CN103383653A - Method and system for managing and dispatching cloud resource - Google Patents

Method and system for managing and dispatching cloud resource Download PDF

Info

Publication number
CN103383653A
CN103383653A CN2012101353571A CN201210135357A CN103383653A CN 103383653 A CN103383653 A CN 103383653A CN 2012101353571 A CN2012101353571 A CN 2012101353571A CN 201210135357 A CN201210135357 A CN 201210135357A CN 103383653 A CN103383653 A CN 103383653A
Authority
CN
China
Prior art keywords
resource
attribute
area
index
user
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
Application number
CN2012101353571A
Other languages
Chinese (zh)
Other versions
CN103383653B (en
Inventor
张健
虎嵩林
刘越
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Computing Technology of CAS
Original Assignee
Institute of Computing Technology of CAS
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Institute of Computing Technology of CAS filed Critical Institute of Computing Technology of CAS
Priority to CN201210135357.1A priority Critical patent/CN103383653B/en
Publication of CN103383653A publication Critical patent/CN103383653A/en
Application granted granted Critical
Publication of CN103383653B publication Critical patent/CN103383653B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention provides a method for managing and dispatching cloud resources, which comprises the following steps: dividing a value range of each attribute of the resource into a plurality of attribute areas, distributing corresponding logic area index number to each attribute area, mapping each resource to a united area index determined by the logic area index number attributed to each attribute value of the resource, when receiving a resource request of a user, determining the logic area index number of the restrict upper limit and the restrict lower limit of each attribute of the requested resource, gaining a plurality of united area indexes satisfying the resource request, then selecting one united area index from the united area indexes, selecting one resource from a logic resource pool corresponding to the selected united area index, and providing the resource to the user. The method can satisfy the requirements of the user resource request with high concurrence degree and the resource renovation with high frequency in the cloud computing environment.

