CN102426542B - Resource management system for data center and operation calling method - Google Patents
Resource management system for data center and operation calling method Download PDFInfo
- Publication number
- CN102426542B CN102426542B CN 201110335469 CN201110335469A CN102426542B CN 102426542 B CN102426542 B CN 102426542B CN 201110335469 CN201110335469 CN 201110335469 CN 201110335469 A CN201110335469 A CN 201110335469A CN 102426542 B CN102426542 B CN 102426542B
- Authority
- CN
- China
- Prior art keywords
- group
- resource
- data center
- scheduling
- job scheduling
- 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.)
- Active
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a resource management for a data center and an operation calling method applied to the system. The system comprises a plurality of physical nodes and a data center resource manager, which are arranged in the data center, wherein the data center resource manager is used for managing and distributing resources on the physical nodes by units of groups, and operation of each group can operate on any physical node meeting operation operating requirements in the data center; when load is high, the plurality of groups can proportionally share resources in the data center; when load is low, the groups with operation to be operated can proportionally apply for idle resources according to resource quota so as to send operation as much as possible; therefore, the resource utilization rate of the data center is improved. In addition, the quick capacity expansion and dynamic shrinkage of resources of multiple groups can be realized through modifying the resource quota and operation scheduling strategies of the groups; therefore, diversity and time for handling emergency, which are required by application operation scheduling, can be realized.
Description
Technical field
The invention belongs to the computer systems technology field, relate in particular to the method that improves data center's resource sharing efficient.
Background technology
Data center (DataCenter) typically refers to the place that focuses on, stores, transmits, exchanges, manages of realizing information in a physical space, is comprised of one or more group of planes that is positioned at same machine room.A group of planes is to couple together by express network the computer system that consists of by numerous servers or workstation (workstation), and the server or the workstation that consist of a group of planes are called as node.
In recent years, the internet, applications development rapidly, present application type many, the large features such as (any active ues amount, visit capacity and business datum amounts) of application scale, make the data center's scale that supports internet, applications sharply enlarge, extend to several ten thousand or even hundreds of thousands platform scale by hundreds of platform, several thousand.And, due to the diversity of application load feature, diversity and the node resource dynamic in use of data center's node resource, brought very large difficulty for data center's cluster operation scheduling.How reasonably (for example: internet, applications), sharing efficiently group of planes resource, improve data center's resource utilization, is the key issue of data center's resource management in schedule job.
General resource management system for data center as shown in Figure 1, adopting the method for resource partitioning is a plurality of independently subregions (a plurality of independently little group of planes) with data center's resource division, feature according to application and resource, for each subregion (group of planes) arranges different job scheduling strategies, under certain rule, the part resource can be leased in different by stages.This method is simple, easily realization, but exists group of planes quantity many, small scale, and operation can not across the phenomenon of the effective shared resource of subregion, cause resource fragmentation; In addition, the resource share method of leasing formula lacks dirigibility and instantaneity, causes operation dilatation and capacity reducing fast, can not adapt to well the demand that current internet is used, and causes resource utilization low, and application cost is higher.
Summary of the invention
Therefore, the object of the invention is to overcome the defective of above-mentioned prior art, a kind of data centre dispatching method of taking into account fair and capacity is provided.
The objective of the invention is to be achieved through the following technical solutions:
on the one hand, the invention provides a kind of resource management system for data center, described system comprises a plurality of physical nodes and the data center resource manager that is positioned at data center, described data center resource manager is managed take group as unit and is distributed resource on physical node, resource on a plurality of physical nodes of one or more groups of shared data centers, each group has resource quota and job scheduling algorithm separately, the resource quota of described each group is the operable minimum resource usage quantity of this group, the operation of each group can run on any in data center and satisfy on the physical node that job run requires.
In said system, data center resource manager can manage group in the configuration file mode, in described configuration file, the attribute of putting for each assembly comprises: title, resource quota, the resource quantity seizing scheduling strategy, priority, used to user, job scheduling strategy, the operation of this group submit job of having the right.
In said system, the attribute that data center resource manager can be set up group, deletion group or change each group by revising configuration file.
In said system, user or use can be to a plurality of group submit jobs that the submit job authority is arranged wherein, but same operation energy and can only submit to a group.
In said system, wherein data center resource manager also is used for carrying out job scheduling.
On the other hand, the invention provides a kind of job scheduling method for said system, described method comprises:
Step 2, selection will participate in the group of job scheduling;
Step 3 is carried out job scheduling to selected group.
In said method, described step 1 can be calculated according to following formula the priority of each group:
The lowest common multiple of the resource quota of the priority of a group=all groups/this group has been used resource quantity.
In said method, also comprised the step that a plurality of groups are sorted from high to low according to priority before described step 2.
In said method, described step 2 selects to participate in the group of job scheduling according to following priority:
(1) select priority the highest and the group of waiting for operation arranged;
(2) select all groups higher than this group priority to wait for that number of jobs is 0, and this group is that the group of waiting for operation is arranged;
(3) select front priority high, participated in job scheduling but unsuccessful group, and this group has used resource less than the high group of the priority of its resource quota.
In said method, step 3) comprise
Step 31 is carried out job scheduling according to the job scheduling strategy of selected group;
Step 32 is if the job scheduling success goes to step 1; If job scheduling is unsuccessful, judges whether to carry out operation and seize scheduling;
Step 33 is carried out operation to selected group and is seized scheduling;
Step 34 is dispatched successfully if operation is seized, and sleeps a period of time, goes to step 1;
Otherwise, judge whether selected group be the minimum group of priority, if so, sleep a period of time, go to step 1, otherwise select next group, and whether judge its quantity of having used resource less than its resource quota, if less than, go to step 31, otherwise sleep a period of time, go to step 1.
In said method, step 32 can adopt following step to judge whether to carry out operation and seize scheduling:
When a group use resource less than the value of " its resource quota * first threshold " time, need to seize scheduling;
When a group use resource greater than the value of " resource quota * Second Threshold " time, it can be preempted scheduling;
Wherein the span of first threshold be (0,1], the span of Second Threshold be [1, natural number).
In said method, first threshold can be 0.9, and Second Threshold can be 1.1.
Compared with prior art, the invention has the advantages that:
Manage with Resources allocation take group as unit and carry out job scheduling, the operation of each group can run on any one of data center to be satisfied on the node that job run required, had enough idling-resources, does not therefore have the phenomenon of division of resources and resource fragmentation.When load is high, share pro rata the resource of data center according to the resource quota for a plurality of groups, when load is low, have the group of waiting for operation to apply for pro rata idling-resource by the resource quota, issue as much as possible operation, thereby improve data center's resource utilization.And can realize quick dilatation and the Dynamic contraction of the resource of a plurality of groups by resource quota and the job scheduling strategy of modification group, satisfy diversity and the demand of reply burst time of application job dispatching requirement.
Description of drawings
Embodiments of the present invention is further illustrated referring to accompanying drawing, wherein:
Fig. 1 is the structural representation of existing resource management system for data center;
Fig. 2 is the structural representation according to the resource management system for data center of the embodiment of the present invention;
Fig. 3 is the process flow diagram according to data center's job scheduling method of 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.
Shown in Figure 2 is the configuration diagram of resource management system for data center according to an embodiment of the invention.This system comprises a plurality of physical nodes and the data center resource manager that is positioned at data center.Wherein, data center resource manager is distributed and manages the resource on a plurality of physical nodes and carry out job scheduling take group (Group) as unit.A plurality of groups of all resources of sharing data center.Can join certain resource quota (Resource Quota) and select the job scheduling strategy for each component.User or use can be to a plurality of group submit jobs that the submit job authority is arranged, but same operation energy and can only submit to a group.A group can represent user's group or application, and for example using can be as the IM instant messaging, happy farm, QQ game, the Internet service of Online Video etc.
