CN104270421B - A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed - Google Patents

A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed Download PDF

Info

Publication number
CN104270421B
CN104270421B CN201410466319.3A CN201410466319A CN104270421B CN 104270421 B CN104270421 B CN 104270421B CN 201410466319 A CN201410466319 A CN 201410466319A CN 104270421 B CN104270421 B CN 104270421B
Authority
CN
China
Prior art keywords
task
cloud platform
virtual
virtual machine
resource
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
Application number
CN201410466319.3A
Other languages
Chinese (zh)
Other versions
CN104270421A (en
Inventor
沈蒙
李凡
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.)
Beijing Institute of Technology BIT
Original Assignee
Beijing Institute of Technology BIT
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 Beijing Institute of Technology BIT filed Critical Beijing Institute of Technology BIT
Priority to CN201410466319.3A priority Critical patent/CN104270421B/en
Publication of CN104270421A publication Critical patent/CN104270421A/en
Application granted granted Critical
Publication of CN104270421B publication Critical patent/CN104270421B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources

Abstract

The present invention relates to a kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed, belong to network virtualization field.Tenant is according to own service demand, the virtual resource request for offering the challenge required, and the request is sent to cloud platform；After cloud platform is connected to the task requests from tenant, the virtual machine quantity and dispensing position of existing task are retrieved, according to the virtual resource demand of new task and the virtual resource allocation situation of existing associated task, constructs resource distribution grid.Calculate and dispatch the virtual resource quantity that the task is actually needed, choose the best alternatives；Finally determine whether to distribute virtual resource.The virtual resource that the present invention can be directed to tenant task is asked, and according to the resource service condition of existing cloud platform, is calculated and is taken the minimum task scheduling approach of resource, improves the resource utilization of cloud platform.

Description

A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed
Technical field
The present invention relates to a kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed, belongs to network virtualization neck Domain.
Background technology
With the appearance using Amazon EC2 as the public data center cloud platform of representative, increasing enterprise is by business Moved to by local data center on cloud platform.In order to support a large number of users to share hardware (such as CPU, storage, bandwidth) money Physical hardware resources are abstracted as virtual resource (such as virtual machine) by source, cloud platform operator generally use multi-tenant technology.Each User (also known as tenant) according to own service demand, can rent the virtual resource of different scales, and according to resource usage amount to Operator pay.Therefore, multi-tenant cloud platform has two big advantages：First, substantial amounts of tenant can be served, physics money is improved The utilization rate in source, so as to improve operator's income；Second, it is allowed to tenant according to business demand flexibly stretch virtual resource use Scale, farthest avoid the wasting of resources.
However, the significant challenge that multi-tenant platform faces is resource isolation.Existing business cloud platform can ensure to meet All kinds of calculating and storage resource (such as CPU, internal memory and memory space) needed for tenant, but any type of network is not provided Bandwidth guaranteed.Then, numerous tenants compete limited network bandwidth resources, cause to operate in the task completion time on virtual net It is unpredictable, have a strong impact on the end-user experience using its business.In the long run, this will weaken user by business migration To the enthusiasm of cloud platform, large-scale promotion and the commercialization of cloud platform are hindered to a certain extent.From this, design is a kind of high The cross-domain virtual network mapping method of effect has huge economy and social value.Therefore, we, which devise, supports Bandwidth guaranteed Multi-tenant cloud platform method for scheduling task, can both ensure the task completion time of tenant, and and can improves the resource profit of cloud platform With rate, cloud platform operator and the doulbe-sides' victory of tenant are realized.
The content of the invention
The present invention is based on following rational hypothesis：Server of cloud platform is connected using tree topology.When new task reaches When, if multi-tenant cloud platform can meet its resource requirement, the task is scheduled to be performed；Otherwise, the task is rejected.
Particular content includes：
Step 1: tenant, according to own service demand, the virtual resource for offering the challenge required is asked, and the request is sent To cloud platform：One complete task requests is by a four-tuple<N,Ba,Be,S>To state, wherein, N represents what task needed Virtual machine quantity, virtual machine are the collection of virtual computing resource (such as CPU) and virtual storage resource (such as memory size, memory capacity) Close, and in cloud platform virtual resource base unit.BaRepresent every virtual machine of internal virtual amount of bandwidth, the i.e. task Other virtual machines with belonging to the task communicate required amount of bandwidth；BeRepresent external Virtual amount of bandwidth, the i.e. task Every virtual machine communicated required amount of bandwidth with belonging to the virtual machine of other tasks；S is represented to have with the task and associated (i.e. Communication) relation other tasks set.
Step 2: after cloud platform is connected to the task requests from tenant, virtual machine quantity, the task of existing task are retrieved In the position of physical machine (such as server) deposited of every virtual machine, the internal virtual amount of bandwidth of existing task and existing The external Virtual amount of bandwidth of task.
Step 3: cloud platform is according to the virtual resource demand of new task and the virtual resource allocation feelings of existing associated task Condition, construct resource distribution grid.
The process of establishing of resource distribution grid is：If new task is designated as Q, and assumes that only have task P associates with Q.To appointing Physical link l in a cloud platform of anticipating,
1) existing virtual machine number of the task in l left subtree is calculated, then, remaining virtual machine is then located at its right son In tree, node corresponding to left subtree and each virtual machine construction that right subtree virtual machine sum is existing task.
2) it is node corresponding to new task Q each virtual machine construction.
3) add source node and destination node, wherein source node are located at link l left subtrees with existing task and new task respectively Virtual machine be connected, and the virtual machine that destination node is located at link l right subtrees with existing task and new task respectively is connected.
4) it is each two internal nodes of task creation and two external nodes.The virtual machine node of each task is distinguished Internal node and external node corresponding to being connected to, the capacity on side is virtual machine internal communication bandwidth quantity and PERCOM peripheral communication band Wide quantity.
5) internal node of same task is interconnected.The multiple tasks for the relation that (communicated) with association, its external node It is interconnected.So far, resource distribution grid construction finishes.
Step 4: cloud platform calculates the virtual resource quantity dispatched the task and be actually needed according to resource distribution grid, take Following policy selection optimal case V：Since the root node of tree topology, for its next layer of some independent subtree, selection The most scheme of virtual machine is accommodated in single subtree.Iteration is carried out successively, until leaf node.
Step 5: whether detection optimal case V is empty.If it is empty, cloud platform do not have enough schedule virtual resources this Business, the task are rejected；If non-NULL, cloud platform is that the task distributes virtual resource according to optimal case V.
Beneficial effect
The virtual resource that the present invention can be directed to tenant task is asked, according to the resource service condition of existing cloud platform, meter Calculate and take the minimum task scheduling approach of resource.Multi-tenant cloud platform method for scheduling task proposed by the present invention can both ensure The task completion time of tenant, and can improve the resource utilization of cloud platform, are advantageous to promote the large-scale application of cloud platform.
Brief description of the drawings
Fig. 1 flow charts；
Fig. 2 present invention applies example schematic diagram；
Fig. 3 dual channel model schematic diagrames；
Fig. 4 resource distribution grid schematic diagrames；
Fig. 5 tasks average completion time tests block diagram；
Fig. 6 cloud platforms resource utilization tests block diagram.
Embodiment
The present invention is specifically described below in conjunction with the accompanying drawings：
The specific workflow of the present invention is as shown in Figure 1.
As shown in Fig. 2 according to step 1 tenant according to own service demand, the virtual resource request for offering the challenge required, And the request is sent to cloud platform.One complete task requests is stated by a four-tuple, as shown in Figure 3；
After Step 2: cloud platform is connected to the task requests from tenant, existing task P virtual machine quantity is retrieved For NP, dispensing position AP, the internal virtual amount of bandwidth of existing task is BP a, the external Virtual amount of bandwidth of existing task is BP e.Wherein, APIt is that length is NPVector, the numbering for the physical machine that each element representation virtual machine is deposited.
According to Step 3: cloud platform according to the virtual resource demand of new task and the virtual resource allocation of existing associated task Situation, resource distribution grid is constructed, as shown in Figure 4.For the ease of statement, new task is referred to as Q.And assume to only have in step 2 and retrieve Existing task P associated with Q.To the physical link l in any one cloud platform, the foundation of resource distribution grid includes following step Suddenly：
(1) VM numbers of the calculating task P in l left subtree, has been designated as p1；So, remaining VM is then located at its right son In tree, p is designated as2。NP=p1+p2.For node corresponding to task P each VM constructions.
(2) Q is made to have q1And q2Individual VM is located in link l left subtree and right subtree respectively, NQ=q1+q2.For the every of task Q Node corresponding to individual VM constructions.
(3) add source node and destination node, wherein source node are located at the VM phases of link l left subtrees with task P and Q respectively Connect, and the VM that destination node is located at link l right subtrees with task P and Q respectively is connected.
(4) for each two internal nodes of task creation, (subscript is represented with intra, such as pi intraAnd po intra) and two it is outer (subscript is represented portion's node with intro, such as pi interAnd po inter).The VM nodes of each task are respectively coupled corresponding inside Node and external node, the capacity on side is VM intercommunications amount of bandwidth and PERCOM peripheral communication amount of bandwidth.
(5) internal node of same task is interconnected.The multiple tasks for the relation that (communicated) with association, its external node It is interconnected.So far, resource distribution grid construction finishes.
Above-mentioned construction process is equally applicable to the scene that multiple existing tasks associated with Q be present.
Cloud platform is according to resource distribution grid, bandwidth demands of the calculating task P and Q on link l.The mistake of computation bandwidth demand The problem of journey is equivalent to calculate the resource distribution grid max-flow.Some task Q scheduling scheme is given, if the program is to every chain The bandwidth demand on road is respectively less than or the available bandwidth equal to the link, then the program is referred to as feasible program.
According to Step 4: cloud platform according to all feasible programs, choose the best alternatives V, that is, determines the N of task Q requirementsQIt is individual The best placement of virtual machine, specifically includes following steps：
1) full null vector A is initializedQ, length NQ, physical machine that each virtual machine for logger task Q is deposited compiles Number.
2) a feasible vector FV is associated for each physical link l in cloud platforml, for representing in the link The virtual machine quantity that can be distributed in subtree.Each feasible vector has NQ+ 1, n-th (0≤n≤NQ) position be 1 represent can be at this N VM is distributed in subtree, being represented for 0 cannot.
3) a numerical value M is associated for each physical link ll, represent FVlIn can be the minimum virtual of task distribution Machine number.
4) link being connected with root node is traveled through, is that the subtree belonging to each of the links distributes MlIndividual virtual machine, and from NQIn subtract Remove Ml.If after the completion of traversal, NQIt is not zero, then remaining virtual machine is assigned in the most subtree of free virtual machine.
5) each link is successively traveled through, until leaf node (leaf node is physical machine).It will distribute to task Q's Leaf node numbering writes vectorial A successivelyQIn.
According to Step 5: detecting vectorial AQWith the presence or absence of neutral element.If in the presence of it is empty, cloud to illustrate optimal scheduling plan V Platform does not have enough resources to support task Q, and task requests are rejected：If being not present, illustrate that optimal scheduling plan V is present, And according to AQVirtual resource is distributed for task Q.
According to above-mentioned embodiment, the performance of the method for scheduling task (referred to as binary channels optimization method) is imitated Very, and with other two kinds can selection scheduling method compare.Two methods for contrast are briefly discussed below：
1) single channel method：Cloud platform describes the resource requirement of tenant using existing one channel model, and selection is first can Capable task scheduling approach is as optimal case.
2) dual channel approaches：Cloud platform describes the resource requirement of tenant using dual channel model, selects first feasible appoint Scheduling scheme be engaged in as optimal case.
Speed is reached by control task, it is possible to achieve different cloud platforms it is expected load.
Fig. 5 is that being obtained using different method for scheduling task for task has been averaged in the case where different cloud platforms it is expected load Into time simulation result.As can be seen that binary channels optimization method proposed by the present invention can effectively shorten the average completion of task Time, the amplitude of shortening it is expected the increase of load with cloud platform and expanded.
Fig. 6 is the cloud platform resource that is obtained using different method for scheduling task in the case where different cloud platform it is expected load Utilization rate simulation result.It can be found that binary channels optimization method proposed by the present invention can improve the resource utilization of cloud platform.
By experimental result as can be seen that multi-tenant cloud platform method for scheduling task proposed by the present invention can both ensure to rent The task completion time at family, and can improve the resource utilization of cloud platform.

Claims (1)

  1. A kind of 1. multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed, it is characterised in that：
    Step 1: tenant, according to own service demand, the virtual resource for offering the challenge required is asked, and the request is sent to cloud Platform：One complete task requests is by a four-tuple<N,Ba,Be,S>To state, wherein, N represents that task needs virtual Machine quantity, BaRepresent the internal virtual amount of bandwidth of every virtual machine；BeRepresent the external Virtual amount of bandwidth of every virtual machine；S Represent the set of other tasks that there is incidence relation with the task；
    Step 2: after cloud platform is connected to the task requests from tenant, retrieve in the virtual machine quantity for having task, the task The external Virtual of the position for the physical machine that every virtual machine is deposited, the internal virtual amount of bandwidth of existing task and existing task Amount of bandwidth；
    Step 3: cloud platform is according to the virtual resource demand of new task and the virtual resource allocation situation of existing associated task, structure Make resource distribution grid：
    New task is designated as Q, to the physical link l in any one cloud platform,
    1) existing virtual machine number of the task in l left subtree is calculated, then, remaining virtual machine is then located at its right subtree In, node corresponding to left subtree and each virtual machine construction that right subtree virtual machine sum is existing task；
    2) it is node corresponding to new task Q each virtual machine construction；
    3) add source node and destination node, wherein source node are located at the void of link l left subtrees with existing task and new task respectively Plan machine is connected, and the virtual machine that destination node is located at link l right subtrees with existing task and new task respectively is connected；
    4) it is respectively coupled for each two internal nodes of task creation and two external nodes, the virtual machine node of each task To corresponding internal node and external node, the capacity on side is virtual machine internal communication bandwidth quantity and PERCOM peripheral communication bandwidth number Amount；
    5) internal node of same task is interconnected, there are the multiple tasks of incidence relation, its external node is interconnected；
    Step 4: cloud platform calculates the virtual resource quantity dispatched the task and be actually needed according to resource distribution grid, take as follows Policy selection optimal case V：Since the root node of tree topology, for its next layer of some independent subtree, select single The most scheme of virtual machine is accommodated in subtree, iteration is carried out successively, until leaf node；
    Step 5: whether detection optimal case V is empty, if it is empty, cloud platform does not have enough schedule virtual resources task, should Task is rejected；If non-NULL, cloud platform is that the task distributes virtual resource according to optimal case V.
CN201410466319.3A 2014-09-12 2014-09-12 A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed Expired - Fee Related CN104270421B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410466319.3A CN104270421B (en) 2014-09-12 2014-09-12 A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410466319.3A CN104270421B (en) 2014-09-12 2014-09-12 A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed

Publications (2)

Publication Number Publication Date
CN104270421A CN104270421A (en) 2015-01-07
CN104270421B true CN104270421B (en) 2017-12-19

Family

ID=52161910

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410466319.3A Expired - Fee Related CN104270421B (en) 2014-09-12 2014-09-12 A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed

Country Status (1)

Country Link
CN (1) CN104270421B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9516552B2 (en) * 2015-01-28 2016-12-06 International Business Machines Corporation Ensuring quality of bandwidth in a Wi-Fi connection to an internet access point
CN105159780B (en) * 2015-08-21 2018-07-20 北京理工大学 The high availability virtual network mapping method and device of oriented multilayer time cloud application
CN105224392B (en) * 2015-10-13 2018-07-27 中国联合网络通信集团有限公司 A kind of virtual computing resource quota management method and platform
CN105915470B (en) * 2016-01-27 2019-05-21 无锡华云数据技术服务有限公司 A kind of elastic bandwidth configuration method based on Linux flow control
CN105577834B (en) * 2016-02-06 2018-10-16 清华大学 Two layers of bandwidth allocation methods of cloud data center with Predicable performance and system
US10609129B2 (en) 2016-02-19 2020-03-31 Huawei Technologies Co., Ltd. Method and system for multi-tenant resource distribution
CN105871751A (en) * 2016-03-25 2016-08-17 中国科学院计算技术研究所 Method and system for guaranteeing network bandwidth of data center
CN107276801A (en) * 2017-06-14 2017-10-20 中国石油化工股份有限公司 A kind of collocation method of the cloud computing platform based on application service
CN109327422B (en) * 2017-08-01 2021-04-02 中国移动通信集团浙江有限公司 Multi-tenant isolation method and device
CN108665157B (en) * 2018-05-02 2021-08-20 中山大学 Method for realizing balanced scheduling of cloud workflow system process instance
CN111199033A (en) * 2020-01-09 2020-05-26 山东浪潮通软信息科技有限公司 Method and tool for processing identity card and certificate information by using cloud technology
CN112073501A (en) * 2020-09-02 2020-12-11 浪潮云信息技术股份公司 Tenant separation type storage and management method

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103747059A (en) * 2013-12-26 2014-04-23 华中科技大学 Method and system for guaranteeing cloud computing server cluster network
CN103812949A (en) * 2014-03-06 2014-05-21 中国科学院信息工程研究所 Task scheduling and resource allocation method and system for real-time cloud platform

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8667171B2 (en) * 2010-05-28 2014-03-04 Microsoft Corporation Virtual data center allocation with bandwidth guarantees
US8671407B2 (en) * 2011-07-06 2014-03-11 Microsoft Corporation Offering network performance guarantees in multi-tenant datacenters

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103747059A (en) * 2013-12-26 2014-04-23 华中科技大学 Method and system for guaranteeing cloud computing server cluster network
CN103812949A (en) * 2014-03-06 2014-05-21 中国科学院信息工程研究所 Task scheduling and resource allocation method and system for real-time cloud platform

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
Network Sharing in Multi-tenant Datacenters;Hitesh Ballani等;《https://www.microsoft.com/en-us/research/publication/network-sharing-in-multi-tenant-datacenters/》;20120201;第1页第1节，第3页第3.1节-第7页第4.1.3节，附图2-6 *

Also Published As

Publication number Publication date
CN104270421A (en) 2015-01-07

Similar Documents

Publication Publication Date Title
CN104270421B (en) A kind of multi-tenant cloud platform method for scheduling task for supporting Bandwidth guaranteed
CN105103506B (en) For the method and system for the non-homogeneous bandwidth request allocation bandwidth in system for cloud computing
CN104410672B (en) Method, the method and device of forwarding service of network function virtualization applications upgrading
CN103827825B (en) Virtual resource object component
CN104881325B (en) A kind of resource regulating method and resource scheduling system
CN103607466B (en) A kind of wide-area multi-stage distributed parallel grid analysis method based on cloud computing
WO2016101638A1 (en) Operation management method for electric power system cloud simulation platform
CN105468435A (en) NFV dynamic resource distribution method
CN103885831B (en) The system of selection of virtual machine host machine and device
CN104503832B (en) A kind of scheduling virtual machine system and method for fair and efficiency balance
CN109962940A (en) A kind of virtualization example scheduling system and dispatching method based on cloud platform
CN105610715B (en) A kind of cloud data center multi-dummy machine migration scheduling method of planning based on SDN
CN103336722B (en) A kind of CPU resources of virtual machine monitoring and dynamic allocation method
CN104683488B (en) Streaming computing system and its dispatching method and device
CN104935628B (en) A kind of method that multiple associated virtual machines are migrated between multiple data centers
CN103023980B (en) A kind of method and system of cloud platform processes user service request
CN103560967A (en) Virtual data center mapping method sensing service requirement
CN107346264A (en) A kind of method, apparatus and server apparatus of virtual machine load balance scheduling
CN105069702B (en) A kind of power grid integrated information processing method
CN102637138B (en) A kind of method for calculating scheduling virtual machine
CN108370328A (en) A kind of management method and device of NFV MANO policy depictions symbol
CN107133083A (en) A kind of virtual data center resource provider method based on virtualization technology
CN106909462A (en) A kind of cloud resource regulating method and device
CN106161603B (en) A kind of method of networking, equipment and framework
CN108900435A (en) A kind of method, apparatus and computer storage medium of service deployment

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: 20171219

Termination date: 20210912