Description

Cloud resource management and dispatching method and system
Technical field
The present invention relates to distributed resource management and scheduling, the high concurrent scheduling and the high-frequency that refer more particularly to magnanimity resource under cloud computing environment are upgraded.
Background technology
In large scale network efficiently scheduling resource be distributed resource management system the key issue that must solve.Under distributed environment, the scale of Internet resources (comprising computational resource, data resource, software resource and Service Source etc.) expands rapidly, these resources are dispersed in each website in network, and dynamic is strong, and are connected to collaborative work on heterogeneous platform by network.Support that from the resource that the grid computing utilization disperses large-scale application is different, cloud computing puts together various network resources carries out unified management and scheduling, and by network with as required, the mode of easily expansion provides various services.Scheduling of resource under cloud computing environment must be upgraded the pressure that brings etc. in the face of the access of magnanimity resource information, high concurrent user resources request and ample resources.
Under cloud computing environment, efficient scheduling resource must satisfy: (1) reads and writes the magnanimity resource information efficiently; (2) navigate to fast resource or the resource set of meeting consumers' demand from the magnanimity resource; (3) alleviate ample resources and update to the pressure that system brings.
In existing cloud computing environment, mainly contain following several resource dispatching strategy: centralized resources scheduling, hierarchy type scheduling of resource, based on the scheduling of resource of P2P route with based on scheduling of resource of DHTs etc.Centralized resources scheduling is by central server unified managing resource and user's request, and concentrated match user request and available resources are carried out resource and distributed.Carry out scheduling of resource based on the scheduling strategy of P2P route by the strategy that mates with route in the whole network territory.Utilizing distributed hashtable based on the resource dispatching strategy of DHTs is that each resource is distributed unique index, and according to index with resource deployment to specific server.Wherein, based on centralized poor with resource management system extendability hierarchy type, be difficult to satisfy the demand that novel large-scale is used.And be difficult to the suitable resource of scheduling in the whole network territory based on the resource regulating method of P2P route, and in large concurrent situation, Internet traffic increases severely.Resource regulating method based on DHTs passes through distributed hash (Hash) policy deployment and locating resource, it has solved the problem of the storage of magnanimity resource and quick locating resource, but consuming time larger when carrying out resource selection, and the expense of the logic index change that brings of resource updates and resource migration is also very large, and the user who is difficult to process highly simultaneous access asks and high-frequency resource updates.
Summary of the invention
Therefore, the object of the invention is to overcome the defective of above-mentioned prior art, a kind of cloud resource regulating method is provided, can process user's request and the high-frequency resource updates of highly simultaneous access.
In order to realize the foregoing invention purpose, adopted following technical proposal:
On the one hand, the invention provides a kind of cloud method for managing resource, described method comprises:
Step 1) is added up all available resources, obtains the upper and lower bound of each attribute of resource, to determine the codomain of each attribute;
Step 2) codomain with each attribute is divided between a plurality of attribute areas, and is to distribute call number between corresponding logic area between each attribute area;
Step 3) is for each resource, determines call number between the logic area under each property value of this resource, to obtain corresponding to index between the association area of this resource;
Step 4) is a plurality of logical resources ponds according to index between association area with resource division, belongs to a logical resource pond corresponding to all resources of index between same association area.
In technique scheme, also can comprise between each attribute area, the user in statistics a period of time inquires about the step of the resource request quantity between this attribute area;
In technique scheme, also can comprise the step of adjusting interval division based on resource request quantity, it comprises following operation:
Step 51), carry out following operation for each attribute of resource:
(511) user that drops in this attribute partition interval of scanning asks quantity successively, asks to record for every scanning RN and just sets a boundary value;
(512) come again the codomain of this attribute is divided between a plurality of attribute areas based on the boundary value of new settings;
(513) for distributing call number between corresponding logic area between each attribute area that obtains after repartitioning;
Step 52), for each resource, determine call number between the new logic area under each property value of this resource, to obtain corresponding to index between the new association area of this resource;
Step 53), be a plurality of logical resources ponds according to index between new association area with resource division, belong to a logical resource pond corresponding to all resources of index between same association area.
In technique scheme, in step 512) also can comprise the step of following correction attribute burst length:
Average burst length ALen between a plurality of attribute areas that obtain after calculating is repartitioned;
Between each attribute area after scanning is repartitioned successively, carry out following operation:
If the length between i attribute area is less than α * Alen, and the length between i+1 attribute area is less than α * Alen, merge between the i attribute area and i+1 attribute area between, 0＜α＜1 wherein;
If the length between i attribute area is less than α * Alen, and the length between i+1 attribute area is greater than α * Alen, the length between i attribute area is extended to α * ALen and length between i+1 attribute area is dwindled corresponding part;
If the length between i attribute area is greater than α * Alen, this burst length remains unchanged.
In technique scheme, can adopt the key-value storage organization to keep index and corresponding the Resources list thereof between association area.
In technique scheme, resource can be virtual machine, physical node, cluster or data center.
Another aspect the invention provides a kind of cloud resource regulating method, and described method comprises:
Step 1 receives the resource request from the user;
Step 2 for each attribute of requested resource, extracts respectively the constraint upper limit and the constraint lower limit of each attribute;
Step 3 is determined call number between the constraint lower limit of each attribute of requested resource and the logic area of constraint under lower limit, is met index between a plurality of association areas of this resource request;
Step 4 is selected index between an association area index between described a plurality of association areas, and select a resource to offer this user between selected association area the corresponding logical resource of index pond.
In technique scheme, also can comprise in statistics a period of time the step for the user resources request quantity of index between each association area.
In technique scheme, described step 4 can be between described a plurality of association areas be selected index between the association area of access temperature minimum index, and select a resource to offer this user between selected association area the corresponding logical resource of index pond, wherein accessing temperature is interior user resources request quantity for index between association area of unit interval.
In technique scheme, can also comprise the step of monitoring and renewal resource status, and the state that need to upgrade this resource when the property value of resource surpasses between the attribute area at its place.
Another aspect the invention provides a kind of cloud resource scheduling system, and described system comprises:
Be used for receiving the module from user's resource request;
Be used for each attribute for requested resource, extract respectively the constraint upper limit of each attribute and the module of constraint lower limit;
Be used for to determine call number between the constraint upper limit of each attribute of requested resource and the logic area under the constraint lower limit, be met the module of index between a plurality of association areas of this resource request;
Be used for selecting index between an association area from index between described a plurality of association areas, and select a resource to offer this user's module between selected association area the corresponding logical resource of index pond.
In said system, can also comprise the module for monitoring and renewal resource status.
Compared with prior art, the invention has the advantages that:
Employing can navigate to rapidly the resource of meeting consumers' demand based on the dispatching method of interval division from the magnanimity resource, and it is interval that user's request balancedly is distributed in each resource division as far as possible, to improve the efficient of scheduling of resource.When resource status changes, adopt the resource updates pattern based on the interval, greatly reduce the pressure that resource updates frequently brings to resource management system.
Description of drawings
Embodiments of the present invention is further illustrated referring to accompanying drawing, wherein:
Fig. 1 is for being used for implementing the cloud resource environment schematic diagram of the embodiment of the present invention;
Fig. 2 is the cloud scheduler course of work schematic diagram 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 the cloud dispatching system configuration diagram based on interval division according to 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.
Embodiment
In order to make purpose of the present invention, technical scheme and advantage are clearer, and the present invention is described in more detail by specific embodiment below in conjunction with accompanying drawing.Should be appreciated that specific embodiment described herein only in order to explain the present invention, is not intended to limit the present invention.
Fig. 1 has provided the cloud resource environment schematic diagram that is used for implementing the embodiment of the present invention.This environment is comprised of several data centers, and each data center comprises several clusters that is comprised of physical node, several virtual machines of operation on each physical node.And resource can be virtual machine, can be also physical node or cluster or data center.As shown in Figure 1, various resources are organized with layer architecture, and the user can select to need the resource granularity of scheduling according to demand, can selection scheduling single virtual machine resource, also can dispatch physical node, cluster or data center etc.Wherein resources of virtual machine is the base unit of scheduling of resource.For each resource R, relate generally to three calculated performance attributes with and position attribution, wherein three calculated performance attributes are respectively: idle CPU(represents with check figure), free memory (GB) and idle hard disk (GB).Like this, resource R can simply be expressed as R={[FCPU, FMemory, FDisk], [DC, Cluster, PN] }, FCPU wherein, FMemory, FDisk are respectively idle CPU(and represent with check figure), free memory (GB) and idle hard disk (GB), DC represents the data center under R, and Cluster represents the cluster under R, and PN represents the physical node under R.In layer architecture shown in Figure 1, the stock number of every layer can obtain by the stock number of its lower one deck is integrated, for example for each physical node, can obtain by all resources of virtual machine that integration is located thereon the calculated performance attribute of this physical node resource.
In one embodiment, provided cloud method for managing resource based on interval division.At first the method provides the resource initial interval division of (comprising virtual machine, physical node, cluster, data center etc.) according to empirical value, then according to this interval division with resource mapping in corresponding logical resource pond.More specifically, the method mainly comprises the following steps:
Step 1 is added up all available resources, obtains the upper and lower bound of each attribute of resource, to determine the codomain of each attribute.
For example, determine the upper and lower bound of each attribute (adding up idle CPU, free memory and idle hard disk) of each statistic unit (for example, virtual machine, physical node, cluster and data center).Take resources of virtual machine as example, at first determine the upper and lower bound of each attribute (for example, idle CPU, free memory, idle hard disk etc.) of each resources of virtual machine, and then obtain the codomain of each attribute.The CPU codomain is designated as RR CPU=[CPUmin, CPUmax], wherein CPUmin is the lower limit of CPU, and CPUmax is the upper limit of CPU, and the codomain of the free memory of resource R is designated as RR Mem=[MEMmin, MEMmax] and the codomain of idle hard disk are designated as RR Disk=[DISKmin, DISKmax].
Step 2 is divided into the codomain of each attribute between a plurality of attribute areas, and is to distribute call number between corresponding logic area between each attribute area.
At first, the codomain with each attribute is divided between a plurality of attribute areas.For example, can adopt at first the method for average division to carry out subregion to the codomain of each attribute.Will introduce the method for each interval adjustment being optimized interval division hereinafter.For example, initial resource partitioning is: CPU(lower limit CPUmin=0, upper limit CPUmax=32) on average be divided into M subregion, internal memory (lower limit MEMmin=0G, upper limit MEMmax=16G) on average be divided into N subregion, hard disk (lower limit DISKmin=0G, upper limit DISKmax=1000G) on average be divided into P subregion, have:
(1) for CPU arbitrarily i, i ∈ [1, M-1] and i are integer, and CPU is arranged i∈ RR CPUAnd CPU i＜CPU i+1, CPU 0=CPUmin, CPU M=CPUmax; T CPU={ [CPU i, CPU i+1) | i ∈ [0, M-1] } consisted of the differentiation (i.e. a subregion, lower with) of [CPUmin, CPUmax] between the idle CPU attribute whole district, B CPU={ CPU i| i ∈ [0, M] } consist of the division border set of CPU, utilize B CPUCan calculate the interval under given CPU value.
(2) for Mem arbitrarily j, j ∈ [1, N-1] and j are integer, and Mem is arranged j∈ RR MemAnd Mem j＜Mem j+1, Mem 0=MEMmin, Mem N=MEMmax; T MEM={ [Mem j, Mem j+1) | j ∈ [0, N-1] } consist of the differentiation of [MEMmin, MEMmax] between the free memory attribute whole district, B MEM={ Mem j| j ∈ [0, N] } consist of the set on the division border of free memory, utilize B MEMCan calculate the interval under given CPU value.
(3) for Disk arbitrarily k, k ∈ [1, P-1] and k are integer, and Disk is arranged k∈ RR DiskAnd Disk k＜Disk k+1, Disk 0=DISKmin, Disk P=DISKmax, T Disk={ [Disk k, Disk k+1) | k ∈ [0, P-1] } consist of the differentiation of [DISKmin, DISKmax] between the idle hard disk attribute whole district, B DISK={ Disk k| k ∈ [0, P] } consisted of the division border set of idle hard disk attribute, utilize B DISKCan calculate the interval under given CPU value.
Then, for distributing call number between corresponding logic area between each attribute area of dividing.For example can refer to that for call number i between this logic area between i attribute area of certain attribute, wherein i is integer.
Step 3 for each resource, is determined call number between the logic area under each property value of this resource, to obtain corresponding to index between the association area of this resource.
In fact be exactly resource R finally is mapped between its corresponding association area on index.Still describe as an example of resources of virtual machine R example.N attribute A for each resources of virtual machine R 1, A 2..., A n, through mapping
Figure BDA00001596350500061
X ∈ [1, n] obtains call number I between each self-corresponding logic area of n attribute 1, I 2..., I n, wherein,
Figure BDA00001596350500062
Expression attribute A xThe set of initial interval division border (namely top each attribute codomain is averaged after, the boundary value set that each is interval, for example B CPU, B MEM, B DISK),
Figure BDA00001596350500063
Expression attribute A xValue, f x() expression acts on attribute A xMapping ruler, then to call number I between n logic area 1, I 2..., I nAgain shine upon and obtain index Joint RangeIndex=F (I between association area corresponding to resource R 1, I 2..., I n), i.e. logical resource pond ID, F () carry out again the map operation rule to each logic index, obtain like this a JointRangeIndex of each resource ownership through twice mapping.
The below is example with three attributes (idle CPU, free memory and idle hard disk) of resource, illustrates how to shine upon.For resource R={＜FreeCPU, FreeCPUValue 〉,＜FreeMemory, FreeMemValue 〉,＜FreeDisk, FreeDiskValue 〉, its three attributes are shone upon according to rule separately respectively, have:
i＝f CPU(B CPU,FreeCPUValue),i∈[1,M];
j=f MEM(B MEM,FreeMemValue),j∈[1,N];
k=f DISK(B DISK,FreeDiskValue),k∈[1,P].
Regular f wherein CPUCan be described below: at the B of ascending order CPUIn set, find the position i of first value that is not less than set-point FreeCPUValue, f MEM, f DISKDefinition similar.
Then, by Quadratic Map rule F () with index i between the logic area of three attributes, j, k carries out Joint Mapping and obtains index JointRangeIndex=F (i between association area, j, k), here F () mapping ruler can adopt cascade mode (for example: with " _ " be connected), namely according to specific attribute order (be CPU-here〉MEM-〉order of DISK), call number between each logic area is combined, obtain index i_j_k between association area, certainly F () also can be defined as other rule, depending on uses.
Step 4 is a plurality of logical resources ponds according to index between association area with resource division, belongs to a logical resource pond corresponding to all resources of index between same association area.In advance with reference to figure 2, as shown in Figure 2, by above-mentioned interval division method, can be a plurality of logical resources ponds with the available resource division under cloud environment, each logical resource pond comprises one or more resources, by (the 6_4_3:{R in Fig. 2 for example of index (also can be called logical resource pond ID) between each corresponding association area in logical resource pond 100, R 50..., 6_5_3:{R 95, R 120, R 80Etc.), can navigate to fast user's requested resource.
In yet another embodiment, the key-value that can utilize key-value (key assignments) storage organization to store to be made of logical resource pond ID and corresponding resource collection thereof is to (being index and corresponding the Resources list thereof between association area), the actual content of storage resources no longer, do like this and can effectively utilize storage space, can also support concurrent inquiry to accelerate the efficient of resource location, also avoid simultaneously the problem of the resource content stored in the frequent updating storage system.
In yet another embodiment of the present invention, provide a kind of cloud resource regulating method based on interval division.The method (also can be described as query requests sometimes in the resource request that receives from the user, in fact be exactly the request of user's query resource) time, resource request is resolved, for example, detect the type (namely dispatching granularity) of resource, as virtual machine, physical node, cluster or data center.Then, for each attribute of requested resource, extract the constraint upper limit and the constraint lower limit of each attribute.Then, determine call number between the constraint upper limit of each attribute of requested resource and the logic area of constraint under lower limit, be met index between a plurality of association areas of this resource request.Select at last index between an association area index between described a plurality of association areas, and select a resource to offer this user between selected association area the corresponding logical resource of index pond.
Wherein, user's resource request can be similar [attribute, operational character, value] the form of attribute constraint, wherein attribute (attribute) is character string (string) type, operational character (operator) have＜,＜=, ,=, value (value) is integer (integer), double type etc.
For example, the attribute when certain resource request is constrained to:
Q={＜CPU, CPUOper, QCPUValue 〉,＜Memory, MemOper, QMemValue 〉,＜Disk, DiskOper, QDiskValue〉} time, wherein, CPUOper, MemOper, DiskOper are the operational character (operator) of each attribute, such as,＜, ,≤〉=etc.Can shine upon according to above same mapping ruler each attribute to this request.
i′=f CPU(B CPU,QCPUValue),i′∈[1,M];
j′=f MEM(B MEM,QMemValue),j′∈[1,N];
k′=f DISK(B DISK,QDiskValue),k′∈[1,P].
When＜i CPUOper i '〉﹠amp;＜j MemOper j '〉﹠amp;＜k DiskOper k '〉when setting up, certain the resource R in the logical resource pond that will inquire about that Q matches that between association area, index i_j_k is corresponding.
In yet another embodiment, in order to support better the interval resource request of multiattribute, can adopt resource request as follows:
Table 1
Figure BDA00001596350500091
wherein, user's resource request has mainly comprised value attribute and position attribution, wherein＜request_type〉resource type of representative request, can be single virtual machine (Virtual Machine) or physical node (Physical Node) or cluster (Cluster) or data center (Data Center),＜num_machine〉machine quantity that needs of expression user,＜compute_attr〉represent computation attribute, comprise that idle CPU(represents with check figure), free memory (GB) and idle hard disk (GB), namely＜cpu_core 〉,＜free_mem 〉,＜free_disk 〉, showed in table 1 that the user asks a cluster, require the attribute characteristics of 4 machines to be: check figure is between 2 to 4, internal memory is less than or equal to 2G, idle hard-disk capacity is between 100 to 200G, when asking, the user also can only comprise one or more in three attributes.In the pattern of this support interval query, the calculated performance attribute of user's request represents with the interval, for example user's request " internal memory is not less than the machine of 2G ", attribute＜free_mem〉be expressed as＜free_mem 2, Max＜/free_mem 〉, namely the constraint condition of each attribute represents with an interval upper and lower bound.
When such resource request arrives, at first by resolving each attribute (for example, idle CPU check figure, free memory and idle hard-disk capacity) and the attribute constraint (the constraint upper limit and constraint lower limit) that can obtain resource request.For example, the attribute A that occurs in scanning user request 1, A 2..., A n, for each attribute A xThe constraint bound of (x ∈ [1, n]) is designated as
Figure BDA00001596350500092
Wherein
Figure BDA00001596350500093
Be designated as attribute A xThe constraint lower limit,
Figure BDA00001596350500094
Represent attribute A xThe constraint upper limit.Then, first by a logical mappings
Figure BDA00001596350500095
With
Figure BDA00001596350500096
F wherein x() expression is to attribute A xMapping ruler,
Figure BDA00001596350500097
Expression attribute A xAn interval differentiation, obtain n lower limit logic index of user's request
Figure BDA00001596350500098
With n upper limit logic index
Figure BDA00001596350500099
Then, carry out resource lookup, namely search index between the association area of the resource that satisfies customer requirements or resource set.For certain user's resource request, can obtain index between the association area of a lot of the resources that satisfy customer requirements, for example, be designated as:
JointRangeIndexSet = Σ I 1 = I 1 min I 1 max Σ I 2 = I 2 min I 2 max . . . Σ I n = I n min I n max F ( I 1 , I 2 , . . . , I n )
Logic index (the I to each attribute that namely above introduces of F () wherein x=(x ∈ [1, n])) the rule of map operation again, F (I 1, I 2..., I n) can obtain index (JointRangeIndex) between the association area of corresponding a resource that satisfies user's request or resource set.Select at last index between an association area index between described a plurality of association areas, and select a resource to offer this user between selected association area the corresponding logical resource of index pond.
In yet another embodiment, after index between a plurality of association areas that find the resource that satisfies user request or resource set, adopt the strategy of least often access to come to select index index between the association area of access temperature minimum between a plurality of association areas, wherein access temperature and be in the unit interval user resources request quantity for index between association area.Thereby, avoid in a period of time user's request to focus on some logical resources pond.
For example, suppose the attribute A of resource R is designated as R A, R AI the interval that is positioned at attribute A is designated as Can arrange The A attribute that represents resource R ' is positioned at i+1 interval, and the A attribute of resource R is positioned at i interval of subregion, and the value of the A attribute of R ' is larger than the value of the A attribute of R.According to the definition of resource lookup, suppose that the constraint condition of CPU, internal memory and the hard disk attribute of user's request is respectively Range through the interval that obtains after shining upon CPU=[T 1, T 2], Range Mem=[S 1, S 2], Range Disk=[P 1, P 2], can be met thus the set of index between user's the association area of resource request JointRangeIndexSet = Σ i = T 1 T 2 Σ j = S 1 S 2 Σ k = P 1 P 2 i _ j _ k . In the present embodiment, come to select index between suitable association area from this set by following interval selection formula, described interval selection formula is: Select_Range={S| δ s=min (δ i), i ∈ Joint RangeIndexSet} that is to say and select to select factor delta less interval.
Wherein, the interval selection factor
Figure BDA00001596350500105
The access temperature between the RAF Representative Region wherein, Wherein N representative is in T user's number of request in the time period, for example, and can be by obtain the value of N in statistics a period of time for the user resources request quantity of index between each association area.T represents the time period (setting constant).M TRepresent the quantity in interval domestic-investment source in the T time period, for interval S ∈ JointRangeIndexSet arbitrarily, Be the selection weight of interval S, hypothesis is worked as here The time δ sBe infinity.
This interval selection mode has avoided within a period of time the concentrated area to access some interval situation, be to access phenomenon between hot area, can also avoid simultaneously the situation of high interval resources idle, and has a feature of carrying out interval selection according to the distribution of resource situation, the number of times that is number of resources more interval access is relatively many, and the less interval access of resource is relatively less.