Different from the subregion in existing resource management system for data center, the resource quota of distributing to a group is a logical concept, and expression guarantees to satisfy its operable minimum resources quantity, does not comprise the node listing that is exclusively used in this group.The operation of each group can run on any one of data center to be satisfied on the node that job run required, had enough idling-resources, does not therefore have the phenomenon of division of resources and resource fragmentation.And the subregion in existing system is physical concept, and subregion has the node of some separately usually, and is separate between subregion.
Wherein, data center's resource division commonly used mainly comprises following several form: (1) physical machine; (2) virtual machine, the computational resource for each virtual machine distributes mainly contains CPU, internal memory, disk and network; (3) groove (Slot) is divided into a plurality of grooves according to the CPU number of node; (4) Linux container (Container).The present invention can support multiple group of planes resource division mode, and namely the resource of any type of data center can manage and distribute according to method provided by the invention.
By the job scheduling method that hereinafter will introduce, when group of planes load is high, shares liberally group of planes resource according to the resource quota, satisfy the minimum resources demand of each group for a plurality of groups; When group of planes load is low, there is the group of waiting for operation to apply for pro rata group of planes idling-resource by the resource quota, issue as much as possible operation, thereby improve group of planes resource utilization, realize the scalable management of group of planes resource.
Data center resource manager can realize by the mode of configuration file (for example its can called after Group.conf) configuration and the management organized.According to this configuration file, data center resource manager reads and the configuration informations such as scheduling strategy and rights management are seized in resource quota, job scheduling strategy, the operation of each group of initialization.By revising configuration file, can dynamically organize a plurality of groups, reach the dirigibility of data center's resource management.Also can organize for each increases various attributes and authority, is for example the suitable job scheduling strategy of each group selection, realizes the variation of data center's resource management, satisfies the diversity of internet, applications job scheduling demand.
In one embodiment, the Group.conf file layout is for example: the configuration to each group comprises following key word (key word and content can be isolated with colon):
Name: the name of group, it can be by the character string of capitalization (A-Z), lowercase (a-z), numeral (0-9) and underscore (_) formation, but its maximum length is no more than constant MAXGROUPNAMELEN (maximum zone name length), and the name of any two groups can not be identical.
ResourceQuota: the resource quota of this group, namely should group minimum resources usage quantity.It must be the natural number greater than 0, and its span is (0, available resources quantity).
Users: have the right to the user profile of this group submit job, can following formal description:
" username|username " wherein: " | " expression " perhaps ".
SchedPolicy: the job scheduling strategy of this group.Can string representation, case-insensitive mainly contains: FIFO, Priority, Capacity, the job scheduling strategies such as BackFill.
PreemptiveSchedPolicy: this group seize the job scheduling strategy.Can string representation, case-insensitive mainly contains: lower priority job is preferential, the running job priority scheduling is seized the job scheduling strategy recently.
Priority: the priority of this group.(the priority computing method of the group of the discussion that can vide infra)
UsedResource: this organize all running job take the summation of resource quantity.
For the account form of the quantity of using resource, at present general has two kinds: (1) is directly calculated according to the number of CPU and is used resource quantity.(2) keystone resources computing method for example suppose that a station server has 4 CPU, 16G internal memory.The resource requirement of operation is 2CPU, the 4G internal memory.It has taken the CPU of whole service 1/2, and 1/4 internal memory is by the CPU calculating shared resource use amount of occupation proportion maximum, i.e. 1/2 station server;
Should point out that the present invention supports the multiple resources computing method, the quantity of using resource that obtains a group that requires in the present invention, as for its circular, the user can according to demand, select voluntarily.In addition, above the setting about the resource quota is relevant to the Resource Calculation method, if adopt the method for CPU number, its quota is as long as specify minimum spendable CPU number just passable; If adopt the keystone resources computing method, its resource quota need to be specified CPU, internal memory, disk and the network bandwidth.
When system starts, read the initial value that dispatching algorithm etc. is seized in resource quota, job scheduling algorithm and operation that configuration file obtains Group, but its value can on-the-fly modify when moving in system, for example can on-the-fly modify by API.During each job scheduling, calculate the priority of this group according to the new resource quota value of group, and configure according to the dispatching algorithm of this group, select different job scheduling algorithms.Each Group can select suitable job scheduling strategy and operation to seize scheduling strategy according to the application load feature, satisfies the diversity of internet, applications job scheduling demand.
Fig. 3 shows the process flow diagram of data center's job scheduling method according to an embodiment of the invention.The method has adopted the job scheduling method of two-stage.First order job scheduling according to the priority of each group with used resource situation, selects to participate in the group of job scheduling, guarantees a plurality of groups according to resource quota shared group of planes resource pro rata; Second level job scheduling according to the configuration of selected group, is selected the job scheduling algorithm, dispatches the wait operation of this group.Mainly comprise the following steps: step 1, at first calculate the priority of each group and according to priority sequence from high to low; Step 2, selection participates in the group of job scheduling; Step 3 according to the configuration of selected group, is carried out job scheduling.In addition, if the present load of a group of planes is high, resource is nervous, utilizes operation to seize scheduling and realizes that a plurality of groups are shared group of planes resource pro rata according to the resource quota.
In one embodiment, the priority of each group can by the resource quota of this group and it all running job take total resources and calculate, numerical value is larger, priority is lower, circular is as follows:
(1) the resource quota of a group is set;
(2) calculate the lowest common multiple of all group resource quotas;
That (3) calculates this group uses resource quantity;
(4) lowest common multiple of the resource quota of these group priority=all groups/this group has been used resource quantity; When this group was 0 with resource, its priority was fixed as 0.
In the priority algorithm of above-mentioned group, by asking lowest common multiple, then the number of resources of using divided by this group can guarantee that a plurality of group are by the shared cluster resource of resource quota equitable proportion.For instance, two group A and B are arranged, the resource quota of group A is 2, the resource quota of group B is 3, moves identical operation, and the resource hypothesis that each operation consumes is 1, so when A the running job number be 1, B running job number is 1 o'clock, and the priority ratio A of B is high, comes the front during job scheduling, but as B when the running job number has been 2, its priority value is 4, and the A priority value is 3, and A comes the front.Again for example, the resource quota of A and B is all operation of 1, A operation, and B does not have running job, and during job scheduling, the priority ratio A of B is high, and when the B running job was two, the priority ratio B of A was high.
As seen, above-mentioned priority computational algorithm about group by the ratio numerical value of using resource and its resource quota, sorts when sharing group of planes resource for a plurality of groups from low to high, has guaranteed to come the right that preferentially obtains job scheduling of front.The one-stop operation scheduling flow, can select the high group of priority to carry out job scheduling, if its job scheduling success, this job scheduling finishes, carry out job scheduling next time, if it does not wait for operation or its back group use resource less than its resource quota, can then dispatch the operation of next group.
In addition, in order to improve the efficient of algorithm, can according to priority sort from high to low, to facilitate the selection step of back.
In one embodiment, can select to carry out according to following priority the group of job scheduling: (1) selects priority the highest and the group of waiting for operation arranged; (2) be chosen in all high groups of this group front priority and wait for that number of jobs is 0, and this group is that the group of waiting for operation is arranged; (3) select the high group of front priority, participated in job scheduling, but job scheduling is unsuccessful, wait for that namely operation does not have the upper suitable computational resource of coupling, and this group has used resource less than its resource quota.Group in addition can not obtain the chance of job scheduling.
In one embodiment, can adopt job scheduling and operation to seize the mode that combines of scheduling and complete scheduling to operation.Carrying out operation, to seize the purpose of scheduling be to guarantee that a plurality of groups are shared cluster resources liberally, prevents that the group from the group of planes free time, issuing a lot of operations, and be all long operation.Group of planes busy, other group can be seized the operation that this group is seized in scheduling by operation, so that the operation of self is issued.
Should point out, the one-stop operation scheduling flow is only dispatched an operation of a group.As can be seen from the above, the one-stop operation scheduling flow is divided into two steps, and the one, selection will be carried out the group of job scheduling, and the 2nd, carry out job scheduling according to the group of selecting, but the job scheduling of this group is once only dispatched an operation.After a job scheduling success, the job scheduling flow process is completed, and can enter next job scheduling flow process, can recomputate the priority of each group, and these groups are resequenced.
More specifically, as shown in Figure 3, the method comprises the following steps:
A) put sign IsSched and equal False;
Wherein, the effect of this sign is to judge whether the job scheduling flow process has selected a group to carry out job scheduling.Usually, can there be three kinds of situations in the possibility of selecting a group to carry out job scheduling: (1) selects the highest group of priority, and this group has the wait operation, carries out job scheduling, and operation is resource matched, operation issues successfully, and this job scheduling flow process finishes; (2) preferentially select the high group of priority to carry out job scheduling, but find that these groups do not wait for operation, then dispatching priority is low, but the group of waiting for operation is arranged, and at this moment the IsSched value is false, because the group of front is not carried out job scheduling; (3) group that front priority is high has the wait operation, and job scheduling is unsuccessful, and this time, the IsSched value was true, and the job scheduling flow process can then be dispatched the group of back, but can add a condition, this group use resource less than its resource quota.The purpose of doing like this is: (a) guarantee the fairness between the job scheduling of a plurality of groups, preferentially obtain the right of job scheduling with the little group of resource/this ratio of resource quota; (b) guarantee the least resource quota of each group.In other words, even if the group job of front scheduling is unsuccessful, come the group of back, if it use resource less than its resource quota, its job scheduling of still having an opportunity to obtain.So not only can satisfy the least resource quota of each group, and when group of planes idling-resource, each is organized according to resource quota ratio, shares liberally.Namely in group of planes busy, share group of planes resource according to resource quota equitable proportion for a plurality of groups; In group of planes idle, the group of waiting for that operation need to issue is arranged, according to resource quota ratio, the fair idling-resource of sharing.
B) calculate the priority of each group;
C) according to priority sort from high to low for group;
D) select the highest group of priority;
E) do you judge that selecteed group has the wait operation? no, jump to step 1);
F) put sign IsSched and equal True;
G) select suitable job scheduling algorithm according to the configuration of selected group, carry out job scheduling; As what above introduce, job scheduling algorithm's initial value arranges in configuration file, but also can dynamically select.
H) do you judge the whether success of job scheduling of selected group? success jumps to step a);
I) do you judge that selected group of needs operation seize scheduling? do not need, jump to step 1);
Can be by two threshold values be set, A and B, judge whether a group needs to carry out operation and seize scheduling and whether can be preempted scheduling, wherein the span of A be (0,1], can be for example 0.9, the span of B is [1, natural number), can be for example 1.1, the purpose that two threshold value A and b are set is to produce shake for fear of seizing scheduling.When a group use resource less than the value of " its resource quota * A " time, need to seize scheduling.When a group use resource greater than the value of " resource quota * B " time, can be preempted scheduling.The value of A and B can be configured according to demand.Group is preempted after scheduling in a period of time (for example 20 seconds) and does not participate in job scheduling, that is to say that the job scheduling flow process can skip this group, directly dispatches the group of back, can prevent from like this that job scheduling from producing to shake.And a period of time after a group is preempted, for example one minute, if it use resource quantity greater than its resource quota, still do not participate in job scheduling.
J) scheduling is seized in the operation of carrying out this group;
K) do you judge whether successfully the Group operation seizes scheduling? success jumps to step o);
L) do you judge that this group is last group? to jump to step o);
M) select next group;
Does n) selecteed group of judgement equal True with resource quantity less than its resource quota or IsSched? be true, jump to step e);
O) sleep a period of time, go to step a);
Can find out from above-mentioned steps, the group of preferentially selecting has the operation of wait and the highest group of priority, moreover is to select to have to wait for operation and used resource to participate in job scheduling less than the group of its resource quota.System can also set a default set (Default Group), the resource quota of this group less (1%-5% that is about group of planes total resources).When group of planes load was high, because the resource quota of default set is few, priority was low, and it be easy to surpass its resource quota with resource quantity, can not get scheduling and carry out so the major part of default set is waited for operation; When group of planes load was low, default set can be applied for group of planes idling-resource pro rata by the resource quota, dispatches and carry out its wait operation, and its suitable user submits the operation that priority is low, there is no the deadline restriction to.
In addition, if a job scheduling success of selected group shows the available free resource of a group of planes, load is not high, enters immediately new job scheduling flow process, can effectively dispatch a large amount of little operations; Otherwise, scheduling flow sleep a period of time, avoid repeating invalid job scheduling.
And the system manager can seize in operation under the help of scheduling by the resource quota of more reorganizing, and realizes quick dilatation and the Dynamic contraction of the resource of a plurality of groups.
Application and the effect thereof of the system and method that provides in the above-described embodiments is provided the below by way of example.For example, due to popularizing of smart mobile phone, increasing people take bus or subway in, the login mobile phone is used (news, music, books etc.) pastime riding time makes the peak period that mobile phone is used concentrate on point seven thirty to nine of morning, five thirty thirty to seven of evening these two intervals.When people opened computer and start working, the pc clients such as portal website, E-mail address, happy farm and web game were used the peak period that has welcome them.Leisure was used and was entered peak period, as online game, Online Video etc. to ten one points at 8 o'clock in evening.So can be according to the difference of using peak period, setting is applied in the resource quota of different time, the resource quota that is applied in seven thirty to nine of morning such as the regulation mobile phone is 20,10 At All Other Times, the be on duty resource quota of period of news website is 50, other period is 20, the resource quota of online game is 50 the time in the evening, 20 At All Other Times, seize scheduling and use under the assistance of the services such as automatic deployment in operation, the use resource of each application realizes the quick dilatation of resource and the Dynamic contraction of each group along with the time changes different and difference.This is due to the resource quota of passing through by each group of dynamic change of period, and each group priority can change, and can seize on one's own initiative the resource of other group or be seized by other group.
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 (9)
1. resource management system for data center, described system comprises a plurality of physical nodes and the data center resource manager that is positioned at data center, described data center resource manager is managed take group as unit and is distributed resource on physical node, resource on a plurality of physical nodes of a plurality of groups of shared data centers, each group has resource quota and job scheduling algorithm separately, the resource quota of described each group is the operable minimum resource usage quantity of this group, do not comprise the physical node list that is exclusively used in this group, the operation of each group can run on any in data center and satisfy on the physical node that job run requires, wherein, the user can be to one or more groups of submit jobs, but same operation energy and can only submit to a group,
Described data center resource manager organizes according to the resource quota of group and each group that the resource quantity of use selects to participate in job scheduling, then according to the job scheduling algorithm of selected group, the operation of this group is dispatched.
2. system according to claim 1, wherein, data center resource manager manages group in the configuration file mode, in described configuration file, the entry of putting for each assembly comprises: title, resource quota, the resource quantity seizing scheduling strategy, priority, used to user, job scheduling strategy, the operation of this group submit job of having the right.
3. system according to claim 2, wherein, the attribute that data center resource manager is set up group, deletion group or changed each group by revising configuration file.
4. one kind is used for as the job scheduling method of above-mentioned arbitrary claim described system, described method is managed take group as unit and is distributed resource on physical node, resource on a plurality of physical nodes of a plurality of groups of shared data centers, each group has resource quota and job scheduling algorithm separately, the resource quota of described each group is the operable minimum resource usage quantity of this group, do not comprise the physical node list that is exclusively used in this group, the operation of each group can run on any in data center and satisfy on the physical node that job run requires, wherein, the user can be to one or more groups of submit jobs, but same operation energy and can only submit to a group, described method also comprises:
Step 1 is calculated the priority of each group, and wherein, the lowest common multiple of the resource quota of the priority of a group=all groups/this group has been used resource quantity;
Step 2, selection will participate in the group of job scheduling;
Step 3 is carried out job scheduling to selected group.
5. method according to claim 4, wherein, also comprised the step that a plurality of groups are sorted from high to low according to priority before described step 2.
6. method according to claim 4, wherein, described step 2 selects to participate in the group of job scheduling according to following priority:
(1) select priority the highest and the group of waiting for operation arranged;
(2) select to have the group of wait operation and all groups higher than this group priority to wait for that numbers of jobs are 0;
(3) select front priority high, participated in job scheduling but unsuccessful group, and this group has used resource less than the high group of the priority of its resource quota.
7. method according to claim 4, wherein, described step 3) comprises
Step 31 is carried out job scheduling according to the job scheduling strategy of selected group;
Step 32 is if the job scheduling success goes to step 1; If job scheduling is unsuccessful, judges whether to carry out operation and seize scheduling; Do not seize scheduling if do not need to carry out operation, go to step 1; Seize scheduling if need to carry out operation, go to step 33;
Step 33 is carried out operation to selected group and is seized scheduling; Step 34 is dispatched successfully if operation is seized, and sleeps a period of time, go to step 1, otherwise, judge whether selected group be the minimum group of priority, if so, sleep a period of time, go to step 1, otherwise select next group, and whether judge its quantity of having used resource less than its resource quota, if less than, go to step 31, otherwise sleep a period of time, go to step 1.
8. method according to claim 7, wherein, the step below step 32 adopts judges whether to carry out operation and seizes scheduling:
When a group use resource less than the value of " its resource quota * first threshold " time, need to seize scheduling;
When a group use resource greater than the value of " resource quota * Second Threshold " time, it can be preempted scheduling;
Wherein the span of first threshold be (0,1], the span of Second Threshold is that [1, n), n is natural number and greater than 1.
9. method according to claim 8, wherein first threshold is 0.9, Second Threshold is 1.1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110335469 CN102426542B (en) | 2011-10-28 | 2011-10-28 | Resource management system for data center and operation calling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110335469 CN102426542B (en) | 2011-10-28 | 2011-10-28 | Resource management system for data center and operation calling method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102426542A CN102426542A (en) | 2012-04-25 |
| CN102426542B true CN102426542B (en) | 2013-06-12 |
Family
ID=45960530
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110335469 Active CN102426542B (en) | 2011-10-28 | 2011-10-28 | Resource management system for data center and operation calling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102426542B (en) |
Families Citing this family (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722413B (en) * | 2012-05-16 | 2017-12-29 | 上海兆民云计算科技有限公司 | The distributed resource scheduling method that a kind of desktop cloud cluster uses |
| CN103279390B (en) * | 2012-08-21 | 2016-09-28 | 中国科学院信息工程研究所 | A kind of parallel processing system (PPS) towards little optimization of job |
| CN102945185B (en) * | 2012-10-24 | 2015-04-22 | 深信服网络科技（深圳）有限公司 | Task scheduling method and device |
| CN103841126B (en) * | 2012-11-20 | 2017-11-21 | 中国移动通信集团辽宁有限公司 | The concocting method and device of a kind of shared resource |
| CN103336714A (en) * | 2013-06-20 | 2013-10-02 | 北京奇虎科技有限公司 | Operation scheduling method and device |
| CN103491024B (en) * | 2013-09-27 | 2017-01-11 | 中国科学院信息工程研究所 | Job scheduling method and device for streaming data |
| CN104750558B (en) * | 2013-12-31 | 2018-07-03 | 伊姆西公司 | The method and apparatus that resource allocation is managed in quota system is layered |
| CN103945004B (en) * | 2014-05-06 | 2017-05-31 | 中国联合网络通信集团有限公司 | Data dispatching method and system between a kind of data center |
| CN105320565B (en) * | 2014-07-31 | 2018-11-20 | 中国石油化工股份有限公司 | A kind of computer scheduling of resource method for a variety of application software |
| CN106170769B (en) * | 2014-12-11 | 2020-01-21 | 华为技术有限公司 | Method, device and program product for determining virtual machine resources |
| CN107688492B (en) * | 2016-08-05 | 2021-11-09 | 阿里巴巴集团控股有限公司 | Resource control method and device and cluster resource management system |
| US10782757B2 (en) * | 2017-08-09 | 2020-09-22 | Equinix, Inc. | Rack level power control |
| CN109992373B (en) * | 2017-12-29 | 2021-04-09 | 华为技术有限公司 | Resource scheduling method, information management method and device and task deployment system |
| CN109086142B (en) * | 2018-09-25 | 2022-03-25 | 郑州云海信息技术有限公司 | Resource scheduling method and device based on Openlava |
| CN110198356A (en) * | 2019-06-10 | 2019-09-03 | 莫毓昌 | A kind of user's request scheduling mechanism based on mixed cloud |
| CN110515737A (en) * | 2019-09-02 | 2019-11-29 | 北京明略软件系统有限公司 | Data management task operation method and device |
| CN114500405A (en) * | 2021-12-27 | 2022-05-13 | 天翼云科技有限公司 | Resource allocation and acquisition method and device for multi-type service application |
| CN118175118A (en) * | 2022-12-08 | 2024-06-11 | 中国科学院声学研究所 | ICN (information and communication network) -based data center distributed resource scheduling method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1517872A (en) * | 2003-01-16 | 2004-08-04 | 国际商业机器公司 | Method and device for dynamic allocation of computer resource |
| CN101326489A (en) * | 2005-12-12 | 2008-12-17 | 微软公司 | OS mini-boot for running multiple environments |
-
2011
- 2011-10-28 CN CN 201110335469 patent/CN102426542B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1517872A (en) * | 2003-01-16 | 2004-08-04 | 国际商业机器公司 | Method and device for dynamic allocation of computer resource |
| CN101326489A (en) * | 2005-12-12 | 2008-12-17 | 微软公司 | OS mini-boot for running multiple environments |
Non-Patent Citations (1)
| Title |
|---|
| 汤小春，刘健.《基于元区间的云计算基础设施服务的资源分配额研究》.《计算机工程与应用》.2010,1－5. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102426542A (en) | 2012-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102426542B (en) | Resource management system for data center and operation calling method | |
| CN104092767B (en) | A kind of publish/subscribe system and its method of work for increasing message queue model | |
| CN111290854B (en) | Task management method, device, system, computer storage medium and electronic equipment | |
| Lu et al. | Join-idle-queue: A novel load balancing algorithm for dynamically scalable web services | |
| CN101146127B (en) | A client buffer update method and device in distributed system | |
| CN103516585A (en) | Method and system for distributing messages according to priorities | |
| CN105786603B (en) | Distributed high-concurrency service processing system and method | |
| CN103927225A (en) | Multi-core framework Internet information processing and optimizing method | |
| US9239736B2 (en) | System and method of providing a fixed time offset based dedicated co-allocation of a common resource set | |
| CN102981973B (en) | Perform the method for request within the storage system | |
| CN109152061A (en) | Channel agility method, apparatus, server and storage medium | |
| US9817698B2 (en) | Scheduling execution requests to allow partial results | |
| CN108510337B (en) | Order generation and inventory updating method, client and server | |
| Tao et al. | Job scheduling optimization for multi-user MapReduce clusters | |
| CN105302650A (en) | Dynamic multi-resource equitable distribution method oriented to cloud computing environment | |
| CN103685492B (en) | Dispatching method, dispatching device and application of Hadoop trunking system | |
| CN113419846A (en) | Resource allocation method and device, electronic equipment and computer readable storage medium | |
| Shi et al. | Multijob associated task scheduling for cloud computing based on task duplication and insertion | |
| CN111461583B (en) | Inventory checking method and device | |
| CN112764909B (en) | Sharing method and system based on cloud architecture workstation | |
| US10893015B2 (en) | Priority topic messaging | |
| CN103051511B (en) | A kind of processing method of message data and device | |
| CN117056064A (en) | Resource allocation method, device, server, storage medium and program product | |
| CN115550284A (en) | Message processing method, device and equipment | |
| CN114489978A (en) | Resource scheduling method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |