CN104881271A - Cloud dynamic management method and device - Google Patents
Cloud dynamic management method and device Download PDFInfo
- Publication number
- CN104881271A CN104881271A CN201410069284.XA CN201410069284A CN104881271A CN 104881271 A CN104881271 A CN 104881271A CN 201410069284 A CN201410069284 A CN 201410069284A CN 104881271 A CN104881271 A CN 104881271A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- load
- sequence
- load value
- 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
- 238000007726 management method Methods 0.000 title claims abstract description 142
- 230000002159 abnormal effect Effects 0.000 claims abstract description 74
- 230000008859 change Effects 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims description 26
- 238000004364 calculation method Methods 0.000 claims description 13
- 238000005516 engineering process Methods 0.000 abstract description 5
- 230000006870 function Effects 0.000 description 11
- 230000008901 benefit Effects 0.000 description 6
- 238000005265 energy consumption Methods 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000003860 storage Methods 0.000 description 4
- 230000003247 decreasing effect Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000014509 gene expression Effects 0.000 description 2
- 230000035772 mutation Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000013468 resource allocation Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Landscapes
- Debugging And Monitoring (AREA)
- Computer And Data Communications (AREA)
Abstract
The invention discloses a cloud dynamic management method and device, and belongs to the cloud technology in the computer field, and solves problems that a present cloud management method is long in time delay, low in accuracy and the like. The cloud dynamic management method comprises that load information of a cloud system is obtained, the load information comprises a first load sequence formed in chronological sequence, whether an abnormal load value which deviates from the overall change rule of the first load sequence exists in the first load sequence or not is judged, if an abnormal load value exists, based on normal load values in the first load sequence and a preset first function relationship, a replacing load value which replaces the abnormal load value is worked out, the abnormal load value in the first load sequence is replaced by the replacing load value, a second load sequence is formed, and cloud management is carried out according to the second load sequence.
Description
Technical Field
The invention relates to a cloud management technology in the field of computers, in particular to a cloud dynamic management method and device.
Background
The cloud system is a system which completes calculation by distributing on a large number of distributed computers, and a cloud system platform is used for providing information service, namely cloud service, for users through a network.
Compared with the traditional software form, the cloud computing in the cloud system has the remarkable advantages of loose coupling, on-demand, controllable cost, virtual resources, heterogeneous cooperation and the like, so that the cloud computing is more suitable for applications of electronic commerce, flexible manufacturing, mobile internet and the like of cash. Cloud computing includes two aspects: the cloud system platform infrastructure constructed at the bottom layer is used for constructing the foundation of an upper application program; and secondly, constructing a cloud system application program on the basic platform.
The cloud system belongs to a new technical field, related technologies, theories and methods are still in a growth period, and the existing cloud resource management technology has a plurality of defects, for example, because the existing cloud resources (specifically, physical machines in the IaaS system) are operated, a decision for increasing or reducing the virtual machine resources is made after the system finds that the access load is increased or reduced, the obvious hysteresis is provided, the situation of sudden load increase or sudden load reduction cannot be responded, and the management efficiency and the management precision are very low.
Disclosure of Invention
In order to solve the above problems, the present invention provides a cloud dynamic management method and apparatus capable of improving the efficiency and accuracy of cloud management by a cloud system.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
the invention provides a cloud dynamic management method in a first aspect, which comprises the following steps:
acquiring load information of a cloud system, wherein the load information comprises a first load sequence;
judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
if the abnormal load value exists, calculating a replacement load value for replacing the abnormal load value according to the normal load value in the first load sequence and a preset first function relation;
replacing abnormal load values in the first load sequence with the replacement load values to form a second load sequence;
and performing cloud management according to the second load sequence.
Preferably, the determining whether there is an abnormal load value deviating from the overall change rule of the first load sequence in the first load sequence includes:
calculating a positive rate of change RP of the first load sequencemAnd negative rate of change RNm；
The load value satisfying the following relation in the first load sequence is an abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence;
the ld isj-1Is the j-1 th load value in the first load sequence;
the ld isj+1Is the j +1 th load value in the first load sequence;
the beta is a coefficient factor;
wherein j is a positive integer greater than 1.
Preferably, the preset first function relationship is:
the ld isj1' to replace the jth load in the first load sequence1Individual load value ldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
Preferably, the cloud management according to the second load sequence includes:
calculating a load increment according to the second load sequence and a preset second functional relation;
and obtaining a management decision according to the load increment, and carrying out cloud management according to the obtained management decision.
Preferably, the second function relationship is:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient; the n is the number of load values in the second load sequence; y is a positive integer less than n;
obtaining a management decision according to the load increment, and performing cloud management according to the obtained management decision, wherein the method specifically comprises the following steps:
when aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine;
when in useThen s in the cloud system2Converting the cold state physical machine into a warm state physical machine;
when aincm<0 is large andthen s in the cloud system3Converting the thermal state physical machine into a temperature state physical machine;
when in useThen s in the cloud system4Converting the temperature state physical machine into a cold state physical machine;
wherein,
a is said1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; the ld ismIs the mth load value in the second load sequence;
the V ishotIs a cloud systemAn upper limit on the available hot state physical machines in the system;
the V iswarmThe upper limit of the hot state physical machine in the cloud system can be provided.
A second aspect of the present invention provides a cloud dynamic management apparatus, including:
the cloud system comprises a load information acquisition unit, a load information acquisition unit and a load information acquisition unit, wherein the load information acquisition unit is used for acquiring load information of the cloud system; the load information comprises a first load sequence;
the judging unit is used for judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
a replacement load value forming unit, configured to calculate, when an abnormal load value exists in the first load sequence, a replacement load value that replaces the abnormal load value according to a normal load value in the first load sequence and a preset first functional relationship;
a second load sequence forming unit, configured to replace the abnormal load value in the first load sequence with the replacement load value to form a second load sequence;
and the management unit is used for carrying out cloud management according to the second load sequence.
Further, the judging unit includes
A calculation module to calculate a positive rate of change RP of the first load sequencemAnd negative rate of change RNm；
The judging module is used for judging the abnormal load value, and the load value in the first load sequence which meets the following formula is the abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence;
the ld isj-1Is the j-1 th load value in the first load sequence;
the ld isj+1Is the j +1 th load value in the first load sequence;
the beta is a coefficient factor;
wherein j is a positive integer greater than 1.
Further, the preset first functional relationship is
The ld isj1' to replace the jth load in the first load sequence1Individual load value ldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
Further, the management unit includes:
the load increment calculation module is used for calculating the load increment according to the second load sequence and a preset second function relation;
and the management module is used for obtaining a management decision according to the load increment and carrying out cloud management according to the obtained management decision.
Further, the second function relationship is:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient; y is; the n is the number of load values in the second load sequence;
the management module is used as an aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine; when in useIn time, will be in the cloud system s2Converting the cold state physical machine into a warm state physical machine; when aincm<0 is large andin time, will be in the cloud system s3Converting the thermal state physical machine into a temperature state physical machine; when in useIn time, will be in the cloud system s4Converting the temperature state physical machine into a cold state physical machine;
wherein, the alpha is1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; the ld ismIs the mth load value in the second load sequence;
the V ishotAn upper limit for a hot state physical machine in the cloud system can be provided;
the V iswarmThe upper limit of the hot state physical machine in the cloud system can be provided.
The embodiment of the invention provides a cloud dynamic management method and device, wherein the cloud dynamic management method is based on a cloud system historical load value to carry out cloud management at the next moment, and in order to further optimize management, the abnormal load value of the historical load value is replaced, so that the management time delay is small, the control efficiency is high, the abnormal load value is eliminated, and the management control is more accurate.
Drawings
Fig. 1 is a schematic flow chart of a cloud dynamic management method according to a first embodiment of the present invention;
fig. 2 is a schematic structural diagram of a cloud dynamic management apparatus according to a second embodiment of the present invention.
Detailed Description
The present application is described in further detail below with reference to the drawings and examples of the specification.
The first embodiment:
the invention provides a cloud dynamic management method, as shown in fig. 1, the method comprises the following steps:
step S110: acquiring load information of a cloud system, wherein the load information comprises a first load sequence;
here, the first load sequence is formed in chronological order;
step S120: judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
step S130: when an abnormal load value exists in the first load sequence, calculating a replacement load value for replacing the abnormal load value according to a normal load value in the first load sequence and a preset first functional relation;
step S140: replacing abnormal load values in the first load sequence with the replacement load values to form a second load sequence;
step S150: and performing cloud management according to the second load sequence.
When the abnormal copy value does not exist in the first load sequence, cloud management can be directly performed according to the first load sequence.
The first load sequence formed in step S110 includes load values corresponding to a plurality of moments of the cloud system, and the load values are arranged according to a time sequence, and a specific physical meaning of the load values may be a ratio of resources in a busy state in the cloud system to total resources in the cloud system. Generally, the time difference between any two adjacent load values is a value period; the value period is from 50 milliseconds to 1000 milliseconds, and specific values can be set according to the load change condition of the cloud system and the management requirement, specifically as follows: 100. 200, 500, 700, or 800 milliseconds, etc.
The load value can be recorded as ldiI is the ldiIn the first load sequence, the higher the time, the smaller the i is. In particular said first load sequence may be denoted ldm-n+1,.....ldm-1,ldm. Generally, a cloud system starts to obtain load values according to a value period when the cloud system starts to operate, and when the trend change at the next moment needs to be predicted so as to perform resource management and scheduling of the cloud system, the load values of the latest n value periods are generally required to be obtained, so that the load values are obtained at the ldm-n+1,.....ldm-1,ldmIn the first load sequence, the ldmUsually the load value of the value cycle corresponding to the current time.
Specifically, how to obtain the load information of the cloud system, data acquisition may be performed by a data acquisition device or other devices, or other detection devices perform statistics on busy and idle states of each resource in the cloud system, or other methods, so as to obtain the load value. In a specific implementation process, the load information may include not only the first load sequence as described above, but also other information related to the operation of the cloud system. The other information may include information such as a time taken for the first load sequence and a load condition analysis of the first load sequence.
When the cloud system operates normally, the change of the busy and idle state of resources in the cloud system should have continuity in time. If a point in the continuous load values has a great change from the load values before and after the point, the change appears very abrupt and usually indicates that an abnormality occurs, and the corresponding load value is the abnormal load value.
There are various reasons for the abnormal load value of the cloud system, which can be roughly divided into two types:
the first is system factors, such as message exception delay, connection bandwidth change, computation resource conflict, etc.;
the second is a non-system factor, such as a system, software or hardware failure, a message loss, and the like.
The overall change rule of the first load sequence in step S120 is a change continuity of the load value according to time, and may be embodied as an increasing trend, a decreasing trend, or one or more various fluctuations within a certain range, where the fluctuations are, for example, increasing first and then decreasing or decreasing first and then increasing. The implementation method of the step S120 also has various methods, specifically, as a drawing method:
firstly, forming a first load sequence waveform diagram in a coordinate system with a vertical coordinate as a load value and a horizontal coordinate as a load value serial number;
secondly, comparing the variation waveform with a preset variation waveform range diagram to find out a mutation point, wherein the mutation point is an abnormal load value deviating from the variation rule of the first load sequence; or finding out the abrupt load value by a direct observation method; or calculating the slope formed by any two adjacent points in the coordinate system, and comparing the slope with a preset threshold, wherein if the slope between a certain load value and two adjacent load values is greater than the threshold, the corresponding load value is considered as an abnormal load value.
There are various specific methods for obtaining the abnormal load value, and details are not repeated here.
In step S130, a replacement load value of the abnormal load value is obtained through calculation, so that the formed second load sequence better conforms to the actual load variation of the cloud system. The method for obtaining the replacement load value is various, specifically, if two normal load values before and after the abnormal load value are summed, the average value is obtained, the average value can be calculated by correction factors such as coefficient factors according to needs, and the method is also various in a specific implementation process, and is not described in detail herein.
In the step S140, replacing the abnormal load value with the replacement load value calculated in the step S130 to obtain a second load sequence, where the second load sequence eliminates the abnormal load value in the first load sequence and retains a normal load value, so as to more closely reflect resources consumed by the operation of the cloud system, and perform cloud management of the cloud system according to the second load sequence, so that resource scheduling is more reasonable, and resource configuration more meets requirements.
In summary, according to the cloud dynamic management method of this embodiment, first, the change trend of resource scheduling acquired by the cloud system is predicted according to the historical load value, and cloud scheduling is performed according to the prediction, so that resource scheduling can be performed in time, and the conditions of load surge and drastic reduction can be quickly dealt with; secondly, the abnormal load value of the load value is considered, and the corresponding replacement load value is adopted to replace the abnormal load value, so that the problem of low cloud management precision caused by the abnormal load value is avoided during cloud management, and the resource allocation of cloud management is further optimized.
There are many ways to obtain the abnormal load value in the first load sequence in step S120, and a method that is simpler, faster, and more accurate than the methods such as the graph drawing method is provided below, which specifically includes the following steps:
step 1: calculating a positive rate of change R of the first load sequencePmAnd negative rate of change RNm；
The ld isvIs the v-th load value in the first load sequence; the ld isuIs the u-th load value in the first load sequence; the n is the number of load values in the first load sequence;
step 2: the load value which meets the following relation in the first load sequence is an abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence; the ld isj-1Is the j-1 th load value in the first load sequence; the ld isj+1Is the j +1 th load value in the first load sequence;
the beta is a coefficient factor; typically β is greater than 1;
wherein j is a positive integer greater than 1.
Repeated experiments prove that the method for acquiring the abnormal load value deviating from the overall change rule of the first load sequence in the first load sequence has the advantages of simplicity, convenience, rapidness and high accuracy.
As a further improvement of this embodiment, this embodiment further provides a preferred preset first functional relationship for calculating the replacement load value, where the preset first functional relationship is:
the ld isj1' to replace the jth load in the first load sequence1A load valueldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
By adopting the preset first functional relation, the calculated replacement load value can well follow the change rule of the load value and can well approach the actual load value at the corresponding moment, so that the resource optimization management and scheduling which are further beneficial to cloud management are realized.
Further, this embodiment embodies the cloud management according to the second load sequence:
step A: calculating a load increment according to the second load sequence and a preset second functional relation;
and B: and obtaining a management decision according to the load increment, and carrying out cloud management according to the obtained management decision.
The load increment may be a positive load increment or a negative load increment. In a specific application process, when an abnormal load value does not exist in the first load sequence, a load increment can be calculated according to the first load sequence and a preset second function relation; and then carrying out cloud management according to the load increment. Here, cloud management is management of cloud resources in a cloud system.
Cloud systems can be classified into three categories according to service types: infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS).
The IaaS mode cloud system virtualizes physical resources into a resource pool through virtualization technologies such as system virtualization, multiprocessor virtualization, memory virtualization, I/O virtualization and the like, and the resources are managed and scheduled by the IaaS system in a unified manner. The IaaS system provides computing resources and storage resources to users.
The IaaS system has the advantages that: the resources are flexibly and dynamically managed, characteristics of imbalance, uncertainty, sparsity and the like of application requests are fully utilized, and fewer computing resources (such as hosts, storage and network bandwidth) can support large-scale application requirements, so that benefit maximization is realized.
In order to balance performance and energy consumption, the IaaS system sets a Physical Machine (PM) state owned by the system to a cold (cold), warm (warm) and hot (hot) state, respectively. The energy consumption of the hot state PM is highest, the energy consumption of the warm state PM is second, and the energy consumption of the cold state PM is lowest. While the task execution speed of the hot state PM is highest (full speed operation), the warm state PM is second (less than full speed operation), and the operation speed of the cold state PM is zero.
As can be seen from the above description, a physical host (PM) in the IaaS cloud system is a center for performing service operation and demand response, a load condition of the system is directly determined by load conditions of all PMs managed by the system, and a quality of a PM load management policy directly determines an operation efficiency of the cloud system. On this background, how to dynamically maintain the state of the PM according to the task load of the IaaS cloud system, thereby realizing the win-win of energy saving and performance, becomes a hotspot and difficulty for research.
The cloud dynamic management method described in this embodiment is particularly suitable for an IaaS system among the three systems. For the IaaS system, the cloud dynamic management method described in this embodiment proposes the optimization of the second functional relationship, which is specifically as follows:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient, is usually a positive number less than 1, and is preferably 0.5; y is; the n is the total number of load values of the second load sequence.
The cloud management according to the load increment specifically includes:
when aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine;
when in useThen s in the cloud system2Converting the cold state physical machine into a warm state physical machine;
when aincm<0 is large andthen s in the cloud system3Converting the thermal state physical machine into a temperature state physical machine;
when in useThen s in the cloud system4Physical machine for converting temperature state into cold state
Wherein, the alpha is1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; typically the said ldmx is less than 100%, typically anywhere between 90% and 99%; the value of the second threshold is not more than 0.1, and preferably is any value between 0.03 and 0.06, such as 0.05, according to the common setting of the existing IaaS system.