In yet another embodiment of the present invention, also provide a kind of cloud resource scheduling system.Fig. 2 has provided the schematic diagram according to a specific works flow process of the cloud resource scheduling system of the embodiment of the present invention.Shown in Figure 2 is the scheduling example of single virtual machine scheduling of resource or physical node (or cluster or data center).In this example, before operation, the subregion situation of resource (virtual machine or physical node or cluster or data center) is that CPU is divided into 8 districts, Memeory is divided into 6 districts, Disk is divided into 8 districts, be stored in the Key-Value storer about available resource information, comprise joint index and corresponding the Resources list thereof.the request that scheduler receives the user judges afterwards its scheduling type and selects suitable scheduler module, then each attribute with user's request resource (for example carries out Interval Maps, to CPU, Mem, three attributes of Disk shine upon), through each attribute of user request after mapping corresponding one or several subregions in the certain numerical value scope all, describe in Fig. 2 be after shining upon CPU the 6th, 7 two intervals, the 4th of internal memory, the 3rd of 5 each intervals and hard disk, the CPU of user's request has been satisfied respectively in 4 intervals, the constraint condition of internal memory and hard disk, between 8 of the demonstration of Fig. 2 association areas, the resource in the corresponding resource collection of index all satisfies user's demand, method according to the interval selection of above introducing, select the interval of nearest least referenced to carry out the scheduling of resource, namely selected index 6_5_3 between association area, then find the Resources list corresponding to 6_5_3 from the Key-Value storage, and random resource R120 of selection therefrom, the R120 as a result that then will select notifies corresponding watch-dog Monitor to carry out the reservation of resource, when the unit of scheduling is physical node (or cluster or data center), watch-dog need to be integrated the resources of virtual machine under R120, then upgrade the dispatch state of integrated resources of virtual machine, if being resources of virtual machine, R120 directly changes the dispatch state of R120, then upgrade the relevant information of relevant higher level's resource (namely managing the data center of this cluster as higher level's resource of cluster), also need to upgrade its block information if change has occured in its interval.The Result as a result that last watch-dog will be dispatched returns to scheduler, after scheduler records scheduling resource information, result is returned to the user, and the Visitor Logs between the resource that will be scheduled simultaneously location adds 1.
What adopt due to above-mentioned attribute interval division is average division, and along with the carrying out of scheduling of resource process may produce " focus " problem, namely some interval frequently is scheduled, and other intervals seldom are scheduled.Therefore, in yet another embodiment of the present invention, can also dynamically adjust interval division, so that the user drops on each interval as far as possible fifty-fifty to the request of resource.That is to say each interval is balancedly assigned in user's request as far as possible, and merge adjacent interval when too small or increase burst length to avoid upgrading frequently migration operation between a large amount of resource-area that brings when interval.
Still so that each resource is mainly considered idle CPU, three attributes of free memory and idle hard disk are that example describes, i.e. A CPU, A Mem, A Disk, to attribute A arbitrarily x, its initial division can be designated as
Figure BDA00001596350500121
L wherein xRepresent respectively the subregion interval number of attribute,
Figure BDA00001596350500122
e 0=MinA x,
Figure BDA00001596350500123
e i-1＜e i, i=1,2 ..., L xConsist of attribute A xThe boundary value set of interval division.User in statistics a period of time inquires about the resource request quantity between this attribute area.For example with querying attributes A in a period of time xUser's statistical conditions of asking to record be expressed as { q xr| r=1,2 ..., L x, q wherein xrExpression request attribute A xValue drop on [e r-1, e r) in the number of request, ask number
Figure BDA00001596350500124
The purpose of dynamically adjusting interval division is again between the planning region, makes each interval user request as far as possible impartial, and each interval average request amount R N is designated as N is the sum of user's request, L xBe attribute A xThe interval quantity of subregion.Mainly comprise the following steps:
(1) from low interval user's request attribute A that scans successively to high interval xEach interval number of request record, a boundary value just set in every scanning RN request record; Come again the codomain of this attribute is divided between a plurality of attribute areas based on the boundary value that sets; So just can obtain A xA new division Thereby balanced user's request is in the distribution in each interval.
(2) consider that adjusting interval division may cause some interval too short, reason is in the original interval division of attribute, dropping on some interval interior user request may be relatively concentrated, and between the back zone, length can become very little through repartitioning, and at this moment will have influence on the update efficiency of resource.So need to again revise the interval: interval of definition [e i-1 *, e i *) length be Len i=e i *-e i-1 *, i ∈ [2, L x-1], Len 1 = e 1 * - e 0 , Len L x = e L x - e L x - 1 * ; Define average burst length
Figure BDA00001596350500133
Scan successively new interval division T x ′ = { [ e 0 , e 1 * ) , [ e 1 * , e 2 * ) , . . . , [ e L x - 1 * , e L x ] } Each subregion, calculate each subregion length of an interval degree, obtain { Len i| i ∈ [1, L x].
Modification rule is as follows: for interval [e arbitrarily i-1 *, e i *) length L en i, i ∈ [1, L x], if Len i＜α * ALen(wherein α (α＜1) is the threshold value coefficient, and this value can be obtained by empirical value), mean that length of an interval spends shortly, the situation of adjustment is as follows:
If a) Len i＜α * ALen and Len i+1＜α * ALen merges i interval and i+1 interval, interval number L simultaneously x-1, then consider Len i+2
B) Len i＜α * ALen, Len i+1α * ALen, Len adjusted iThen consider Len to α * ALen i+1C) Len i〉=α * ALen continues to consider Len i+1
At last, for distributing call number between corresponding logic area between each attribute area that obtains after repartitioning; For each resource, redefine call number between the new logic area under each property value of this resource, to obtain corresponding to index between the new association area of this resource; And be a plurality of logical resources ponds according to index between new association area with resource division, belong to a logical resource pond corresponding to all resources of index between same association area.
The schematic diagram of the concrete example of above-mentioned dynamic adjustment interval division process that shown in Figure 3 is.As shown in Figure 3, before operation, the interval division situation is: interval number is 8, and the threshold value coefficient is 0.5, and average burst length is 1, and the attribute codomain is [0,8], the total amount of user request is 264, asks quantity to be followed successively by 10,20 the user of the distribution in 8 intervals, 30,60,80,35,20,9.As seen the quantity of user's request in each interval is very unbalanced.
Dynamic adjustment interval division optimization method according to above-mentioned can be divided into for two steps:
(1) often be added to 33 resources and just set an interval border to each user's request of high interval scanning from low interval successively, such 264 resources can be divided and are evenly distributed in each interval, as in Fig. 2 through the interval division after interval restructuring.Can find out through after the restructuring of interval, the boundary value of interval division also changes, and variation has also occured in each length of an interval degree, some intervals are elongated, some intervals shorten, each length of an interval degree has become by the average 1 of preliminary examination the value that varies in size, and the length of the 4th, 5,6 three subregion in new subregion is all lower than threshold alpha * ALen=0.5, and the length of the 1st and the 8th subregion is again far away higher than threshold value 0.5.
(2) through the interval restructuring of the first step, some interval becomes too short, need to revise for too short interval, live the 4th, 5,6 interval as dotted line frame circle in figure, according to the interval modification rule of above introducing, scan successively each interval from low interval beginning, the length of the Three regions of front all is not less than threshold value, the 4th interval [3.7,4.4] and the length in the 5th interval [4.4,4.5] be all 0.4, lower than threshold value, two intervals are merged into [3.7,4.5].Continue the burst length in the 6th interval of scanning [4.5,4.9] also lower than 0.5, but the 7th burst length only needs the length of the 6th subregion is become 0.5 to the 7th interval the extension this moment greater than 0.5, the 7th interval shortens to 0.9 simultaneously.So far interval correction operation is completed.7 intervals have been become through revised resource division situation, user's number of request in each is interval is more balanced with respect to the situation under initial interval division, can find out also that simultaneously in the resource partitioning interval after adjustment, some length of an interval degree has shortened than initial interval, this is also that balance user request is in the distribution in each interval and the trade-off problem between resource updates.
Fig. 4 has provided according to the present invention the configuration diagram of the cloud resource scheduling system of another embodiment.This system comprises scheduling granularity Detection device, interval division scheduler, resource partitioning table storer, modulated degree resource memory, watch-dog and query note storer, and the effect of modules is as follows:
Scheduling granularity Detection device: be used for the scheduling granularity of judgement user request, such as scheduling single virtual machine resource, scheduling physical node resource, scheduling cluster resource or data dispatching center resources.
Interval division scheduler: be used for the scheduling granularity according to user's request, select different scheduler modules, (single resources of virtual machine scheduler for example, the physical node Resource Scheduler, the cluster resource scheduler, data center's Resource Scheduler), in the logical resource pond and in module, scheduling is satisfied the resource of user's request and returns to the user.Corresponding different scheduling of resource objects, the interval division scheduler comprises single resources of virtual machine scheduler, physical node Resource Scheduler, cluster resource scheduler and data center's Resource Scheduler etc.
Resource partitioning table storer: the logical resource pond ID(that mainly is responsible for recording resources of various types is index between association area) and each logical resource pond in key-value pair of consisting of of resource collection, store with Key-Value Store framework, improve index speed, it comprises resources of virtual machine partition table, physical node partition table, cluster resource partition table and data center's partition table.
Modulated degree resource memory: the resource that essential record has been dispatched, as the basis for estimation that prevents that resource from reentrying and dispatching.
Watch-dog: need to record on the one hand the property value information of resource, also need to monitor the dynamic change of resource and upgrade Resource Properties information, the integration that secondly watch-dog also needs to carry out resource in the process of scheduling of resource is met the resource collection that the user asks.It data structure that comprises is as follows:
Figure BDA00001596350500151
Query note storer: be responsible for the resource request of recording user and the corresponding relation of watch-dog, be used for the dispatch state that finds corresponding watch-dog and revise shared resource when user's release busy resource.
In the present embodiment, the scheduling of resource of this system mainly comprises the following steps:
1) detection is from user's the granularity to resource request
When a user asks to arrive, at first detected user's request granularity by scheduling granularity Detection device, judge that namely it is to carry out single scheduling virtual machine, or physical node scheduling or colony dispatching or data center's scheduling.
2) obtain index between association area
The interval division scheduler is according to the corresponding scheduler of scheduling granularity selection (the single resources of virtual machine scheduler that detects, the physical node Resource Scheduler, cluster resource scheduler or data center's Resource Scheduler), this scheduler is according to the interval division rule of respective type resource, find index set JointRangeIdexSet between the association area that satisfies user's constraint requirements, and therefrom find call number JointRangeIdex between " suitable " association area.
3) Gains resources collection
Call number JointRangeIdex between the association area that scheduler obtains according to previous step, find the Resources list corresponding to JointRangeIdex in corresponding resource partitioning table from resource partitioning table storer, find by certain resource selection mechanism (as random selection) the suitable resource (may be the single virtual resource, may be also physical node or cluster or data center) that satisfies user's request again.
4) resource consolidation
This system passes to corresponding watch-dog with this resource as scheduling result, if this resource is the single virtual resource, watch-dog is directly revised the dispatch state of this resource; If this resource is physical node or cluster or data center, watch-dog is implement resource integration to subordinate's dummy node of this resource, finds the less resource collection of meeting consumers' demand.The step of described resource consolidation is for example:
(4-1) first initialize queue SubResource is for waiting the collection R that reallocates resources.
(4-2) scan queue SubResource successively then when scan element is not virtual machine, adds SubResource(can have access to the nethermost virtual machine set of scan element all resources of scan element next stage so always); When being virtual machine, scan element adds result queue.
(4-3) often carry out once (4-2), whether judged result meets user's request, as meets namely and return results, and does not continue to search if meet, and returns to (4-2), until find the results set that meets the demands.
5) resource reservation
Watch-dog needs to carry out the reservation of resource after having integrated resource, it is mainly concerned with the modification (may cause the modification of information of the data center on its upper strata such as the scheduling of cluster) of higher level's resource information of the modification of the resource information that is scheduled and the resource that is scheduled.Namely revise the dispatch state of each resource in selected resource or resource set, need simultaneously to revise the available resource information of corresponding physical node, cluster and data center.Comprise that mainly step is as follows:
(5-1) each resource in the resource collection of traversal watch-dog integration, revise its dispatch state, and the notice scheduler is reserved resource R in Key-Value Store.
(5-2) then consider successively its superior node (such as higher level's cluster and the data center of physical node) along with the variation of the dispatching office generation of resource R, if between its association area, the index change notifies scheduler to upgrade it at index information corresponding to Key-Value Store.
(5-3) scheduling result is returned to the user.
6) scheduling of resource record
Watch-dog need to be at the corresponding relation of local record user query requests and scheduling result in the query note table after returning to scheduling result (resource or resource set).Simultaneously, dispatching system deposits scheduling result in modulated degree resource memory in, and the corresponding relation of the watch-dog under user's request and scheduling result is added in the query note storer.
In yet another embodiment, the scheduling of resource of this system can also comprise the step of resource updates, and for example when the attribute of certain resources of virtual machine changed, more new technological process was as follows for it:
Watch-dog upgrades the attribute information of this resource in the local virtual resource table;
Watch-dog upgrades available resource information and the interval index information that comprises physical node, cluster and the data center of this resource in physical node resource table, cluster resource table and data center's resource table;
When relating to the demarcation interval index change of resource (this virtual resource or comprise physical node, cluster or the data center of this virtual resource), watch-dog need to notify the interval division scheduler to upgrade the interval of the resource in the related resource partition table of storing in resource partitioning table storer.
After the interval division scheduler is received the updating message of watch-dog, upgrade in resource partitioning table storer the partitioned record about this resource (this virtual resource or comprise physical node, cluster or the data center of this virtual resource) in the respective resources partition table.
When resource extent is very large, when the concurrency of user access was higher, the renewal meeting of ample resources brought huge renewal pressure to system, has affected greatly the efficient of scheduling of resource, and the delay of resource updates simultaneously also can bring consistency problem.
Below in conjunction with Fig. 5 and Fig. 6, prior art and resource updates pattern of the present invention are analyzed.As shown in Figure 5, traditional resource regeneration method need to be along with the dynamic change of Resource Properties the information of new resources more frequently.Fig. 6 has provided according to the employing of the embodiment of the present invention resource updates pattern based on interval migration, and namely only having when the property value of resource surpasses between the attribute area at its place just needs to upgrade.Can find out from the contrast of Fig. 5 and Fig. 6, the value of resource Rx attribute a when state 1 is 5, be mapped to the 6th interval [4,7] in, the value of the attribute a of resource Rx becomes 6 when state 2, needs the value of the attribute a of the resource Rx that stores in the modification system in traditional resource management system, and in the resource updates pattern based on interval migration of the present embodiment, the value 6 of attribute a still is positioned at the 6th interval, does not at this moment just need logic index corresponding to Rx of storing in the modification system.Only have the attribute a as resource Rx to change to state 3, namely value becomes at 8 o'clock, and interval under a attribute of Rx just changes, and becomes the 7th interval [7,9], and just needing the logic index of the resource Rx that stores in the modification system at this moment is Rx:7.This shows, the resource updates pattern based on interval migration in the present embodiment seems very effective in solution high-frequency dynamic resource replacement problem.
Although the present invention is described by preferred embodiment, yet the present invention is not limited to embodiment as described herein, also comprises without departing from the present invention various changes and the variation done.

Claims (13)

1. cloud method for managing resource, described method comprises:
Step 1) is added up all available resources, obtains the upper and lower bound of each attribute of resource, to determine the codomain of each attribute;
Step 2) codomain with each attribute is divided between a plurality of attribute areas, and is to distribute call number between corresponding logic area between each attribute area;
Step 3) is for each resource, determines call number between the logic area under each property value of this resource, to obtain corresponding to index between the association area of this resource;
Step 4) is a plurality of logical resources ponds according to index between association area with resource division, belongs to a logical resource pond corresponding to all resources of index between same association area.
2. method according to claim 1, also comprise between each attribute area, and the user in statistics a period of time inquires about the step of the resource request quantity between this attribute area.
3. method according to claim 2, also comprise the step of adjusting interval division based on resource request quantity, and it comprises following operation:
Step 51), carry out following operation for each attribute of resource:
(511) user that drops in this attribute partition interval of scanning asks quantity successively, asks to record for every scanning RN and just sets a boundary value;
(512) come again the codomain of this attribute is divided between a plurality of attribute areas based on the boundary value of new settings;
(513) for distributing call number between corresponding logic area between each attribute area that obtains after repartitioning;
Step 52), for each resource, determine call number between the new logic area under each property value of this resource, to obtain corresponding to index between the new association area of this resource;
Step 53), be a plurality of logical resources ponds according to index between new association area with resource division, belong to a logical resource pond corresponding to all resources of index between same association area.
4. method according to claim 3 is in step 512) also comprise the step of following correction attribute burst length:
Average burst length ALen between a plurality of attribute areas that obtain after calculating is repartitioned;
Between each attribute area after scanning is repartitioned successively, carry out following operation:
If the length between i attribute area is less than α * Alen, and the length between i+1 attribute area is less than α * Alen, merge between the i attribute area and i+1 attribute area between, 0＜α＜1 wherein;
If the length between i attribute area is less than α * Alen, and the length between i+1 attribute area is greater than α * Alen, the length between i attribute area is extended to α * ALen and length between i+1 attribute area is dwindled corresponding part;
If the length between i attribute area is greater than α * Alen, this burst length remains unchanged.
5. method according to claim 1, wherein adopt the key-value storage organization to keep index and corresponding the Resources list thereof between association area.
6. according to the described method of above-mentioned arbitrary claim, wherein, resource is virtual machine, physical node, cluster or data center.
7. cloud resource regulating method, described method comprises:
Step 1 receives the resource request from the user;
Step 2 for each attribute of requested resource, extracts respectively the constraint upper limit and the constraint lower limit of each attribute;
Step 3 is determined call number between the constraint upper limit of each attribute of requested resource and the logic area of constraint under lower limit, is met index between a plurality of association areas of this resource request;
Step 4 is selected index between an association area index between described a plurality of association areas, and select a resource to offer this user between selected association area the corresponding logical resource of index pond.
8. method according to claim 7 also comprises in statistics a period of time the step for the user resources request quantity of index between each association area.
9. method according to claim 7, wherein said step 4 is selected index between the association area of access temperature minimum index between described a plurality of association areas, and select a resource to offer this user between selected association area the corresponding logical resource of index pond, wherein accessing temperature is interior user resources request quantity for index between association area of unit interval.
10. method according to claim 7, also comprise the step of monitoring and upgrading resource status.
11. method according to claim 10 wherein only has when the property value of resource surpasses between the attribute area at its place and just need to upgrade the state of this resource.
12. a cloud resource scheduling system, described system comprises:
Be used for receiving the module from user's resource request;
Be used for each attribute for requested resource, extract respectively the constraint upper limit of each attribute and the module of constraint lower limit;
Be used for to determine call number between the constraint upper limit of each attribute of requested resource and the logic area under the constraint lower limit, be met the module of index between a plurality of association areas of this resource request;
Be used for selecting index between an association area from index between described a plurality of association areas, and select a resource to offer this user's module between selected association area the corresponding logical resource of index pond.
13. method according to claim 12 also comprises the module for monitoring and renewal resource status.
CN201210135357.1A 2012-05-02 2012-05-02 Cloud resource management and dispatching method and system Expired - Fee Related CN103383653B (en)

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 true CN103383653A (en) 2013-11-06
CN103383653B 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)

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104391697A (en) * 2014-11-11 2015-03-04 百度在线网络技术（北京）有限公司 System and method for managing cloud resources of application program
CN104679595A (en) * 2015-03-26 2015-06-03 南京大学 Application-oriented dynamic resource allocation method for IaaS (Infrastructure As A Service) layer
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
CN104881322A (en) * 2015-05-18 2015-09-02 中国科学院计算技术研究所 Method and device for dispatching cluster resource based on packing model
CN105204921A (en) * 2014-06-25 2015-12-30 北京奇虎科技有限公司 Task distribution method, device, system and server
CN106127382A (en) * 2016-06-22 2016-11-16 财付通支付科技有限公司 Method for managing resource and device
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
CN107276914A (en) * 2017-06-01 2017-10-20 千寻位置网络有限公司 The method of self-service resource allocation scheduling based on CMDB
CN107404527A (en) * 2017-07-26 2017-11-28 郑州云海信息技术有限公司 A kind of resource accesses and control method 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
CN108334409A (en) * 2018-01-15 2018-07-27 北京大学 A kind of fine-grained high-performance cloud resource management dispatching method
CN108572871A (en) * 2017-12-13 2018-09-25 北京金山云网络技术有限公司 A kind of resource allocation methods, device, electronic equipment and storage medium
CN108701061A (en) * 2016-02-25 2018-10-23 瑞典爱立信有限公司 The resources control of interconnection hardware infrastructure
CN109376006A (en) * 2018-09-04 2019-02-22 西安电子科技大学 Resource integrated method based on user demand time-varying characteristics under a kind of cloud computing environment
CN111694672A (en) * 2020-06-12 2020-09-22 北京字节跳动网络技术有限公司 Resource allocation method, task submission method, device, electronic equipment and medium
CN111934958A (en) * 2020-07-29 2020-11-13 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 (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080216081A1 (en) * 2005-03-11 2008-09-04 Cluster Resources, Inc. System and Method For Enforcing Future Policies in a Compute Environment
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
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

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20080216081A1 (en) * 2005-03-11 2008-09-04 Cluster Resources, Inc. System and Method For Enforcing Future Policies in a Compute Environment
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
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

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
JINBAO WANG，ET AL.: "Indexing Multi-dimensional Data in a Cloud System", 《PROCSSDINGS OF THE 2010 ACM SIGMOD INTERNATIONAL CONFERENCE ON MANAGEMENT OF DATA》 *
王国英: "基于P2P的资源搜索方法研究", 《广西大学硕士研究生学位论文》 *

Cited By (27)

* Cited by examiner, † Cited by third party
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
CN105204921A (en) * 2014-06-25 2015-12-30 北京奇虎科技有限公司 Task distribution method, device, system and server
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
CN104391697A (en) * 2014-11-11 2015-03-04 百度在线网络技术（北京）有限公司 System and method for managing cloud resources of application program
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
CN104679595A (en) * 2015-03-26 2015-06-03 南京大学 Application-oriented dynamic resource allocation method for IaaS (Infrastructure As A Service) layer
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
CN104881322A (en) * 2015-05-18 2015-09-02 中国科学院计算技术研究所 Method and device for dispatching cluster resource based on packing model
CN108701061B (en) * 2016-02-25 2022-02-18 瑞典爱立信有限公司 Resource control of interconnected hardware infrastructure
CN108701061A (en) * 2016-02-25 2018-10-23 瑞典爱立信有限公司 The resources control of interconnection hardware infrastructure
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
CN106127382A (en) * 2016-06-22 2016-11-16 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
CN107276914A (en) * 2017-06-01 2017-10-20 千寻位置网络有限公司 The method of self-service resource allocation scheduling based on CMDB
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
CN107404527A (en) * 2017-07-26 2017-11-28 郑州云海信息技术有限公司 A kind of resource accesses and control method and device
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
CN108572871A (en) * 2017-12-13 2018-09-25 北京金山云网络技术有限公司 A kind of resource allocation methods, device, electronic equipment and storage medium
CN108334409A (en) * 2018-01-15 2018-07-27 北京大学 A kind of fine-grained high-performance cloud resource management dispatching method
CN109376006A (en) * 2018-09-04 2019-02-22 西安电子科技大学 Resource integrated method based on user demand time-varying characteristics under a kind of cloud computing environment
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
CN111694672A (en) * 2020-06-12 2020-09-22 北京字节跳动网络技术有限公司 Resource allocation method, task submission method, device, electronic equipment and medium
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
CN111934958A (en) * 2020-07-29 2020-11-13 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

Also Published As

Publication number Publication date
CN103383653B (en) 2017-07-04

Similar Documents

Publication Publication Date Title
CN103383653A (en) Method and system for managing and dispatching cloud resource
CN100399301C (en) Mass storage device and method for dynamically managing a mass storage device
US8424059B2 (en) Calculating multi-tenancy resource requirements and automated tenant dynamic placement in a multi-tenant shared environment
CN107743611B (en) System, method, and computer storage medium for optimizing dynamic resource allocation
CN104050042B (en) The resource allocation methods and device of ETL operations
JP4614956B2 (en) Dynamic reassignment of data ownership
US9558207B1 (en) Versioning of database partition maps
US8386540B1 (en) Scalable relational database service
US9912741B2 (en) Optimization of computer system logical partition migrations in a multiple computer system environment
US20150113540A1 (en) Assigning resources among multiple task groups in a database system
US9390146B2 (en) Dynamic structural management of a distributed caching infrastructure
CN100428131C (en) Method for distributing resource in large scale storage system
CN110196851A (en) A kind of date storage method, device, equipment and storage medium
JP2005196602A (en) System configuration changing method in unshared type database management system
Chockler et al. Design and implementation of caching services in the cloud
CN104679594B (en) A kind of middleware distributed computing method
CN104731528B (en) A kind of construction method and system of cloud computing block storage service
CN103714098B (en) Method and system for carrying out subregion to database
CN104639594A (en) System and method for allocating physical resources and virtual resources
US20120265569A1 (en) Determining work placement preferences
CN103095788A (en) Cloud resource scheduling policy based on network topology
CN107480254B (en) Online load balancing method suitable for distributed memory database
CN102158533A (en) Distributed web service selection method based on QoS (Quality of Service)
CN104035807A (en) Metadata cache replacement method of cloud storage system
CN104765572B (en) The virtual storage server system and its dispatching method of a kind of energy-conservation

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
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20170704

Termination date: 20200502