table, is only protected in table
Stay the task record for being not carried out or executing failure, the executing failure of the task also can process periodically, reattempt to execution or to
Client reflects task execution failure.
Further, the selection algorithm of simulated annealing is expressed as in the step S5p(ai) it is select probability, n is population G={ a1,a2,...,anNumber, wherein
Individual aiAdaptive value be f (ai), T is annealing temperature.
Further, with probability P in the step S6cCarrying out crossover operation includes：
S61. it strikes a bargain in current population by the high group of individuals of one group of adaptive value of certain rule selection and matches pond, from mating pond
Randomly select a pair of of individual；
S62. one or more positions are randomly selected from chromosome bit string as position to be intersected；
S63. according to crossover probability PcIntersected, wherein crossover probability PcIt is set as 0.4~0.9, the individual of pairing is existed
At crossover location, respective bit string content is exchanged, forms two new individuals.
Further, the mutation probability P in the step S7mIt is 0.001~0.1.
The beneficial effects of the present invention are：
According to the particularity and complexity of GIS operation and application, establishes task based access control priority and execute the GIS of time
Task model divides priority to request task as genetic algorithm gene coding basis, in scheduling, allow the response time it is short,
The operation being commonly used preferentially is executed, to improve user satisfaction；The genetic algorithm for load balancing is devised, including
Fitness function, genetic operator and control parameter, and form algorithm flow so that the hardware resource in cluster obtain it is optimal
The utilization of change avoids the occurrence of certain excessively high and some nodes of node resource utilization rate and the relatively low or even idle feelings of utilization rate occurs
Condition.
Detailed description of the invention
Fig. 1 is that the present invention is based on the step schematic diagrames of the cloud computing cluster load dispatching method of GA algorithm；
Fig. 2 is mixed scheduling strategy flow chart of the present invention；
Fig. 3 is service role execution time model table of the present invention；
Fig. 4 is initial population schematic diagram of the present invention；
Fig. 5 is load information securing mechanism figure of the present invention；
Fig. 6 is task execution real-time status snapshot table of the present invention；
Fig. 7 is 20 concurrent user's average response time comparison diagrams of the invention；
Fig. 8 is 50 concurrent user's average response time comparison diagrams of the invention.
Specific embodiment
In the following, being described further in conjunction with attached drawing and specific embodiment to the present invention：
" GA algorithm " is expressed as genetic algorithm, as shown in Figure 1, the cloud computing cluster load dispatching method based on GA algorithm,
Including：
S1：Task queue is read, the operating parameter of each task is found out from " service role execution time model table ".
The case where above-mentioned " service role execution time model table ", is as follows：
As shown in figure 3, recite mission number, GIS service function, data mark in " service role execution time model table "
Knowledge, average performance times and priority, each task are expressed with t (l, p), are respectively indicated mission number, task scale and are appointed
Business priority.The task scale that l is indicated is reached using the expection implementation schedule of task here.P is a Boolean, with 1
(True) it indicates high priority, can be prioritized in intra-node task execution, 0 (False) indicates low priority, can quilt
Postpone execution." service role execution time model table " illustrates the example model of task presentation, the average performance times in table
The average value that will be tested according to cluster single node.
At the same time, selection Genetic Strategies are also needed, the cluster that the present invention constructs is a symmetrical cluster, i.e., each node
As computing capability with service type is, in addition, starting more complicated Dynamic Load-balancing Algorithm and can expend and certain be
System resource, leads to some delays, the invention proposes a kind of by poll and the scheduling plan combined based on genetic algorithm approach
Slightly, as shown in Figure 2, it is desired to be able to adapt to various cluster pressure scenes, improve overall performance.
1. requesting density Dr=Nr/Tapan
Indicate the GIS number of requests in the unit time.Wherein TapanThe sampling time section of request, NrFor in the period
Request number.
2. load strategy converts threshold values C=K*Ns
Indicate the critical value of switching load strategy.Wherein Ns is clustered node quantity, and it is one just whole that K, which is to adjust constant,
Numerical value.
The strategy is when cluster load pressure is smaller, using static robin manner.As request density DrReach one
Determine threshold values C when, enable the load balance scheduling strategy based on GA algorithm.
S2：Operating parameter is encoded, operating parameter set is converted into chromosome bit string set, produces initial population
Group (g), g=0.
A gene represents a task to be allocated in above-mentioned chromosome bit string, and a gene representation is (t, l, p, n),
Wherein t indicates mission number, and l indicates task scale, i.e. the expected of task executes the time, and p indicates task priority, and p is one
Boolean indicates high priority with 1 (True), can be prioritized in intra-node task execution that 0 (False) indicates low
Priority, can be postponed execution, and n indicates that the clustered node server number that the task is assigned to, the value of n take random side
Formula generates, i.e., from the value range of n 0<n<Positive integer is randomly selected in Ns and gives n assignment, forms a chromosome；One chromosome
It is made of the task that one group of band distributes, the corresponding load dispatch scheme of a chromosome, the expression way of a chromosome is
{(t,l,p,n)|t>C,l>0, p=0 | 1,0<n<Ns }, wherein C is that load strategy converts threshold values (C=K*Ns), and 1 is positive integer,
Ns is clustered node quantity.
It wherein needs to determine population scale before initialization of population, and population scale meaning is as follows：
Population scale, i.e. chromosome number have a significant impact to performance of genetic algorithms, and the population that a scale is n is lost
The complexity for passing the mode that operator is therefrom generated and examined is O (n3).Population scale is bigger, and group more has diversity, algorithm
A possibility that falling into local convergence is with regard to smaller, but the expansion of population scale also brings along dramatically increasing for calculation amount, to reduce
Convergence rate.Population scale is too small, then will limit the search space of algorithm, it is likely that the case where Premature Convergence occur.General feelings
Scholar suggests that population scale is between 20 to 200 under condition.Task execution time prediction model itself in view of the present embodiment
Non-precision, furthermore the such scheduling system real time of cluster is stronger, needs quickly to make scheduling decision, therefore receive from improving
It holds back speed to set out, population scale takes 20 here.
The individual of initial population is generated by random fashion here.T, l and p may be used in chromosome four-tuple (t, l, p, n)
With determination, random manner is taken to generate the value of n here, i.e., from the value range of n 0<n<Positive integer is randomly selected in Ns to n
Assignment forms a chromosome.Initial population schematic diagram is as shown in Figure 4.
S3：Calculate the fitness value of each individual in initial population.
Above-mentioned fitness function is assessed the chromosome adaptability of population, and executing the survival of the fittest is existence rule,
Directly determine Evolutionary direction and the behavior of group.The chromosome for having higher adaptive value is considered to have preferable bit string knot
A possibility that structure has preferable survival ability, generates outstanding offspring is bigger.For the present invention, fitness value is higher
Load dispatch scheme is easier to be retained in scheme pond (population), and more easily output optimization scheme.
Since forementioned gene coding can not reflect the sequence of scheduling, the present invention by optimal scheduling conceptual internal according to
The mode of priority ranking realizes the target of user satisfaction, i.e., in clustered node internal task queue, priority is high to appoint
Business first carries out.Here fitness function is considered from the optimization aim of cluster server state, on each node server
Utilization rate CPU is positively related in number of tasks and task execution time and server, can reflect the load state of CPU, in turn
Reflect node overall load situation.The scheduling scheme that dispatching algorithm obtains is a desired value of node present load, uses task
The summation of time is executed to quantify the value.
The fitness value is indicated that the fitness function is positioned as by fitness functionWherein reflection collection
The node load amount deviation of group node load balance degreeClustered node average load amount isNode load total amount isWherein clustered node load capacity
(Li) it is current distribution load value (Lc) and expected (algorithm distribution) load value (Le) the sum of i.e. Li=Lc+Le, node present load
Number of tasks is Tc, n-th of task is expected to execute the time as ctn, it is contemplated that loading commissions number is Te, n-th of task is expected to execute the time
For etn.Under normal circumstances, a fitness value always nonnegative value, and the bigger adaptability of numerical value is better.And node load amount deviation
It is smaller, illustrate that allocation plan is more outstanding, chromosome fitness is higher, therefore, expresses here using the inverse of load deviation suitable
Angle value is answered, numerical value is bigger, and load distribution scheme is better.
Above-mentioned current distribution load value (Lc) and node present load number of tasks TcIt is referred to as clustered node current load information,
Clustered node current load information is obtained by the way of load capacity snapshot.
Node present load amount is also the factor for having to consider, is obtained used here as the mode of load capacity snapshot current
Load information.The mechanism that load information obtains is executing scheduling scheme as shown in figure 5, the mode of the load capacity snapshot refers to
When, by task distribution condition synchronous recording into task execution real-time status snapshot table, each node is also anti-by execution status of task
It is fed in load balancer, with the record for the state of timely updating；And when executing fitness evaluation, task real-time status can be generated
Then snapshot table is calculated and is analyzed to task real-time status snapshot table, obtains the value of present load amount.
As shown in fig. 6, the task real-time status snapshot table record task ID, task priority, it is expected execute the time,
Time started and state.The expected execution time is by the average performance times in " GIS service task execution time model table "
It obtaining, Jian time beginning is the time that task is distributed to node, and state value includes " 0 " and " 1 ", and " 0 " indicates task execution failure,
" 1 " expression task is allocated successfully, if task execution success, deletes the scheduler task from real-time task table, is only protected in table
Stay the task record for being not carried out or executing failure, the executing failure of the task also can process periodically, reattempt to execution or to
Client reflects task execution failure.
S4：Judge termination condition g>=Gmax (Gmax is maximum reproductive order of generation), if condition is unsatisfactory for, executes S5, into
Otherwise row genetic manipulation goes to S8, terminate algorithm：
S5：It carries out selection operation and forms next-generation population Group (g), g+=1, using the selection algorithm of simulated annealing；Institute
The selection algorithm for stating simulated annealing in step S5 is expressed asp(ai) it is select probability, n
For population G={ a1,a2,...,anNumber, wherein individual aiAdaptive value be f (ai), T is annealing temperature.The algorithm is stronger
Local search ability the characteristics of allow to from falling into locally optimal solution in search process, and in the difference of genetic evolution
Annealing temperature can be used to control different pressure in stage.
S6：With probability PcCarry out crossover operation；Operation simulation nature evolve in sexual reproduction process, realize
The recombination of gene, on one side by original outstanding gene genetic to the next generation.The process of intersection includes following steps：
S61. it strikes a bargain in current population by the high group of individuals of one group of adaptive value of certain rule selection and matches pond, from mating pond
Randomly select a pair of of individual；
S62. one or more positions are randomly selected from chromosome bit string as position to be intersected；
S63. according to crossover probability PcIntersected, wherein crossover probability PcIt is set as 0.4~0.9, the individual of pairing is existed
At crossover location, respective bit string content is exchanged, forms two new individuals.
Crossover operator include a bit, two o'clock and multiple point crossover, consistent intersection etc..Here using the algorithm unanimously intersected, i.e.,
Each in chromosome bit string is all pressed into equal probabilities and carries out random uniform crossover.The general value of crossover probability be 0.4~
0.9, higher probability can accelerate convergence rate, but also will increase the probability of " precocity ", and crossover probability takes median 0.6 here.
S7：With probability PmMutation operation is carried out, S3 is gone to；Operation simulation nature evolve in gene mutation it is existing
As, mutation operator often can be to avoiding " precocity " from generating good preventive effect, on the one hand, mutation operator to evolve in lose
The allele of mistake is restored, and to keep population diversity, while the variation of appropriateness, also can be improved the local search of algorithm
Efficiency.For the stability for guaranteeing population, so that variation individual and its male parent are unlikely to generate excessive difference, mutation probability value
It is typically small, usually 0.001~0.1.
Here mutation operation is reduced to, the variation to chromosome (t, l, p, n), only need to make a variation to the value of n, i.e., from
0, which randomly selects a value into Ns, replaces original value.
S8：Algorithm terminates, and exports current optimal scheduling scheme.
Inventive algorithm is tested：Algorithm of the invention is inserted into cloud computing cluster load balance module, with before
Polling type, Smallest connection number algorithm be compared, the advance of algorithm is verified.The cluster of experiment is 5 physics
Server, 4 are used as GIS server, two virtual machines of every server disposition, 8 dummy nodes formed altogether, every node deployment
One GIS service process；1 is used as Web server, disposes load balance scheduling module.The hardware configuration of server such as following table：
Server software configuration such as following table：
Experimental data is:
Tile data：Wuhan topographic map, 1-9 grades.
Vector data：1 to 5000 topographic map, point, line, area's figure layer each one, element sum is about 2500.
Test scene：The operation of map view, attribute query and spatial analysis, operation time are carried out to WebGIS system at random
Number such as following table,
| Map view | Hear to tile browsing | Attribute query | Cutting pattern analysis | Buffer area buffer area | |
| Number | 5 | 5 | 5 | 1 | 1 |
Foundation is recorded in this, as LoadRunner test scene.Test is directed to 20 users, 50 users, test run respectively
Time is 10 minutes.
Can most reflect performance indicator in the performance test results is average response time, and test result is aggregated, as schemed institute
Show, by Fig. 7, Fig. 8 as it can be seen that based on heredity dispatching algorithm can effectively shorten the task under multi-user concurrent situation complete when
Between, hence it is evident that improve the quality of GIS service.When user concurrent amount is little, effect is not particularly evident, but when user volume after
Continuous when increasing, improvement is with regard to obvious.
It will be apparent to those skilled in the art that can make various other according to the above description of the technical scheme and ideas
Corresponding change and deformation, and all these changes and deformation all should belong to the protection scope of the claims in the present invention
Within.
Claims (10)
1. the cloud computing cluster load dispatching method based on GA algorithm, which is characterized in that including：
S1：Task queue is read, the operating parameter of each task is found out from " service role execution time model table "；
S2：Operating parameter is encoded, operating parameter set is converted into chromosome bit string set, produces initial population
Group (g), g=0；
S3：Calculate the fitness value of each individual in initial population；
S4：Judge termination condition g>=Gmax (Gmax is maximum reproductive order of generation), if condition is unsatisfactory for, executes S5, otherwise turns
To S8：
S5：It carries out selection operation and forms next-generation population Group (g), g+=1, using the selection algorithm of simulated annealing；
S6：With probability PcCarry out crossover operation；
S7：With probability PmMutation operation is carried out, S3 is gone to；
S8：Algorithm terminates, and exports current optimal scheduling scheme.
2. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that the dye
A gene represents a task to be allocated in colour solid bit string, and a gene representation is (t, l, p, n), and wherein t indicates task
Number, l indicate task scale, i.e. the expected of task executes the time, and p indicates task priority, and p is a Boolean, with 1
(True) it indicates high priority, can be prioritized in intra-node task execution, 0 (False) indicates low priority, can quilt
Postpone execution, n indicates that the clustered node server number that the task is assigned to, the value of n take random manner to generate, i.e., from
The value range 0 of n<n<Positive integer is randomly selected in Ns and gives n assignment, forms a chromosome；One chromosome is by one group of band point
The matching of the task is constituted, the corresponding load dispatch scheme of a chromosome, the expression way of a chromosome be (t, l, p, n) |
t>C,l>0, p=0 | 1,0<n<Ns }, wherein C is that load strategy converts threshold values (C=K*Ns), and 1 is positive integer, and Ns is cluster section
Point quantity.
3. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that described suitable
Angle value is answered to be indicated by fitness function, the fitness function is positioned asWherein reflect clustered node load balancing
The node load amount deviation of degreeClustered node average load amount is
Node load total amount isWherein clustered node load capacity (Li) it is current distribution load value
(Lc) and expected (algorithm distribution) load value (Le) the sum of i.e. Li=Lc+Le, node present load number of tasks is Tc, n-th of task
It is expected that executing the time is ctn, it is contemplated that loading commissions number is Te, n-th of task is expected to execute the time as etn。
4. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that described to work as
Preceding distribution load value (Lc) and node present load number of tasks TcIt is referred to as clustered node current load information, clustered node is currently born
Information carrying breath is obtained by the way of load capacity snapshot.
5. the cloud computing cluster load dispatching method according to claim 4 based on GA algorithm, which is characterized in that described negative
The mode of carrying capacity snapshot refers to when executing scheduling scheme, and task distribution condition synchronous recording is fast to task execution real-time status
According in table, each node is also by execution status of task feedback into load balancer, with the record for the state of timely updating；And it is executing
When fitness evaluation, task real-time status snapshot table can be generated, then task real-time status snapshot table is calculated and analyzed,
Obtain the value of present load amount.
6. the cloud computing cluster load dispatching method according to claim 5 based on GA algorithm, which is characterized in that described
State snapshot table records task ID, task priority, expected execution time, time started and state when pragmatic.
7. the cloud computing cluster load dispatching method according to claim 6 based on GA algorithm, which is characterized in that described pre-
Phase execute the time by " GIS service task execution time model table " average performance times obtain, Jian begin the time be task to
The time of node distribution, state value include " 0 " and " 1 ", and " 0 " indicates task execution failure, and " 1 " expression task is allocated successfully, if
Task execution success, then delete the scheduler task from real-time task table, only retains in table and is not carried out or executes appointing for failure
Business record, the executing failure of the task also can be processed periodically, reattempted to execution or failed to client reflection task execution.
8. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that the step
The selection algorithm of simulated annealing is expressed as in rapid S5p(ai) it is select probability, n is kind
Group G={ a1,a2,...,anNumber, wherein individual aiAdaptive value be f (ai), T is annealing temperature.
9. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that the step
With probability P in rapid S6cCarrying out crossover operation includes：
S1. it strikes a bargain in current population by the high group of individuals of one group of adaptive value of certain rule selection and matches pond, it is random from mating pond
Choose a pair of of individual；
S2. one or more positions are randomly selected from chromosome bit string as position to be intersected；
S3. according to crossover probability PcIntersected, wherein crossover probability PcIt is set as 0.4~0.9, the individual of pairing is being intersected
At position, respective bit string content is exchanged, forms two new individuals.
10. the cloud computing cluster load dispatching method according to claim 1 based on GA algorithm, which is characterized in that described
Mutation probability P in step S7mIt is 0.001~0.1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810622604.8A CN108881432A (en) | 2018-06-15 | 2018-06-15 | Cloud computing cluster load dispatching method based on GA algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810622604.8A CN108881432A (en) | 2018-06-15 | 2018-06-15 | Cloud computing cluster load dispatching method based on GA algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108881432A true CN108881432A (en) | 2018-11-23 |
Family
ID=64339471
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810622604.8A Pending CN108881432A (en) | 2018-06-15 | 2018-06-15 | Cloud computing cluster load dispatching method based on GA algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108881432A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726001A (en) * | 2018-12-29 | 2019-05-07 | 中山大学 | A kind of genetic algorithm for heterogeneous system |
| CN110109753A (en) * | 2019-04-25 | 2019-08-09 | 成都信息工程大学 | Resource regulating method and system based on various dimensions constraint genetic algorithm |
| CN110389819A (en) * | 2019-06-24 | 2019-10-29 | 华中科技大学 | A kind of dispatching method and system of computation-intensive batch processing task |
| CN110888719A (en) * | 2019-09-18 | 2020-03-17 | 广州市巨硅信息科技有限公司 | Distributed task scheduling system and method based on web service |
| CN111683132A (en) * | 2020-06-04 | 2020-09-18 | 重庆英楼维兴信息科技有限公司 | Business distribution method based on micro-service architecture and related device |
| CN112765463A (en) * | 2021-01-12 | 2021-05-07 | 陈漩 | Data management method for big data and user requirements and cloud computing server |
| CN113361833A (en) * | 2020-03-02 | 2021-09-07 | 联芯集成电路制造(厦门)有限公司 | Chemical mechanical polishing system and related dispatching management method |
| CN113411369A (en) * | 2020-03-26 | 2021-09-17 | 山东管理学院 | Cloud service resource collaborative optimization scheduling method, system, medium and equipment |
| CN117272838A (en) * | 2023-11-17 | 2023-12-22 | 恒海云技术集团有限公司 | Government affair big data platform data acquisition optimization method |
| CN117688102A (en) * | 2023-12-28 | 2024-03-12 | 江苏北联宝德计算机系统有限公司 | Operation data management system and method based on artificial intelligence |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102063339A (en) * | 2010-12-21 | 2011-05-18 | 北京高森明晨信息科技有限公司 | Resource load balancing method and equipment based on cloud computing system |
| CN103024048A (en) * | 2012-12-17 | 2013-04-03 | 南京邮电大学 | Resources scheduling method under cloud environment |
| CN103365727A (en) * | 2013-07-09 | 2013-10-23 | 南京大学 | Host load forecasting method in cloud computing environment |
| CN103902375A (en) * | 2014-04-11 | 2014-07-02 | 北京工业大学 | Cloud task scheduling method based on improved genetic algorithm |
| CN104811491A (en) * | 2015-04-17 | 2015-07-29 | 华南理工大学 | Cloud computing resource scheduling method based on genetic algorithm |
| CN104932938A (en) * | 2015-06-16 | 2015-09-23 | 中电科软件信息服务有限公司 | Cloud resource scheduling method based on genetic algorithm |
| CN105740051A (en) * | 2016-01-27 | 2016-07-06 | 北京工业大学 | Cloud computing resource scheduling realization method based on improved genetic algorithm |
-
2018
- 2018-06-15 CN CN201810622604.8A patent/CN108881432A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102063339A (en) * | 2010-12-21 | 2011-05-18 | 北京高森明晨信息科技有限公司 | Resource load balancing method and equipment based on cloud computing system |
| CN103024048A (en) * | 2012-12-17 | 2013-04-03 | 南京邮电大学 | Resources scheduling method under cloud environment |
| CN103365727A (en) * | 2013-07-09 | 2013-10-23 | 南京大学 | Host load forecasting method in cloud computing environment |
| CN103902375A (en) * | 2014-04-11 | 2014-07-02 | 北京工业大学 | Cloud task scheduling method based on improved genetic algorithm |
| CN104811491A (en) * | 2015-04-17 | 2015-07-29 | 华南理工大学 | Cloud computing resource scheduling method based on genetic algorithm |
| CN104932938A (en) * | 2015-06-16 | 2015-09-23 | 中电科软件信息服务有限公司 | Cloud resource scheduling method based on genetic algorithm |
| CN105740051A (en) * | 2016-01-27 | 2016-07-06 | 北京工业大学 | Cloud computing resource scheduling realization method based on improved genetic algorithm |
Non-Patent Citations (2)
| Title |
|---|
| 李源林: "基于服务器虚拟化的网络GIS集群关键技术研究", 《中国博士学位论文全文数据库 基础科学辑》 * |
| 胡艳华;唐新来: "基于改进遗传算法的云计算任务调度算法", 《计算机技术与发展》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726001A (en) * | 2018-12-29 | 2019-05-07 | 中山大学 | A kind of genetic algorithm for heterogeneous system |
| CN110109753A (en) * | 2019-04-25 | 2019-08-09 | 成都信息工程大学 | Resource regulating method and system based on various dimensions constraint genetic algorithm |
| CN110389819A (en) * | 2019-06-24 | 2019-10-29 | 华中科技大学 | A kind of dispatching method and system of computation-intensive batch processing task |
| CN110888719A (en) * | 2019-09-18 | 2020-03-17 | 广州市巨硅信息科技有限公司 | Distributed task scheduling system and method based on web service |
| CN113361833A (en) * | 2020-03-02 | 2021-09-07 | 联芯集成电路制造(厦门)有限公司 | Chemical mechanical polishing system and related dispatching management method |
| US11397425B2 (en) | 2020-03-02 | 2022-07-26 | United Semiconductor (Xiamen) Co., Ltd. | CMP polishing system and associated pilot management system |
| CN113411369A (en) * | 2020-03-26 | 2021-09-17 | 山东管理学院 | Cloud service resource collaborative optimization scheduling method, system, medium and equipment |
| CN111683132A (en) * | 2020-06-04 | 2020-09-18 | 重庆英楼维兴信息科技有限公司 | Business distribution method based on micro-service architecture and related device |
| CN111683132B (en) * | 2020-06-04 | 2022-07-19 | 重庆金窝窝网络科技有限公司 | Business distribution method based on micro-service architecture and related device |
| CN112765463A (en) * | 2021-01-12 | 2021-05-07 | 陈漩 | Data management method for big data and user requirements and cloud computing server |
| CN117272838A (en) * | 2023-11-17 | 2023-12-22 | 恒海云技术集团有限公司 | Government affair big data platform data acquisition optimization method |
| CN117272838B (en) * | 2023-11-17 | 2024-02-02 | 恒海云技术集团有限公司 | Government affair big data platform data acquisition optimization method |
| CN117688102A (en) * | 2023-12-28 | 2024-03-12 | 江苏北联宝德计算机系统有限公司 | Operation data management system and method based on artificial intelligence |
| CN117688102B (en) * | 2023-12-28 | 2024-06-07 | 江苏北联宝德计算机系统有限公司 | Operation data management system and method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108881432A (en) | Cloud computing cluster load dispatching method based on GA algorithm | |
| CN111027736B (en) | Micro-service combined deployment and scheduling method under multi-objective optimization | |
| Wang et al. | Load balancing task scheduling based on genetic algorithm in cloud computing | |
| CN110134495B (en) | Container cross-host online migration method, storage medium and terminal equipment | |
| CN105704255B (en) | A kind of server load balancing method based on genetic algorithm | |
| CN103631657B (en) | A kind of method for scheduling task based on MapReduce | |
| CN110231976B (en) | Load prediction-based edge computing platform container deployment method and system | |
| CN108170530B (en) | Hadoop load balancing task scheduling method based on mixed element heuristic algorithm | |
| CN105471985A (en) | Load balance method, cloud platform computing method and cloud platform | |
| CN105718364A (en) | Dynamic assessment method for ability of computation resource in cloud computing platform | |
| CN104461739B (en) | A kind of virtual machine batch dispositions method based on cloudsim platforms | |
| CN108845886B (en) | Cloud computing energy consumption optimization method and system based on phase space | |
| Zhou et al. | Strategy optimization of resource scheduling based on cluster rendering | |
| Shukla et al. | FAT-ETO: Fuzzy-AHP-TOPSIS-Based efficient task offloading algorithm for scientific workflows in heterogeneous fog–cloud environment | |
| Song et al. | A deep reinforcement learning-based task scheduling algorithm for energy efficiency in data centers | |
| Li | Optimization of task offloading problem based on simulated annealing algorithm in MEC | |
| Tang | Load balancing optimization in cloud computing based on task scheduling | |
| CN114201306B (en) | Multi-dimensional geographic space entity distribution method and system based on load balancing technology | |
| Khattar et al. | Multi-criteria-based energy-efficient framework for VM placement in cloud data centers | |
| Mohd. Zin et al. | Replication techniques in data grid environments | |
| Mao et al. | A fine-grained and dynamic MapReduce task scheduling scheme for the heterogeneous cloud environment | |
| CN109040283A (en) | A kind of modified load-balancing algorithm based on difference reaction type | |
| Zhu et al. | Load balancing algorithm for web server based on weighted minimal connections | |
| Park et al. | Reinforcement Learning based load balancing in a distributed heterogeneous storage system | |
| Jarvis et al. | Performance-based middleware services for grid computing |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181123 |