The ld ismIs the mth load value in the second load sequence;
the V ishotAn upper limit for available thermal state physical machines in the cloud system;
the V iswarmAn upper limit for the available thermal state physical machines in the cloud system.
By adopting the cloud dynamic management method of the embodiment to perform cloud management on the IaaS system, compared with the existing method that the change of three states of physical machine heat, cold and temperature is performed after the increase or decrease of the load is specifically checked, the delay is smaller; meanwhile, the method also changes the switching among the hot state, the cold state and the warm state of the physical machine according to the preset proportion and quantity in the traditional IaaS system, the system resources consumed by the switching are obviously smaller, and the operation is obviously more precise and more in line with the specific actual needs.
In this embodiment, when each state of the physical machine is converted, the method further includes counting the number of the physical machines in the current state to be converted, so as to determine whether the state conversion of the physical machine can be completed, and implement dynamic management of the cloud system. In particular, if necessary, s2The individual cold state physical machine is converted into the warm state physical machine, and before the conversion is carried out, whether s exists needs to be counted in advance2The cold state physical machine can be used for conversionIf the judgment result is yes, the physical machine state can be directly converted, otherwise, the abnormal phenomena such as excessive resource consumption and the like need to be further judged, and the abnormal processing step is carried out.
In a specific implementation process, the cloud dynamic management method described in this embodiment is a real-time management method or a periodic management method, that is, the steps S110 to S150 are repeatedly executed, and specifically, the step S110 may be immediately executed after the step S150 is executed; or after step S150 is executed, waiting for a cycle period and then entering step S110; specifically, which method is selected may be set according to the load change speed of the cloud system, the amount of cloud system resources, and the requirement for the fineness of cloud management.
In summary, the present embodiment provides a different past cloud dynamic management method, which performs dynamic management on cloud resources of a current cloud system by obtaining a historical load value of a cloud system, where the obtaining of the load value is continuous in time, so that the obtaining speed is fast, and the historical load value is used to determine cloud resource scheduling at a next time, and it is obvious that a management scheduling decision is generated before the next time, so that the management scheduling decision is executed directly according to the formed management scheduling decision, and a current load condition is obtained in advance for an existing management time, and then the management is performed, so that it is obvious that the management efficiency is higher, the time delay is smaller, and thus it is possible to better cope with drastic increase and drastic decrease of load; meanwhile, in order to further optimize cloud dynamic management, abnormal load values in historical load values are removed, and the load values are replaced by replacement load values which accord with a first load sequence change rule, so that the accuracy of management and scheduling of the transport resources is improved.
Second embodiment:
the present embodiment provides a cloud dynamic management apparatus, which may be used in the cloud dynamic management method described in the first embodiment, as shown in fig. 2, the apparatus includes:
a load information obtaining unit 210, configured to obtain load information of the cloud system; the load information comprises a first load sequence;
a determining unit 220, configured to determine whether an abnormal load value deviating from an overall change rule of the first load sequence exists in the first load sequence;
a replacement load value forming unit 230, configured to calculate, when an abnormal load value exists in the first load sequence, a replacement load value that replaces the abnormal load value according to a normal load value in the first load sequence and a preset first functional relationship;
a second load sequence forming unit 240, configured to replace the abnormal load value in the first load sequence with the replacement load value to form a second load sequence;
a management unit 250, configured to perform cloud management according to the second load sequence.
The specific structure of the load information obtaining unit 210 may be a data collector or a detection and statistics device for idle and busy states of cloud system resources, and the like, and is used to obtain load information, where the load information at least includes a first load sequence, and load values in the first load sequence are usually arranged according to a time sequence, and the load sequence is composed of load values at the first n times of the current time, and the specific composition is, for example, ldm-n+1,.....ldm-1,ldm。
The specific structure of the replacement load value star unit 230 may be any device having a calculation function, such as a calculator.
The management unit 250 may be any management device or management module in the prior art.
The embodiment provides a cloud dynamic management device in different prior arts, which predicts the load change at the next moment according to the historical load value of a cloud system to perform cloud management, and compared with a management device which obtains the current load condition and performs cloud management, the cloud dynamic management device has the advantages that the management time delay is short, and the drastic change of the load can be rapidly coped with; meanwhile, when load change at the next moment is predicted, an abnormal load value in historical load values is provided, and the problems of improper management and insufficient optimization of resource allocation caused by the abnormal load value are solved.
As a further improvement of this embodiment, the determining unit includes a calculating module and a determining module.
The calculation module is used for calculating the positive change rate RP of the first load sequencemAnd negative rate of change RNm；
The judging module is used for judging the abnormal load value, and the load value in the first load sequence which meets the following formula is the abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence; the ld isj-1Is the j-1 th load value in the first load sequence; the ld isj+1Is the j +1 th load value in the first load sequence; the beta is a coefficient factor; usually, β is a predetermined number not less than 1.
Wherein j is a positive integer greater than 1.
In particular, the calculation module may be configured to perform the positive rate of change RP according to the following formulamAnd negative rate of change RNmCalculating (1);
the ld isvIs the v-th load value in the first load sequence; the ld isuIs the u-th load value in the first load sequence; and n is the number of load values in the first load sequence.
The calculation module may be any functional module capable of completing the calculation, such as a calculator; the judging unit may be a comparator or the like for implementing the above logical relationship comparison and judgment.
The specific structure of the judging unit is various, a structure which is fast, simple and convenient to implement and high in accuracy is provided, and other structures are not described in further detail herein.
As a further improvement of this embodiment, there are a plurality of specific expressions of the preset first functional relationship, and one of the following provides a simple relational expression with a small calculation amount, specifically as follows:
the ld isj1' to replace the jth load in the first load sequence1Individual load value ldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
In addition, this embodiment further provides a preferred management unit, where the management unit includes:
the load increment calculation module is used for calculating the load increment according to the second load sequence and a preset second function relation;
and the management module is used for obtaining a management decision according to the load increment and carrying out cloud management according to the obtained management decision.
The specific second function relationship is as follows:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient; y is a positive integer less than n;
the management module is used as an aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine; when in useThen s in the cloud system2Converting the cold state physical machine into a warm state physical machine; when aincm<0 is large andthen s in the cloud system3Converting the thermal state physical machine into a temperature state physical machine; when in useThen s in the cloud system4Converting the temperature state physical machine into a cold state physical machine;
wherein, theα1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; the ld ismIs the mth load value in the second load sequence;
the V ishotAn upper limit for a hot state physical machine in the cloud system can be provided;
the V iswarmThe upper limit of the hot state physical machine in the cloud system can be provided.
The management unit in this embodiment is mainly provided for a cloud management mode in which a physical machine is switched between a "hot" state and a "cold" state in an IaaS system, and the management unit is used for performing cloud management on the physical machine in the IaaS system, and after a current load is successively obtained, the three states are converted according to a preset proportion or quantity, so that the time delay is small, and the management has the advantages of small resource consumption and the like.
The cloud dynamic management device described in this embodiment may be a management device integrated in one or more servers of the cloud system, or may be a management device independent from the cloud system, and is connected to the cloud system in a wired or wireless connection manner to acquire load information and output management information.
The cloud dynamic management apparatus may also be another embodiment of a device including a processor, at least one communication interface, a storage medium, and a bus. The medium has stored therein software or a program. The processor can be a central processing unit CPU, a single chip microcomputer MCU, a digital signal processor DSP, a programmable logic array and the like. The bus connects the processor, the storage medium, and the communication interface. When the processor runs the software or the program, at least the following functions can be completed:
acquiring load information of a cloud system; the load information comprises a first load sequence formed according to the time sequence;
judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
if the abnormal load value exists, calculating a replacement load value for replacing the abnormal load value according to the normal load value in the first load sequence and a preset first function relation;
replacing abnormal load values in the first load sequence with the replacement load values to form a second load sequence;
and performing cloud management according to the second load sequence.
In summary, the present embodiment provides a cloud dynamic management apparatus, a method for performing cloud management at the next time based on a historical load value of a cloud system, and for further optimization management, and performing replacement processing on an abnormal load value of the historical load value, so that management time delay is small, control efficiency is high, the abnormal load value is eliminated, and management control is more accurate.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention.
Claims (10)
1. A cloud dynamic management method, the method comprising:
acquiring load information of a cloud system, wherein the load information comprises a first load sequence;
judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
if the abnormal load value exists, calculating a replacement load value for replacing the abnormal load value according to the normal load value in the first load sequence and a preset first function relation;
replacing abnormal load values in the first load sequence with the replacement load values to form a second load sequence;
and performing cloud management according to the second load sequence.
2. The cloud dynamic management method according to claim 1, wherein the determining whether the abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence comprises:
calculating a positive rate of change RP of the first load sequencemAnd negative rate of change RNm；
The load value satisfying the following relation in the first load sequence is an abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence;
the ld isj-1Is the j-1 th load value in the first load sequence;
the ld isj+1Is the j +1 th load value in the first load sequence;
the beta is a coefficient factor;
wherein j is a positive integer greater than 1.
3. The cloud dynamic management method according to claim 1 or 2, wherein the preset first function relationship is:
the ld isj1' to replace the jth load in the first load sequence1Individual load value ldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
4. The cloud dynamic management method according to claim 1 or 2, wherein the cloud management according to the second load sequence comprises:
calculating a load increment according to the second load sequence and a preset second functional relation;
and obtaining a management decision according to the load increment, and carrying out cloud management according to the obtained management decision.
5. The dynamic cloud management method of claim 4, wherein the second function relationship is:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient; n is the number of load values in the second load sequenceCounting; y is a positive integer less than n;
obtaining a management decision according to the load increment, and performing cloud management according to the obtained management decision, wherein the method specifically comprises the following steps:
when aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine;
when in useThen s in the cloud system2Converting the cold state physical machine into a warm state physical machine;
when aincm<0 is large andthen s in the cloud system3Converting the thermal state physical machine into a temperature state physical machine;
when in useThen s in the cloud system4Converting the temperature state physical machine into a cold state physical machine;
wherein,
a is said1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; the ld ismIs the mth load value in the second load sequence;
the V ishotAn upper limit for a hot state physical machine in the cloud system can be provided;
the V iswarmThe upper limit of the hot state physical machine in the cloud system can be provided.
6. An apparatus for cloud dynamics management, the apparatus comprising:
the cloud system comprises a load information acquisition unit, a load information acquisition unit and a load information acquisition unit, wherein the load information acquisition unit is used for acquiring load information of the cloud system; the load information comprises a first load sequence;
the judging unit is used for judging whether an abnormal load value deviating from the overall change rule of the first load sequence exists in the first load sequence;
a replacement load value forming unit, configured to calculate, when an abnormal load value exists in the first load sequence, a replacement load value that replaces the abnormal load value according to a normal load value in the first load sequence and a preset first functional relationship;
a second load sequence forming unit, configured to replace the abnormal load value in the first load sequence with the replacement load value to form a second load sequence;
and the management unit is used for carrying out cloud management according to the second load sequence.
7. The cloud dynamic management device according to claim 6, wherein the judging unit comprises
A calculation module to calculate a positive rate of change RP of the first load sequencemAnd negative rate of change RNm；
The judging module is used for judging the abnormal load value, and the load value in the first load sequence which meets the following formula is the abnormal load value;
ldj-ldj-1>β×PRmand ldj+1-ldj<β×NRm
Or
ldj-ldj-1<β×NRmAnd ldj+1-ldj>β×PRm
The ld isjIs the jth load value in the first load sequence;
the ld isj-1Is the j-1 th load value in the first load sequence;
the ld isj+1Is the j +1 th load value in the first load sequence;
the beta is a coefficient factor;
wherein j is a positive integer greater than 1.
8. The cloud dynamic management device according to claim 6 or 7, wherein the preset first functional relationship is
The ld isj1' to replace the jth load in the first load sequence1Individual load value ldj1And the replacement load value of, and the ldj1Is an abnormal load value;
the ld iskIs the kth load value in the first load sequence and is the normal load value;
the ld islIs the l-th load value in the first load sequence and is the normal load value.
9. The cloud dynamic management apparatus according to claim 8, wherein the management unit includes:
the load increment calculation module is used for calculating the load increment according to the second load sequence and a preset second function relation;
and the management module is used for obtaining a management decision according to the load increment and carrying out cloud management according to the obtained management decision.
10. The cloud dynamics management apparatus of claim 9, wherein the second function relationship is:
wherein said aincmIs the load increment; the ld ism-x+1Is the m-x +1 load value in the second load sequence; the ld ism-xIs the m-x load value in the second load sequence; gamma is an attenuation coefficient; y is; the n is the number of load values in the second load sequence;
the management module is used as an aincm>0 is large andthen s in the cloud system1Converting the temperature state physical machine into a thermal state physical machine; when in useIn time, will be in the cloud system s2Converting the cold state physical machine into a warm state physical machine; when aincm<0 is large andin time, will be in the cloud system s3Converting the thermal state physical machine into a temperature state physical machine; when in useIn time, will be in the cloud system s4Converting the temperature state physical machine into a cold state physical machine;
wherein, the alpha is1Is a first threshold value; a is said2Is a second threshold value, and 0<α1<α2；
The ldmx is a preset maximum load value in the cloud system; the ld ismIs the mth load value in the second load sequence;
the V ishotIn a cloud systemAn upper bound on the hot state physical machine can be provided;
the V iswarmThe upper limit of the hot state physical machine in the cloud system can be provided.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410069284.XA CN104881271B (en) | 2014-02-27 | 2014-02-27 | Cloud dynamic management approach and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410069284.XA CN104881271B (en) | 2014-02-27 | 2014-02-27 | Cloud dynamic management approach and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104881271A true CN104881271A (en) | 2015-09-02 |
| CN104881271B CN104881271B (en) | 2018-11-23 |
Family
ID=53948780
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410069284.XA Active CN104881271B (en) | 2014-02-27 | 2014-02-27 | Cloud dynamic management approach and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104881271B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103442087A (en) * | 2013-09-12 | 2013-12-11 | 重庆大学 | Web service system access volume control device and method based on response time trend analysis |
| CN103607459A (en) * | 2013-11-21 | 2014-02-26 | 东北大学 | Dynamic resource monitoring and scheduling method of cloud computing platform IaaS layer |
-
2014
- 2014-02-27 CN CN201410069284.XA patent/CN104881271B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103442087A (en) * | 2013-09-12 | 2013-12-11 | 重庆大学 | Web service system access volume control device and method based on response time trend analysis |
| CN103607459A (en) * | 2013-11-21 | 2014-02-26 | 东北大学 | Dynamic resource monitoring and scheduling method of cloud computing platform IaaS layer |
Non-Patent Citations (1)
| Title |
|---|
| 杨伟 等: "基于时间序列的服务器负载预测", 《计算机工程》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104881271B (en) | 2018-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103605567B (en) | Cloud computing task scheduling method facing real-time demand change | |
| US20190324819A1 (en) | Distributed-system task assignment method and apparatus | |
| CN104915407B (en) | A kind of resource regulating method based under Hadoop multi-job environment | |
| WO2017167025A1 (en) | Method and device for realizing task scheduling, and computer storage medium | |
| CN103327072B (en) | Cluster load balancing method and system | |
| CN109976917B (en) | Load scheduling method, device, load scheduler, storage medium and system | |
| CN104102543B (en) | The method and apparatus of adjustment of load in a kind of cloud computing environment | |
| CN108845874B (en) | Dynamic resource allocation method and server | |
| WO2016082370A1 (en) | Distributed node intra-group task scheduling method and system | |
| CN104035542B (en) | A kind of virtual machine migration method and system calculated with energy consumption for cooling balance | |
| CN109799956B (en) | Memory controller and IO request processing method | |
| CN103699433B (en) | One kind dynamically adjusts number of tasks purpose method and system in Hadoop platform | |
| US20170010919A1 (en) | Dynamic weight accumulation for fair allocation of resources in a scheduler hierarchy | |
| US20190138354A1 (en) | Method for scheduling jobs with idle resources | |
| CN104021040A (en) | Cloud computing associated task scheduling method and device based on time constraint | |
| CN104881322A (en) | Method and device for dispatching cluster resource based on packing model | |
| CN103425536A (en) | Test resource management method oriented towards distributed system performance tests | |
| Tao et al. | Load feedback-based resource scheduling and dynamic migration-based data locality for virtual hadoop clusters in openstack-based clouds | |
| CN114780244A (en) | Container cloud resource elastic allocation method and device, computer equipment and medium | |
| CN104571931A (en) | I/O (input/output) request combination scheduling system and method based on system resources | |
| CN103440113A (en) | Disk IO (Input/output) resource allocation method and device | |
| CN110825212B (en) | Energy-saving scheduling method and device and computer storage medium | |
| CN106708624B (en) | Self-adaptive adjustment method for multi-working-domain computing resources | |
| CN104881271B (en) | Cloud dynamic management approach and device | |
| CN108073449B (en) | Dynamic virtual machine placement method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |