CN103383653B - Cloud resource management and dispatching method and system - Google Patents
Cloud resource management and dispatching method and system Download PDFInfo
- Publication number
- CN103383653B CN103383653B CN201210135357.1A CN201210135357A CN103383653B CN 103383653 B CN103383653 B CN 103383653B CN 201210135357 A CN201210135357 A CN 201210135357A CN 103383653 B CN103383653 B CN 103383653B
- Authority
- CN
- China
- Prior art keywords
- resource
- interval
- attribute
- joint
- index
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides cloud resource management and dispatching method.The codomain of each attribute is wherein divided into multiple attributes intervals by resource, and be the corresponding logic interval call number of the interval distribution of each attribute, and by each resource impact to the interval index of joint determined by the logic interval call number belonging to each property value as the resource.When the resource request of user is received, the logic interval call number belonging to the constraint upper limit and constraint lower limit of each attribute of requested resource is determined, be met the interval index of multiple joints of the resource request;Then, the interval index of one joint of selection from the multiple joint interval index, and select a resource to be supplied to the user from the corresponding logical resource pond of the interval index of selected joint.The method can meet user resources request and the demand of high-frequency resource updates of highly simultaneous access in cloud computing environment.
Description
Technical field
The present invention relates to the high concurrent of vast resources under distributed resource management and scheduling, more particularly to cloud computing environment
Scheduling and high-frequency update.
Background technology
Efficiently scheduling resource is the distributed resource management system key issue to be solved in large scale network.
Under distributed environment, Internet resources(Including computing resource, data resource, software resource and Service Source etc.)Scale it is rapid
Expansion, these resources dispersion each website in a network, dynamic is strong, and work is cooperateed with heterogeneous platform by network connection
Make.Support that large-scale application is different using scattered resource from grid computing, cloud computing by various Internet resources concentrate get up into
Row unified management and dispatch, and by network with demand, easy extension way various services are provided.Resource under cloud computing environment
Scheduling must face the access of vast resources information, the user resources request of high concurrent and ample resources and update the pressure for bringing
Deng.
Efficient scheduling resource must is fulfilled under cloud computing environment:(1)Efficient read-write vast resources information;(2)It is quick from
The resource or resource set for meeting user's request are navigated in vast resources;(3)Alleviate ample resources and update the pressure brought to system
Power.
In existing cloud computing environment, mainly there are following several resource dispatching strategies:Centralized resources scheduling, hierarchy type money
Source scheduling, the scheduling of resource based on P2P routes and the scheduling of resource based on DHTs etc..Centralized resources scheduling passes through center service
Device unified managing resource and user's request, and concentrate matching user's request and available resources carry out resource allocation.Based on P2P routes
Scheduling strategy scheduling of resource is carried out by the strategy for being matched and being route in the whole network domain.Scheduling of resource plan based on DHTs
It is each resource allocation unique index slightly using distributed hashtable, and according to index by resource deployment to specific server
On.Wherein, the resource management system autgmentability based on centralization and hierarchy type is poor, it is difficult to meet the need of new large-scale application
Ask.And the resource regulating method for being based on P2P routes is difficult to dispatch suitable resource in the whole network domain, and in big complications
Under, Internet traffic increases severely.Resource regulating method based on DHTs passes through distributed hash(Hash)Policy deployment and positioning are provided
Source, which solve vast resources storage and quick locating resource problem, but taken when resource selection is carried out it is larger, and
The expense of the change of logic that resource updates bring index and resource migration is also very big, it is difficult to process highly simultaneous access user's request and
High-frequency resource updates.
The content of the invention
Therefore, it is an object of the invention to overcome the defect of above-mentioned prior art, there is provided a kind of cloud resource regulating method, energy
Process the user's request and high-frequency resource updates of highly simultaneous access.
In order to realize foregoing invention purpose, following technical proposal is employed:
On the one hand, the invention provides a kind of cloud method for managing resource, methods described includes:
Step 1)All available resources are counted, the upper and lower bound of each attribute of resource is obtained, to determine
The codomain of each attribute;
Step 2)The codomain of each attribute is divided into multiple attributes intervals, and is patrolled for the interval distribution of each attribute is corresponding
Collect interval call number;
Step 3)For each resource, the logic interval call number belonging to each property value of the resource is determined, to obtain
Corresponding to the interval index of joint of the resource;
Step 4)Multiple logical resource ponds are divided resources into according to the interval index of joint, corresponding to the interval rope of same joint
The all resources drawn belong to a logical resource pond.
In above-mentioned technical proposal, may also include for each attribute interval, the user in statistics a period of time inquires about the category
Property interval resource request quantity the step of;
In above-mentioned technical proposal, the step of may also include based on resource request quantity to adjust interval division, under it includes
Row operation:
Step 51), following operation is performed for each attribute of resource:
(511)Scan the user's request quantity in the interval of the attribute partition successively, often scan RN request record just
One boundary value of setting;
(512)It is interval that the codomain of the attribute is divided into multiple attributes by the boundary value based on new settings come again;
(513)It is the corresponding logic interval call number of the interval distribution of each attribute obtained after repartitioning;
Step 52), for each resource, determine the new logic interval call number belonging to each property value of the resource,
To obtain corresponding to the interval index of new joint of the resource;
Step 53), multiple logical resource ponds are divided resources into according to the interval index of new joint, corresponding to same joint
All resources of interval index belong to a logical resource pond.
In above-mentioned technical proposal, in step 512) may also include it is following amendment attribute siding-to-siding block length the step of:
The average area length ALen in multiple attributes interval that calculating is obtained after repartitioning;
Each attribute after scanning is repartitioned successively is interval, performs following operation:
If ith attribute length of an interval degree is less than α * Alen, and the interval length of i+1 attribute is less than α * Alen,
Then merge the i-th attribute interval interval with i+1 attribute, wherein 0<α<1;
If ith attribute length of an interval degree is less than α * Alen, and the interval length of i+1 attribute is more than α * Alen,
Ith attribute length of an interval degree is then extended to the corresponding part of length reduction in α * ALen and i+1 attribute interval;
If ith attribute length of an interval degree is more than α * Alen, the siding-to-siding block length keeps constant.
In above-mentioned technical proposal, can keep combining interval index and its corresponding using key-value storage organizations
The Resources list.
In above-mentioned technical proposal, resource can be virtual machine, physical node, cluster or data center.
Another aspect, the invention provides a kind of cloud resource regulating method, methods described includes:
Step 1, receives the resource request from user;
Step 2, for each attribute of requested resource, extracts under the constraint upper limit and constraint of each attribute respectively
Limit;
Step 3, determines the logic interval rope belonging to the constraint lower limit and constraint lower limit of each attribute of requested resource
Quotation marks, are met the interval index of multiple joints of the resource request;
Step 4, the interval index of one joint of selection from the multiple joint interval index, and from selected association area
Between index and select a resource to be supplied to the user in corresponding logical resource pond.
In above-mentioned technical proposal, the user resources that may also include in statistics a period of time for the interval index of each joint please
The step of seeking quantity.
In above-mentioned technical proposal, the step 4 can select access temperature minimum from the interval index of the multiple joint
The interval index of joint, and select a resource to provide from the corresponding logical resource pond of the interval index of selected joint
The user is given, wherein it is the user resources number of requests in the unit time for the interval index of joint to access temperature.
In above-mentioned technical proposal, the step of monitoring and update resource status can also be included, and when the property value of resource
Need to update the state of the resource when interval more than attribute where it.
Another aspect, the invention provides a kind of cloud resource scheduling system, the system includes:
Module for receiving the resource request from user;
For each attribute for requested resource, the constraint upper limit and constraint lower limit of each attribute are extracted respectively
Module;
For determining the constraint upper limit of each attribute of requested resource and constraining the interval index of logic belonging to lower limit
Number, it is met the module of the interval index of multiple joints of the resource request;
Combine interval index for the selection one from the multiple joint interval index, and it is interval from selected joint
A resource is selected to be supplied to the module of the user in the corresponding logical resource pond of index.
In said system, the module for monitoring and updating resource status can also be included.
Compared with prior art, the advantage of the invention is that:
Using the dispatching method based on interval division, the money for meeting user's request can be navigated to rapidly from vast resources
Source, and the request of user is balancedly distributed in each resource demarcation interval as far as possible, to improve the efficiency of scheduling of resource.Work as resource
When state changes, using based on interval resource updates pattern, substantially reduce frequently resource updates and give resource management system
The pressure that regiment commander comes.
Brief description of the drawings
Embodiments of the present invention is further illustrated referring to the drawings, wherein:
Fig. 1 is the cloud resource environment schematic diagram for implementing the embodiment of the present invention;
Fig. 2 is the cloud scheduler works process schematic according to the embodiment of the present invention;
Fig. 3 is the dynamic adjustment interval division schematic diagram according to the embodiment of the present invention;
Fig. 4 is to dispatch system architecture schematic diagram according to the cloud based on interval division of the embodiment of the present invention.
Fig. 5 is traditional resource updates schematic diagram;
Fig. 6 is the resource updates schematic diagram according to the embodiment of the present invention.
Specific embodiment
In order that the purpose of the present invention, technical scheme and advantage become more apparent, below in conjunction with accompanying drawing by specific real
The present invention is described in more detail to apply example.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention,
It is not intended to limit the present invention.
Fig. 1 gives the cloud resource environment schematic diagram for implementing the embodiment of the present invention.The environment is by several data
The heart is constituted, the cluster that each data center is made up of comprising several physical node, several is run on each physical node empty
Plan machine.And resource can be virtual machine, or physical node or cluster or data center.As shown in figure 1, various resources are
Organized with layer architecture, user can according to demand select the resource granularity for needing to dispatch, can be with selection scheduling list
Individual resources of virtual machine, it is also possible to dispatch physical node, cluster or data center etc..Wherein resources of virtual machine is the base of scheduling of resource
Our unit.For each resource R, relate generally to three and calculate attribute of performance and its position attribution, wherein three calculating performances
Attribute is respectively:Idle CPU(Represented with check figure), free memory(GB)With free hard disk (GB).So, resource R can be simple
Be expressed as R={ [FCPU, FMemory, FDisk], [DC, Cluster, PN] }, wherein FCPU, FMemory, FDisk difference
It is idle CPU(Represented with check figure), free memory(GB)With free hard disk (GB), DC represents the data center belonging to R,
Cluster represents the cluster belonging to R, and PN represents the physical node belonging to R.In the layer architecture shown in Fig. 1, every layer of resource
Amount can be obtained by being integrated to its next layer stock number, such as each physical node, can be by integrating
All resources of virtual machine disposed thereon and obtain the calculating attribute of performance of the physical node resource.
In one embodiment, the cloud method for managing resource based on interval division is given.The method is first according to experience
Value provides resource(Including virtual machine, physical node, cluster, data center etc.)Initial interval division, then according to the interval
Division is mapped resources in corresponding logical resource pond.More specifically, the method is mainly included the following steps that:
All available resources are counted by step 1, obtain the upper and lower bound of each attribute of resource, to determine
The codomain of each attribute.
For example, determining each statistic unit(For example, virtual machine, physical node, cluster and data center)Each attribute(System
Meter free time CPU, free memory and free hard disk)Upper and lower bound.By taking resources of virtual machine as an example, it is first determined each virtual machine
Each attribute of resource(For example, free time CPU, free memory, free hard disk etc.)Upper and lower bound, and then obtain each attribute
Codomain.CPU codomains are designated as RRCPU=[CPUmin, CPUmax], wherein CPUmin are the lower limit of CPU, and CPUmax is upper for CPU's
Limit, the codomain of the free memory of resource R is designated as RRMemThe codomain of=[MEMmin, MEMmax] and free hard disk is designated as RRDisk=
[DISKmin,DISKmax]。
Step 2, is divided into multiple attributes intervals, and patrol for the interval distribution of each attribute is corresponding by the codomain of each attribute
Collect interval call number.
First, the codomain of each attribute is divided into multiple attributes intervals.For example, initially can be using the average side for dividing
Method carries out subregion to the codomain of each attribute.Will introduce below and each interval is adjusted to optimize the side of interval division
Method.For example, initial resource partitioning is:CPU(Lower limit CPUmin=0, upper limit CPUmax=32)M subregion is divided equally into, it is interior
Deposit(Lower limit MEMmin=0G, upper limit MEMmax=16G,)It is divided equally into N number of subregion, hard disk(Lower limit DISKmin=0G, the upper limit
DISKmax=1000G)P subregion is divided equally into, is then had:
(1) for arbitrary CPUi, i ∈ [1, M-1] and i are integer, there is CPUi∈RRCPUAnd CPUi<CPUi+1, CPU0=
CPUmin,CPUM=CPUmax;Then TCPU={[CPUi,CPUi+1) | i ∈ [0, M-1] } constitute the idle CPU attributes whole district between
One differentiation of [CPUmin, CPUmax](I.e. one subregion, similarly hereinafter), BCPU={CPUi| i ∈ [0, M] } constitute CPU division side
Boundary gathers, using BCPUThe interval belonging to given CPU values can be calculated.
(2) for arbitrary Memj, j ∈ [1, N-1] and j are integer, there is Memj∈RRMemAnd Memj<Memj+1, Mem0=
MEMmin,MemN=MEMmax;Then TMEM={[Memj,Memj+1) | j ∈ [0, N-1] } constitute the free memory attribute whole district between
One differentiation of [MEMmin, MEMmax], BMEM={Memj| j ∈ [0, N] } constitute free memory division border set, profit
Use BMEMThe interval belonging to given CPU values can be calculated.
(3) for arbitrary Diskk, k ∈ [1, P-1] and k are integer, there is Diskk∈RRDiskAnd Diskk<Diskk+1,
Disk0=DISKmin,DiskP=DISKmax, then TDisk={[Diskk,Diskk+1) | k ∈ [0, P-1] } constitute free hard disk attribute
One differentiation of [DISKmin, DISKmax], B between the whole districtDISK={Diskk| k ∈ [0, P] } constitute drawing for free hard disk attribute
Divide border set, using BDISKThe interval belonging to given CPU values can be calculated.
Then, it is that corresponding logic interval call number is distributed in each attribute interval for being divided.For example for the logic area
Between call number i can refer to that the ith attribute of certain attribute is interval, wherein i is integer.
Step 3, for each resource, determines the logic interval call number belonging to each property value of the resource, to obtain
Corresponding to the interval index of joint of the resource.
On the interval index of joint that actually resource R is finally mapped to corresponding to it.Still it is with resources of virtual machine R
Example is illustrated.The n attribute A for each resources of virtual machine R1,A2,...,An, by mappingx
∈ [1, n] obtains call number I between n each self-corresponding logic area of attribute1,I2,...,In, wherein,Represent attribute AxJust
Gather on beginning interval division border(After being carried out averagely to each attribute codomain above, each interval border value set, such as BCPU、
BMEM、BDISK),Represent attribute AxValue, fx() represents and acts on attribute AxMapping ruler, it is then interval to n logic
Call number I1,I2,...,InCarry out mapping again and obtain the interval index Joint RangeIndex=F of the corresponding joints of resource R
(I1,I2,...,In), i.e. logical resource pond ID, F () they are to carry out map operation rule again to each logic index, are so passed through
Mapping twice obtains a JointRangeIndex of each resource ownership.
Below with three attributes of resource(Idle CPU, free memory and free hard disk)As a example by illustrate how to be reflected
Penetrate.Then for resource R=<FreeCPU,FreeCPUValue>,<FreeMemory,FreeMemValue>,<FreeDisk,
FreeDiskValue>, its three attributes are mapped according to respective rule respectively, then have:
I=fCPU(BCPU,FreeCPUValue),i∈[1,M];
j=fMEM(BMEM,FreeMemValue),j∈[1,N];
k=fDISK(BDISK,FreeDiskValue),k∈[1,P].
Wherein rule fCPUCan be described as follows:In the B of ascending orderCPUIn set, first is found not less than set-point
The position i of the value of FreeCPUValue, fMEM,fDISKDefinition be similar to.
Then, by Quadratic Map rule F () by three interval index i of the logic of attribute, j, k carry out Joint Mapping and obtain
The interval index JointRangeIndex=F (i, j, k) of joint, F () mapping ruler can be by the way of cascade (for example here:
With " _ " be connected), i.e., according to specific attribute sequentially(Here it is CPU->MEM->The order of DISK)By each logic interval call number
Combine, obtain combining interval index i_j_k, certain F () can also be defined as other rules, depending on the application.
Step 4, multiple logical resource ponds are divided resources into according to the interval index of joint, corresponding to the interval rope of same joint
The all resources drawn belong to a logical resource pond.Fig. 2 is referred in advance, as shown in Fig. 2 by above-mentioned interval division method, can
Multiple logical resource ponds are divided into by the available resource under cloud environment, each logical resource pond provides including one or more
Source, by the interval index of joint corresponding to each logical resource pond(It is referred to as logical resource pond ID)(Such as in Fig. 2 6_
4_3:{R100,R50... ..., 6_5_3:{R95,R120,R80... etc.), can quickly navigate to user's requested resource.
In yet another embodiment, it is possible to use key-value (key assignments) storage organizations are stored by logical resource pond ID
And its key-value pairs that corresponding resource collection is constituted(Combine interval index and its corresponding the Resources list), no longer deposit
The actual content of resource is stored up, do so can effectively utilize memory space, moreover it is possible to which support is concurrently inquired about to accelerate Resource orientation
Efficiency, while the problem of the resource content stored in also avoid frequent updating storage system.
In yet another embodiment of the present invention, there is provided a kind of cloud resource regulating method based on interval division.The party
Method is receiving the resource request from user(Sometimes be alternatively referred to as inquiry request, actually user's query resource please
Ask)When, resource request is parsed, for example, the type of detection resource(Dispatch granularity), such as virtual machine, physical node, collection
Group or data center.Then, for each attribute of requested resource, extract under the constraint upper limit and constraint of each attribute
Limit.Then, it is determined that the constraint upper limit of each attribute of requested resource and the logic interval call number belonging to constraint lower limit, obtain
To the interval index of multiple joints for meeting the resource request.An association area is selected in the last interval index from the multiple joint
Between index, and select a resource to be supplied to the use from the corresponding logical resource pond of the interval index of selected joint
Family.
Wherein, the resource request of user can be the form of the attribute constraint of similar [attribute, operator, value], wherein belonging to
Property(attribute)It is character string(string)Type, operator(operator)Have<,<=,>,>=, value(value)For whole
Number(integer), double types etc..
For example, when the attribute constraint of certain resource request is:
Q={<CPU,CPUOper,QCPUValue>,<Memory,MemOper,QMemValue>,<Disk,DiskOper,
QDiskValue>When, wherein, CPUOper, MemOper, DiskOper are the operator (operator) of each attribute, for example>、
<,≤>=etc..Each attribute of the request can be mapped according to mapping ruler same above.
i′=fCPU(BCPU,QCPUValue),i′∈[1,M];
j′=fMEM(BMEM,QMemValue),j′∈[1,N];
k′=fDISK(BDISK,QDiskValue),k′∈[1,P].
When<i CPUOper i′>&<j MemOper j′>&<k DiskOper k′>During establishment, then inquiry Q is matched
Certain resource R in the corresponding logical resource ponds of the interval index i_j_k of joint.
In yet another embodiment, in order to preferably support many attribute intervals resource request, can use as follows
Resource request:
Table 1
Wherein, the resource request of user mainly contains value attribute and position attribution, wherein<request_type>Represent
The resource type of request, can be single virtual machine(Virtual Machine)Or physical node (Physical Node) or collection
Group (Cluster) or data center (Data Center),<num_machine>The machine quantity that user needs is represented,<
compute_attr>Represent computation attribute, including idle CPU(Represented with check figure), free memory(GB)With free hard disk (GB),
I.e.<cpu_core>,<free_mem>,<free_disk>, one cluster of user's request is illustrated in table 1, it is desirable to 4 machines
Attribute feature is:Between 2 to 4, internal memory is less than or equal to 2G to check figure, and, 100 between 200G, user please for free hard disk capacity
Can also be only comprising one or more in three attributes when asking.In the pattern of this support interval query, user's request
Calculate attribute of performance to be represented with interval, for example user's request " machine of the internal memory not less than 2G ", then attribute<free_mem>Represent
Into<free_mem>2,Max</free_mem>, i.e., the constraints of each attribute with an interval upper and lower bound come table
Show.
When such resource request arrives, each attribute of resource request can be obtained by parsing first(For example, empty
Not busy CPU core number, free memory and free hard disk capacity)And attribute constraint(The constraint upper limit and constraint lower limit).For example, scanning user
The attribute A occurred in request1,A2..., An, for each attribute Ax(x∈[1,n])Constraint bound be designated asWhereinIt is designated as attribute AxConstraint lower limit,Represent attribute AxThe constraint upper limit.Then, first pass through
Logical mappingsWithWherein fx() is represented to attribute AxMapping rule
Then,Represent attribute AxAn interval differentiation, n lower limit logic for obtaining user's request indexAnd n
Individual upper limit logic index
Then, resource lookup is carried out, that is, searches the interval index of joint of the resource or resource set for meeting user's requirement.For
The resource request of certain user, can obtain the interval index of joint of many resources for meeting user's requirement, for example, being designated as:
Wherein F () is that the logic to each attribute described above indexes (Ix=(x ∈ [1, n])) map operation again rule
Then, F (I1,I2..., In) it is that can obtain correspond to the interval index of the joint of a resource for meeting user's request or resource set
(JointRangeIndex).The selection interval index of one joint in the last interval index from the multiple joint, and from selected
The corresponding logical resource pond of the interval index of joint in select a resource to be supplied to the user.
In yet another embodiment, multiple joint intervals index of the resource or resource set for meeting user's request is being found
Afterwards, select to access the minimum joint interval rope of temperature from the interval index of multiple joint using the strategy for least often accessing
Draw, wherein it is the user resources number of requests in the unit time for the interval index of joint to access temperature.So as to, it is to avoid at one section
Interior user's request focuses on some logical resource ponds.
For example, it is assumed that the attribute A of resource R is designated as into RA,RAI-th interval positioned at attribute A is designated asCan arrangeThe A attributes for representing resource R ' are located at i+1 interval, and the A attributes of resource R are located at i-th interval of subregion, and
The value of A attribute of the value of the A attributes of R ' than R is big.According to the definition of resource lookup, it is assumed that the CPU of user's request, internal memory and hard disk
The constraints of attribute is by the interval respectively Range that is obtained after mappingCPU=[T1,T2],RangeMem=[S1,S2],
RangeDisk=[P1,P2], it can thus be concluded that the set of the interval index of joint to the resource request for meeting userIn the present embodiment, it is public by following interval selections
Formula to select suitable joint interval to index from the set, and the interval selection formula is:Select_Range={S|δs=
min(δi), i ∈ Joint RangeIndexSet }, that is to say, that the less intervals of selection selective factor B δ.
Wherein, the interval selection factorWherein RAF represents the access temperature in interval,Wherein N generations
Table user's request number in T time section, for example, the user of interval index can be combined in a period of time by counting for each
Resource request quantity obtains the value of N.T represents the time period(Setting constant).MTRepresent the number in interval domestic-investment source in T time section
Amount, then for arbitrary interval S ∈ JointRangeIndexSet,The as choosing of interval S
Select weight, it is assumed here that whenWhen δsIt is infinity.
This interval selection mode is avoided and intensively accessed between some interval situation, i.e. hot areas within a period of time
Phenomenon is accessed, while the situation of interval resources idle high can also be avoided, and with being carried out according to distribution of resource situation
The more interval number of times for accessing of the feature of interval selection, i.e. number of resources is relatively more, and the less interval access of resource is relatively
It is few.
In yet another embodiment of the present invention, a kind of cloud resource scheduling system is additionally provided.Fig. 2 is given according to this hair
One schematic diagram of specific workflow of the cloud resource scheduling system of bright embodiment.Shown in Fig. 2 is single virtual machine resource
Scheduling or physical node(Or cluster or data center)Scheduling instance.In this example, resource before operation(Virtual machine
Or physical node or cluster or data center)Partitioning scenario be divided into 8 areas for CPU, Memeory is divided into 6 areas, Disk
It is divided into 8 areas, is stored in Key-Value memories on available resource information, including joint index and its correspondence
The Resources list.Judge its scheduling type after the request of scheduler reception user and select suitable scheduler module, then will use
Each attribute of family request resource carries out Interval Maps(For example, to CPU, tri- attributes of Mem, Disk are mapped), by mapping
Each attribute of user's request all correspond to one or several subregions in the range of certain numerical value afterwards, describe in fig. 2
The 6th, 7 two intervals of CPU, the 3rd, 4 intervals of the 4th of internal memory the, 5 each interval and hard disks meet respectively after being mapped
The constraints of the CPU, internal memory and hard disk of user's request, then the 8 of the display of Fig. 2 joint is interval indexes corresponding resource set
Resource in conjunction all meets the demand of user, according to the method for interval selection described above, then selects nearest least referenced
Interval carries out the scheduling of resource, that is, have selected the interval index 6_5_3 of joint, then finds 6_5_3 correspondences from Key-Value storages
The Resources list, then the result R120 of selection is notified corresponding watch-dog by and therefrom random one resource R120 of selection
Monitor carries out the reserved of resource, when the unit of scheduling is physical node(Or cluster or data center)When, it is right that watch-dog needs
Resources of virtual machine under R120 is integrated, and then updates the dispatch state of the resources of virtual machine being integrated, if R120 is virtual
Machine resource then directly changes the dispatch state of R120, then updates higher level's resource of correlation(Higher level's resource such as cluster is management
The data center of the cluster)Relevant information, if its interval there occurs variation also need to update its block information.Last watch-dog
The result Result of scheduling is returned into scheduler, user is returned result to after scheduler records scheduling resource information, while
Access record interval where scheduled resource Jia 1.
Because above-mentioned attribute interval division uses average division, as the carrying out of scheduling of resource process may produce
Raw " focus " problem, i.e., some are interval frequently scheduled, and other are interval seldom scheduled.Therefore, it is of the invention another
In embodiment, dynamic interval division can also be adjusted, so as to obtain request of the user to resource fifty-fifty fall at each as far as possible
It is interval.That is to try one's best balancedly assign to each interval by the request of user, and merges adjacent region when interval too small
Between or increase siding-to-siding block length avoiding frequently updating the substantial amounts of resource brought interval migration operation.
Still mainly to consider each resource idle CPU, said as a example by three attributes of free memory and free hard disk
It is bright, i.e. ACPU,AMem,ADisk, to arbitrary attribute Ax, its initial division can be designated as
Wherein LxThe subregion interval number of attribute is represented respectively, thene0=MinAx,ei-1<ei,i
=1,2 ..., LxConstitute attribute AxThe border value set of interval division.It is interval that user in statistics a period of time inquires about the attribute
Resource request quantity.For example will a period of time interior querying attributes AxUser's request record statistical conditions be expressed as { qxr|r=
1,2,...,Lx, wherein qxrRepresent request attribute AxValue fall in [er-1,er) in request number, then ask number
The purpose of dynamic adjustment interval division is to plan interval again so that each interval user's request is tried one's best
Deng each interval average request amount R N is designated asN is the sum of user's request, LxIt is attribute AxSubregion it is interval
Quantity.Mainly include the following steps:
(1)From low interval to interval scanning user's request attribute A successively highxEach interval number of request record, often sweep
Retouch RN request record and just set a boundary value;The codomain of the attribute is divided into come again based on set boundary value many
Individual attribute is interval;So can be obtained by AxA new divisionFrom
And balanced user's request is in each interval distribution.
(2)In view of adjustment interval division some intervals may be caused too short, reason is the original interval division in attribute
In, the user's request in some intervals that falls may be concentrated relatively, and siding-to-siding block length can become very little after repartitioning,
At this moment the renewal efficiency of resource is just influenced whether.So needs are corrected again to interval:Interval of definition [ei-1 *,ei *) length
It is Len to spendi=ei *-ei-1 *, i ∈ [2, Lx-1],Define average area lengthNew interval division is scanned successivelyEach point
Area, calculates the interval length of each subregion, obtains { Leni|i∈[1,Lx]}。
Modification rule is as follows:For arbitrary interval [ei-1 *,ei *) length Leni, i ∈ [1, Lx], if Leni<α*ALen
(Wherein α(α<1)It is threshold coefficient, the value can be obtained by empirical value), then mean that length of an interval spends short, the situation of adjustment
It is as follows:
If a) Leni<α * ALen and Leni+1<α * ALen, merge i-th interval and i+1 interval, while interval number Lx-
1, then consider Leni+2;
b)Leni<α * ALen, Leni+1>α * ALen, adjust LeniThen consider Len to α * ALeni+1;c)Leni≥α*
ALen, continues with Leni+1。
Finally, it is the corresponding logic interval call number of the interval distribution of each attribute obtained after repartitioning;For each
Resource, redefines the new logic interval call number belonging to each property value of the resource, to obtain corresponding to the resource
The interval index of new joint;And multiple logical resource ponds are divided resources into according to the interval index of new joint, corresponding to same
All resources of the interval index of one joint belong to a logical resource pond.
Shown in Fig. 3 be above-mentioned dynamic adjustment interval division process specific example schematic diagram.As shown in figure 3,
Before operation, interval division situation is:Interval number is 8, and threshold coefficient is 0.5, average area length be 1, attribute codomain for [0,
8], the total amount of user's request is 264, and 10,20,30,60,80,35 are followed successively by the user's request quantity of 8 distributions in interval,
20,9.It can be seen that the quantity of the user's request in each interval is very unbalanced.
According to above-mentioned dynamic adjustment interval division optimization method, two steps can be divided into:
(1) 33 resources are often added to and just set an area from low interval to interval each user's request of scanning high successively
Between border, such 264 resources can divide and be evenly distributed in that each is interval, be drawn by the interval after the restructuring of interval in such as Fig. 2
Point.It can be seen that by after the restructuring of interval, the boundary value of interval division also changes, and each length of an interval degree also occurs
Change, some intervals are elongated, and some intervals shorten, each length of an interval degree becomes value of different sizes by the average 1 of preliminary examination,
And the length of the 4th, 5,6 three subregions in new subregion is below threshold alpha * ALen=0.5, the 1st and the 8th length of subregion
Degree and significantly larger than threshold value 0.5.
(2) by the interval restructuring of the first step, some intervals become too short, are modified for too short interval needs, such as
Dotted line frame circle lives the 4th, 5,6 intervals in figure, according to interval modification rule described above, is scanned successively since low interval each
Interval, three length of an interval degree above are all not less than threshold value, the 4th interval [3.7,4.4] and the 5th interval [4.4,4.5]
Length all be 0.4, less than threshold value, then two intervals are merged into [3.7,4.5].Continue to scan on the 6th interval [4.5,
4.9] siding-to-siding block length is also below 0.5, but the 7th siding-to-siding block length is more than 0.5, now only need to by the 6th length of subregion to
7th interval extension is changed into 0.5, while the 7th interval shorten to 0.9.So far interval amendment operation is completed.By after amendment
Resource dividing condition become 7 intervals, the user's request number in each interval is relative to the situation under initial interval division
It is more balanced, while it is also seen that adjustment after resource partitioning interval in some length of an interval degree shortened than initial interval
, this is also trade-off problem of the balancing user request between each interval distribution and resource updates.
Fig. 4 gives the configuration diagram of the cloud resource scheduling system according to another embodiment of the invention.The system bag
Include scheduling granularity Detection device, interval division scheduler, resource partitioning table memory, scheduled resource memory, watch-dog and look into
Consultation record memory, the effect of modules is as follows:
Scheduling granularity Detection device:Scheduling granularity for judging user's request, such as dispatch single virtual machine resource, scheduling
Physical node resource, scheduling cluster resource or scheduling data center resource.
Interval division scheduler:For the scheduling granularity according to user's request, different scheduler modules are selected,(For example it is single
Resources of virtual machine scheduler, physical node Resource Scheduler, cluster resource scheduler, data center resource scheduler), and in mould
Scheduling meets the resource of user's request and returns to user in logical resource pond in block.The different scheduling of resource object of correspondence,
Interval division scheduler includes single resources of virtual machine scheduler, physical node Resource Scheduler, cluster resource scheduler and data
Center resources scheduler etc..
Resource partitioning table memory:The main logical resource pond ID for being responsible for record all kinds resource(Combine interval rope
Draw)Key-value pairs constituted with the resource collection in each logical resource pond, is stored with Key-Value Store frameworks,
Index speed is improved, it includes resources of virtual machine partition table, physical node partition table, cluster resource partition table and data center point
Area's table.
Scheduled resource memory:The scheduled resource of essential record, as prevent resource reentry scheduling judgement according to
According to.
Watch-dog:On the one hand the attribute value information of record resource is needed, in addition it is also necessary to monitor the dynamic change of resource and update
Resource Properties information, the integration that next watch-dog during scheduling of resource is also required to carry out resource is met user's request
Resource collection.The data structure that it includes is as follows:
Inquiry record storage:It is responsible for the resource request of record user and the corresponding relation of watch-dog, is accounted in user's release
For finding corresponding watch-dog and changing the dispatch state of shared resource during with resource.
In the present embodiment, the scheduling of resource of the system is mainly included the following steps that:
1) granularity to resource request from user is detected
When a user's request arrives, the request granularity of user is detected by scheduling granularity Detection device first, that is, judge it
It is to carry out single scheduling virtual machine, or physical node scheduling or colony dispatching or data center's scheduling.
2) the interval index of joint is obtained
Interval division scheduler is according to the corresponding scheduler of scheduling granularity selection for being detected(Single resources of virtual machine scheduling
Device, physical node Resource Scheduler, cluster resource scheduler or data center resource scheduler), the scheduler is according to respective class
The interval division rule of type resource, finds the joint interval indexed set conjunction JointRangeIdexSet for meeting user's constraint requirements,
And therefrom find the interval call number JointRangeIdex of " suitable " joint.
3) resource set is obtained
The joint interval call number JointRangeIdex that scheduler is obtained according to previous step, from resource partitioning table memory
In find the corresponding the Resources lists of JointRangeIdex in corresponding resource partitioning table, then by certain resource selection mechanism
(As randomly choosed)Find that to meet the suitable resource of user's request (be probably single virtual resource, it is also possible to physical node
Or cluster or data center).
4) resource consolidation
The resource is transmitted to corresponding watch-dog by the system, if the resource is single virtual resource, prison
Control device directly changes the dispatch state of the resource;If the resource is physical node or cluster or data center, watch-dog is to this
Subordinate's dummy node of resource is implement resource integration, and finds the less resource collection for meeting user's request.The resource consolidation
The step of be, for example,:
(4-1) first initializes queue SubResource to wait the collection R that reallocates resources.
(4-2) and then successively scan queue SubResource, when scan element is not virtual machine, by under scan element
All resources of one-level add SubResource(The nethermost virtual machine set of scan element can be so had access to always);
Result queue is added when scan element is virtual machine.
(4-3) often carries out once (4-2), whether judged result meets the request of user, such as meets i.e. returning result, if not
Meet and continue to search for, return to (4-2), until finding untill meeting desired results set.
5) resource reservation
Watch-dog needs to carry out the reserved of resource after resource has been integrated, and it is mainly concerned with repairing for scheduled resource information
Change and be scheduled resource higher level's resource information modification(The scheduling of such as cluster may cause the letter of the data center on its upper strata
The modification of breath).The dispatch state of each resource in selected resource or resource set is namely changed, while needing modification corresponding
The available resource information of physical node, cluster and data center.Mainly include that step is as follows:
Each resource in the resource collection that (5-1) traversal watch-dog is integrated, changes its dispatch state, and notify scheduler
Resource R in Key-Value Store is reserved.
(5-2) and then its superior node is considered successively(Higher level's cluster of such as physical node and data center)With resource
Change produced by the scheduling of R, notifies that scheduler updates it in Key-Value Store couple if the interval index of its joint changes
The index information answered.
Scheduling result is returned to user by (5-3).
6) scheduling of resource record
Watch-dog is returning to scheduling result(Resource or resource set)After need local record user's inquiry request and scheduling
The corresponding relation of result to inquiry record sheet in.Meanwhile, be stored in scheduling result in scheduled resource memory by scheduling system, and
The corresponding relation of the watch-dog belonging to user's request and scheduling result is added in inquiry record storage.
In yet another embodiment, the step of scheduling of resource of the system can also include resource updates, such as when certain
When the attribute of resources of virtual machine changes, its more new technological process is as follows:
Watch-dog updates the attribute information of the resource in local virtual resource table;
Watch-dog updates the thing comprising the resource in physical node resource table, cluster resource table and data center resource table
The available resource information of reason node, cluster and data center and interval index information;
When being related to resource(The virtual resource or the physical node comprising the virtual resource, cluster or data center)'s
During demarcation interval index change, watch-dog needs to notify that interval division scheduler updates the phase stored in resource partitioning table memory
Close the interval of the resource in resource partitioning table.
After interval division scheduler is connected to the new information of watch-dog, respective resources in resource partitioning table memory are updated
In partition table on the resource(The virtual resource or the physical node comprising the virtual resource, cluster or data center)Point
Area records.
When resource extent is very big, when the concurrency that user accesses is higher, the renewal of ample resources can bring huge to system
Renewal pressure, the strong influence efficiency of scheduling of resource, while the delay of resource updates can also bring consistency problem.
Prior art and resource updates pattern of the present invention are analyzed with reference to Fig. 5 and Fig. 6.As shown in figure 5,
Traditional resource regeneration method needs to change with the dynamic of Resource Properties and the frequently information of more new resources.Fig. 6 gives
The resource updates pattern employed based on interval migration according to embodiments of the present invention, that is, only when the property value of resource surpasses
Attribute where crossing it just needs to update when interval.As can be seen that resource Rx attributes in state 1 from the contrast of Fig. 5 and Fig. 6
The value of a is 5, is mapped in the 6th interval [4,7], and in state 2, the value of the attribute a of resource Rx is changed into 6, in traditional resource
Need the value of the attribute a of the resource Rx stored in modification system in management system, and in the present embodiment based on interval migration
In resource updates pattern, the value 6 of attribute a is still located in the 6th interval, at this moment avoids the need for the Rx stored in modification system
Corresponding logic index.Only when the attribute a of resource Rx changes to state 3, i.e., when value is changed into 8, the interval belonging to a attributes of Rx
Just change, become the 7th interval [7,9], the logic index for now just needing the resource Rx stored in modification system is Rx:
7.It can thus be seen that the resource updates pattern based on interval migration is asked in the dynamic resource renewal of solution high-frequency in the present embodiment
Topic seems very effective.
Although the present invention has been described by means of preferred embodiments, however the present invention be not limited to it is described here
Embodiment, done various changes and change is also included without departing from the present invention.
Claims (13)
1. a kind of cloud method for managing resource, methods described includes:
Step 1) all available resources are counted, the upper and lower bound of each attribute of resource is obtained, to determine each
The codomain of attribute;
Step 2) codomain of each attribute is divided into multiple attributes intervals, and be the corresponding logic area of the interval distribution of each attribute
Between call number;
Step 3) for each resource, the logic interval call number belonging to each property value of the resource is determined, to obtain correspondence
In the interval index of the joint of the resource;
Step 4) multiple logical resource ponds are divided resources into according to the interval index of joint, corresponding to the interval index of same joint
All resources belong to a logical resource pond.
2. method according to claim 1, also including interval for each attribute, counts the user's inquiry in a period of time
The step of resource request quantity in attribute interval.
3. method according to claim 2, also including based on resource request quantity to adjust interval division the step of, its bag
Include following operation:
Step 51), perform following operation for each attribute of resource:
(511) scan the user's request quantity in the interval of the attribute partition successively, often scan RN request record with regard to setting
One boundary value;
(512) that the codomain of the attribute is divided into multiple attributes come again is interval for the boundary value based on new settings;
(513) it is that corresponding logic interval call number is distributed in each attribute interval obtained after repartitioning;
Step 52), for each resource, determine the new logic interval call number belonging to each property value of the resource, with
To the interval index of new joint corresponding to the resource;
Step 53), multiple logical resource ponds are divided resources into according to the interval index of new joint, it is interval corresponding to same joint
All resources of index belong to a logical resource pond.
4. method according to claim 3, in step 512) the step of also include following amendment attribute siding-to-siding block length:
The average area length ALen in multiple attributes interval that calculating is obtained after repartitioning;
Each attribute after scanning is repartitioned successively is interval, performs following operation:
If ith attribute length of an interval degree is less than α * Alen, and the interval length of i+1 attribute is less than α * Alen, then close
And the i-th interval and i+1 attribute interval of attribute, wherein 0<α<1;
If ith attribute length of an interval degree is less than α * Alen, and the interval length of i+1 attribute is more than α * Alen, then will
Ith attribute length of an interval degree is extended to the corresponding part of length reduction in α * ALen and i+1 attribute interval;
If ith attribute length of an interval degree is more than α * Alen, the siding-to-siding block length keeps constant.
5. method according to claim 1, wherein keep combining using key-value storage organizations interval index and its
Corresponding the Resources list.
6. the method according to any of the above-described claim, wherein, during resource is virtual machine, physical node, cluster or data
The heart.
7. a kind of cloud resource regulating method, methods described includes:
Step 1, receives the resource request from user;
Step 2, for each attribute of requested resource, extracts the constraint upper limit and constraint lower limit of each attribute respectively;
Step 3, determines the logic interval call number belonging to the constraint upper limit and constraint lower limit of each attribute of requested resource,
It is met the interval index of multiple joints of the resource request;
Step 4, one joint of selection is interval in being indexed from the multiple joint interval indexes, and combines interval rope from selected
A resource is selected to be supplied to the user in logical resource pond corresponding to drawing.
8. method according to claim 7, also combines the user of interval index including statistics for each in for a period of time
The step of resource request quantity.
9. method according to claim 7, wherein the step 4 selects to access heat from the interval index of the multiple joint
The interval index of minimum joint is spent, and a resource is selected from the corresponding logical resource pond of the interval index of selected joint
To be supplied to the user, wherein it is the user resources number of requests in the unit time for the interval index of joint to access temperature.
10. method according to claim 7, also including monitoring and updating resource status the step of.
11. methods according to claim 10, wherein only when the attribute where the property value of resource exceedes it is interval
Just need the state of the renewal resource.
A kind of 12. cloud resource scheduling systems, the system includes:
Module for receiving the resource request from user;
For each attribute for requested resource, the constraint upper limit of each attribute and the mould of constraint lower limit are extracted respectively
Block;
For determining the constraint upper limit of each attribute of requested resource and constraining the logic interval call number belonging to lower limit, obtain
To the module of the interval index of multiple joints for meeting the resource request;
Combine interval index for the selection one from the multiple joint interval index, and indexed from selected joint is interval
A resource is selected to be supplied to the module of the user in corresponding logical resource pond.
13. systems according to claim 12, also including the module for monitoring and updating resource status.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210135357.1A CN103383653B (en) | 2012-05-02 | 2012-05-02 | Cloud resource management and dispatching method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210135357.1A CN103383653B (en) | 2012-05-02 | 2012-05-02 | Cloud resource management and dispatching method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103383653A CN103383653A (en) | 2013-11-06 |
CN103383653B true CN103383653B (en) | 2017-07-04 |
Family
ID=49491452
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210135357.1A Expired - Fee Related CN103383653B (en) | 2012-05-02 | 2012-05-02 | Cloud resource management and dispatching method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103383653B (en) |
Families Citing this family (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104753984A (en) * | 2013-12-26 | 2015-07-01 | 乐视网信息技术(北京)股份有限公司 | Data transmission method and data transmission device |
CN105204921B (en) * | 2014-06-25 | 2019-03-01 | 北京奇虎科技有限公司 | Method for allocating tasks, device, system and server |
CN104391697B (en) * | 2014-11-11 | 2017-12-22 | 百度在线网络技术(北京)有限公司 | The cloud resource management system and method for application program |
CN104679595B (en) * | 2015-03-26 | 2017-10-13 | 南京大学 | A kind of application oriented IaaS layers of dynamic resource allocation method |
CN104881322B (en) * | 2015-05-18 | 2018-10-09 | 中国科学院计算技术研究所 | A kind of cluster resource dispatching method and device based on vanning model |
US10182019B2 (en) * | 2016-02-25 | 2019-01-15 | Telefonaktiebolaget Lm Ericsson (Publ) | Interconnected hardware infrastructure resource control |
CN106612318A (en) * | 2016-04-29 | 2017-05-03 | 四川用联信息技术有限公司 | Cloud resource searching method in cloud computing |
CN106127382B (en) * | 2016-06-22 | 2018-09-07 | 财付通支付科技有限公司 | Method for managing resource and device |
CN108241716A (en) * | 2016-12-27 | 2018-07-03 | 北京金山云网络技术有限公司 | A kind of resource introduction method and device |
CN107276914B (en) * | 2017-06-01 | 2020-12-01 | 千寻位置网络有限公司 | Self-service resource allocation scheduling method based on CMDB |
CN107404527B (en) * | 2017-07-26 | 2020-05-29 | 郑州云海信息技术有限公司 | Resource access and control method and device |
CN108572871B (en) * | 2017-12-13 | 2021-03-26 | 北京金山云网络技术有限公司 | Resource allocation method and device, electronic equipment and storage medium |
CN108334409B (en) * | 2018-01-15 | 2020-10-09 | 北京大学 | Fine-grained high-performance cloud resource management scheduling method |
CN109376006B (en) * | 2018-09-04 | 2021-09-21 | 西安电子科技大学 | Resource integration method based on time-varying characteristics of user requirements in cloud computing environment |
CN111694672B (en) * | 2020-06-12 | 2023-04-25 | 抖音视界有限公司 | Resource allocation method, task submission method, device, electronic equipment and medium |
CN111934958B (en) * | 2020-07-29 | 2022-03-29 | 深圳市高德信通信股份有限公司 | IDC resource scheduling service management platform |
CN112395334A (en) * | 2020-11-30 | 2021-02-23 | 浪潮云信息技术股份公司 | Method for dividing centerless distributed database cluster into a plurality of logic sub-clusters |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719155A (en) * | 2009-12-29 | 2010-06-02 | 北京航空航天大学 | Method of multidimensional attribute range inquiry for supporting distributed multi-cluster computing environment |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2006098725A2 (en) * | 2005-03-11 | 2006-09-21 | Cluster Resources, Inc. | System and method for enforcing future policies in a compute environment |
CN102360314A (en) * | 2011-10-28 | 2012-02-22 | 中国科学院计算技术研究所 | System and method for managing resources of data center |
-
2012
- 2012-05-02 CN CN201210135357.1A patent/CN103383653B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719155A (en) * | 2009-12-29 | 2010-06-02 | 北京航空航天大学 | Method of multidimensional attribute range inquiry for supporting distributed multi-cluster computing environment |
Non-Patent Citations (1)
Title |
---|
基于P2P的资源搜索方法研究;王国英;《广西大学硕士研究生学位论文》;20040831;第31页 * |
Also Published As
Publication number | Publication date |
---|---|
CN103383653A (en) | 2013-11-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103383653B (en) | Cloud resource management and dispatching method and system | |
CN100399301C (en) | Mass storage device and method for dynamically managing a mass storage device | |
US20190230000A1 (en) | Intelligent analytic cloud provisioning | |
CN104050042B (en) | The resource allocation methods and device of ETL operations | |
CN102053982B (en) | A kind of database information management method and equipment | |
US8291424B2 (en) | Method and system of managing resources for on-demand computing | |
CN109684074A (en) | Physical machine resource allocation methods and terminal device | |
CN103176849B (en) | A kind of dispositions method of the cluster virtual machine based on resource classification | |
CN103345514A (en) | Streamed data processing method in big data environment | |
CN107864211B (en) | Cluster resource dispatching method and system | |
Chockler et al. | Design and implementation of caching services in the cloud | |
US20070220028A1 (en) | Method and system for managing load balancing in data-processing system | |
CN105975345B (en) | A kind of video requency frame data dynamic equalization memory management method based on distributed memory | |
CN106095531B (en) | A kind of dispatching method of virtual machine loaded based on grade and physical machine in cloud platform | |
CN111752678A (en) | Low-power-consumption container placement method for distributed collaborative learning in edge computing | |
WO2020259082A1 (en) | Cache allocation method and device, storage medium, and electronic device | |
CN102331948A (en) | Resource state-based virtual machine structure adjustment method and adjustment system | |
CN105683941A (en) | Regulating enterprise database warehouse resource usage | |
CN1602480A (en) | Managing storage resources attached to a data network | |
CN102567077A (en) | Virtualized resource distribution method based on game theory | |
CN111737168A (en) | Cache system, cache processing method, device, equipment and medium | |
US20120265569A1 (en) | Determining work placement preferences | |
CN107480254B (en) | Online load balancing method suitable for distributed memory database | |
Fan et al. | Intelligent resource scheduling based on locality principle in data center networks | |
Jiang et al. | Resource-utilization-aware task scheduling in cloud platform using three-way clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170704 Termination date: 20200502 |
|
CF01 | Termination of patent right due to non-payment of annual fee |