CN101943943A - Method for optimizing energy consumption of computation array based on similar resource aggregation - Google Patents
Method for optimizing energy consumption of computation array based on similar resource aggregation Download PDFInfo
- Publication number
- CN101943943A CN101943943A CN 201010281682 CN201010281682A CN101943943A CN 101943943 A CN101943943 A CN 101943943A CN 201010281682 CN201010281682 CN 201010281682 CN 201010281682 A CN201010281682 A CN 201010281682A CN 101943943 A CN101943943 A CN 101943943A
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- node
- control module
- unit
- dormancy
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention discloses a method for optimizing the energy consumption of a computation array based on similar resource aggregation and aims to reduce the energy consumption of the computation array. The technical scheme is that: the method comprises the following steps of: defining a node energy consumption control unit, a blade energy consumption control unit and a cabinet energy consumption control unit; constructing a system for optimizing the energy consumption of the computation array, which consists of a similar resource aggregation module and an energy consumption state control module; allocating idle computation nodes and dormant computation nodes according to the principle of keeping resources in the similar energy consumption states aggregated on the same energy consumption control unit and allocating computation nodes according to the levels of the energy consumption control units from high level to low level by the similar resource aggregation module; and implementing energy consumption optimizing control, detecting the states of the energy consumption control units, closing the energy consumption control units in the consistent internal energy consumption states and starting the energy consumption control units to which operation is allocated by the energy consumption state control module. The method can keep the aggregation characteristic of the energy consumption units in the similar energy consumption states and improve the probability of closing the energy consumption control units so as to reduce the energy consumption.
Description
Technical field
The present invention relates to the energy optimization method of high performance computing system computing array, especially a kind of method that cuts down the consumption of energy by similar resource polymerization.
Background technology
Along with the continuous development of high performance computing system technology, the development of large-scale parallel computing system of new generation faces many new challenges.Particularly along with the continuous lifting of system performance, scale is more and more huger, and the energy consumption of system rapidly increases, and has caused the energy consumption crisis of large-scale computing systems.The ratio of system energy consumption and system performance is just becoming one of main evaluation index of high-performance calculation.
High performance computing system generally includes front end services array, computing array, three ingredients of rear end storage array.Computing array is made up of a large amount of calculating nodes, and the execution service of operation externally is provided.Computing array is the emphasis of energy optimization research and engineering design as the main power consumption parts of high-performance calculation array.Computing array is made up of system's component units of a plurality of different stages usually, and common system's component units is rudimentary to be comprised to senior: 1) node unit: share the calculating node of an operating system, comprise 1 to a plurality of CPU, share main memory between the CPU.The node unit can place the S3 dormant state and wake up from the S3 dormant state; 2) blade unit: be made up of C1 node unit, the value of C1 is 2 integer power
,
xBe natural number.Blade power supply and other cooling fan of blade level are shared in node unit in the blade unit.When all the node unit in and if only if the blade unit are in dormant state, can close the power supply and blower of blade unit, make whole blade unit be in closed condition, reach energy-conservation purpose; 3) rack unit: be made up of C2 blade unit, the value of C2 is 2 integer power
,
yBe natural number.Blade unit in the rack unit is shared rack power supply, other cooling infrastructure of rack level, communication module.When all blade units were in closed condition in and if only if the rack unit, power supply, cooling, the communication module that can close the rack unit made whole rack be in closed condition, reach energy-conservation purpose.
Common computing array energy optimization method comprises at present:
1) based on the energy optimization method of CPU frequency modulation.When cpu load is high, heighten the clock frequency of CPU, satisfy the performance requirement of using.When cpu load is low, turn down the clock frequency of CPU, when satisfying the application performance requirement, reduce the energy consumption of processor.Energy consumption when on the main computation optimization node of this method task run being arranged.When calculating node in the time of the free time, still power consumption is bigger to drop to the CPU of low-limit frequency, has bigger optimization space.
2) close the power-economizing method of idle node.Because the unbalanced characteristic that user job is submitted to and the part order-preserving characteristic of job scheduling strategy exist a large amount of free time to calculate node in the computing array operational process.When calculating after node free time reaches parameter preset, idle node is placed the S3 dormant state.The S3 dormant state is a kind of low power consumpting state of the computing system of ACPI (Advanced Configuration and Power Interface Specification) standard definition, compares the energy consumption that open state can be saved 90%-95%.This method is the power-economizing method of a kind of node unit, does not consider the energy optimization of system's component units of higher level.
The energy optimization method of high performance computing system computing array is the technical matters that those skilled in the art pay close attention to.Thereby there is not open source literature to relate to the method that similar resource polymerization is cut down the consumption of energy in the research of existing high performance computing system computing array energy consumption.
Summary of the invention
The technical problem to be solved in the present invention is: at the energy optimization problem of high performance computing system computing array, a kind of energy optimization method based on similar resource polymerization is proposed, keep the computational resource that places dormancy or closed condition as much as possible, reach the purpose that reduces the computing array energy consumption.
In order to solve the problems of the technologies described above, technical scheme of the present invention is: the energy consumption control module in the hierarchical definition computing array, the resource that power consumption state is similar is aggregated in the energy consumption control module in the computing array operational process, and increase can place the number of the energy consumption control module of dormancy or closed condition.The power consumption state of energy consumption control module comprises duty, idle condition, closed condition, S3 dormant state.The similar energy consumption control module that is meant of so-called power consumption state is in identical or similar power consumption state, and closed condition is similar power consumption state with the S3 dormant state.
Concrete technical scheme is:
The first step, according to energy consumption control module in system's component units definition computing array of computing array, the energy consumption control module comprises from the subordinate to higher level: node energy consumption control module, blade energy consumption control module, rack energy consumption control module.For the higher level's energy consumption control module that comprises subordinate's energy consumption control module, only all be in consistent dormancy or during the low power consumpting state of closing, higher level's energy consumption control module just can place closed condition when its subordinate's energy consumption control module.In system's operational process, keep the similar resource of power consumption state to be aggregated in same energy consumption control module as far as possible, can place the number of the energy consumption control module of dormancy or closed condition, reduce system's operation energy consumption with effective increase.
Second step, structure computing array energy optimization system.The energy optimization system is deployed in the software package on the server in the front end services array of high performance computing system, is made up of similar resource polymerization module and power consumption state control module.Similar resource polymerization module is carried out the energy optimization decision-making, assembles the energy consumption control module with similar power consumption state in system's operational process, for the power consumption state control module provides the energy consumption control module that places dormancy or closed condition as much as possible.The power consumption state control module is implemented energy optimization control, detects the state of energy consumption control module, closes the energy consumption control module of inner power consumption state unanimity, opens and has distributed the energy consumption control module of operation.
In the 3rd step, similar resource polymerization module receives the operation that client is submitted to, and it is M that the node demand is calculated in this operation, and M is a positive integer.
The 4th step, similar resource polymerization module retrieves the calculating node of all idle conditions from computing array resource status storehouse, distribute the idle node that calculates according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, calculate node according to the order assignment of level from higher level to the subordinate of energy consumption control module.Method is:
4.1 initialization operation calculating node quantity to be allocated is k, k=M.
Be formation jqueue 4.2 all rack unit are sorted from high to low according to the quantity of idle node, the identical rack unit of idle node quantity sorts from low to high according to the average of the free time of all idle nodes in the rack unit.Other resource selection of rack level is from the jqueue queue heads.The initial value of current rack unit pointer j is the rack unit of jqueue queue heads, and the idle node quantity of rack unit j is jtmp.
If 4.3 k 〉=jtmp and jtmp〉0, changeed for 4.4 steps, otherwise changeed for 4.5 steps.
4.4 give current operation, k=k-jtmp with the whole idle node that calculates of current rack unit j.The j pointer towards rack unit of jqueue rear of queue direction reach, was changeed for 4.3 steps.
Be the dqueue of formation 4.5 the blade unit that rack unit j is comprised sorts from high to low according to the quantity of idle node, the identical blade unit of idle node quantity sorts from low to high according to the average of the free time of all idle nodes in the blade unit.The initial value of current blade unit pointer d is the blade unit of dqueue queue heads, and the idle node quantity of blade unit d is dtmp.
If 4.6 k 〉=dtmp and dtmp〉0, changeed for 4.7 steps, otherwise changeed for 4.8 steps.
4.7 give current operation, k=k-dtmp with the whole idle node that calculates of current blade unit d.The d pointer is moved a blade unit towards dqueue rear of queue direction, changeed for 4.6 steps.
Be formation nqueue 4.8 the node unit that blade unit d is comprised sorts from low to high according to the size of free time, the node unit of queue heads is designated as n, and active node element number is ntmp.
If 4.9 k〉0 and k＜ntmp, changeed for 4.10 steps, otherwise changeed for the 5th step.
4.10 give current operation, k=0 with preceding k the idle node of blade unit d.
In the 5th step, check whether the resource allocation request of operation is satisfied.If k=0, i.e. the resource allocation request of operation is satisfied, changes for the 7th step.
The 6th step, similar resource polymerization module retrieves the calculating node of all dormant states from computing array resource status storehouse, distribute dormancy to calculate node according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, calculate node according to the order assignment of level from higher level to the subordinate of energy consumption control module.Method is:
Be formation jsqueue 6.1 all rack unit are sorted from low to high according to the quantity of dormancy node, the identical rack unit of dormancy node quantity sorts from high to low according to the average of the dormancy time of all dormancy nodes in the rack unit.Other resource selection of rack level is from the jsqueue queue heads.The initial value of current rack unit pointer js is the rack unit of jsqueue queue heads, and the dormancy node quantity of rack unit js is jstmp.
If 6.2 k 〉=jstmp and jstmp〉0, then changeed for 6.3 steps, otherwise changeed for 6.4 steps.
6.3 being calculated node, whole dormancy of current rack unit js give current operation, k=k-jstmp.The js pointer towards rack unit of jsqueue rear of queue direction reach, was changeed for 6.2 steps.
Be the dsqueue of formation 6.4 the blade unit that rack unit js is comprised sorts from low to high according to the quantity of dormancy node, the identical blade unit of dormancy node quantity sorts from high to low according to the average of the dormancy time of all dormancy nodes in the blade unit.The initial value of current blade unit pointer ds is the blade unit of dsqueue queue heads, and the dormancy node quantity of blade unit ds is dstmp.
If 6.5 k 〉=dstmp and dstmp〉0, then changeed for 6.6 steps, otherwise changeed for 6.7 steps.
6.6 being calculated node, whole dormancy of current blade unit ds give current operation, k=k-dstmp.The d pointer is moved a blade unit towards dsqueue rear of queue direction, changeed for 6.5 steps.
Be formation nsqueue 6.7 the node unit that blade unit ds is comprised sorts from high to low according to the size of dormancy time, the node unit of queue heads is designated as ns, and the node element number of dormancy is nstmp.
If 6.8 k〉0 and k＜nstmp, changeed for 6.9 steps, otherwise changeed for the 7th step.
6.9 give current operation, k=0 with preceding k the dormancy node of blade unit ds.
In the 7th step, the power consumption state control module is waken the energy consumption control module that is in dormancy or closed condition that distributes operation up according to the order of rack energy consumption control module, blade energy consumption control module, node energy consumption control module.If distributed the rack unit at the calculating node place of operation to be in closed condition, called the high performance computing system Control Software and wake the rack unit up.If distributed the blade unit at the calculating node place of operation to be in closed condition, called the high performance computing system Control Software and wake blade unit up.If distributed the calculating node of operation to be in dormant state, the ACPI control interface that the call operation system provides wakes the dormant state that distributes operation up and calculates node.The high performance computing system Control Software is the application software that is used to open or close the energy consumption unit in the high performance computing system, is to be deployed in the software package on the server in the front end services array of high performance computing system.
In the 8th step, the power consumption state control module is every the state of time interval T detection energy consumption control module, and the ACPI control interface of call operation system calculates node to Idle state and carries out sleep operation.The span of time interval T is 1 minute to 3 minutes.High performance computing system Control Software execution blade shutoff operation is called in all dormancy of all node unit in the blade unit.All blade units are buttoned-up in the rack unit, call the high performance computing system Control Software and carry out the rack shutoff operation.
Compared with prior art, adopt the present invention can reach following technique effect:
1. the 4th step of the present invention and the 6th step are supported the level Distribution Calculation node of the present invention according to the energy consumption control module, have kept the polymerization property of the similar energy consumption unit of power consumption state, have improved the probability that the energy consumption control module is closed;
2. the present invention supports the idle node of priority allocation of the present invention the 4th step, has reduced the power consumption state switching times that calculates node;
3. the present invention supported the present invention preferentially to select short node distribution of free time in the 4th step, had increased the probability of node dormancy;
4. the present invention supported the present invention preferentially to select the long node of dormancy time to distribute in the 6th step, had kept the isotropism of node dormancy time.
Description of drawings
Fig. 1 is an energy optimization method flow diagram of the present invention.
Embodiment
Fig. 1 is an energy optimization method flow diagram of the present invention, may further comprise the steps:
The first step, according to energy consumption control module in system's component units definition computing array of computing array, the energy consumption control module comprises from the subordinate to higher level: node energy consumption control module, blade energy consumption control module, rack energy consumption control module.
Second step, structure computing array energy optimization system.The energy optimization system is deployed in the software package on the server in the front end services array of high performance computing system, is made up of similar resource polymerization module and power consumption state control module.
In the 3rd step, similar resource polymerization module receives the operation that client is submitted to, and it is M that the node demand is calculated in this operation, and M is a positive integer.
The 4th step, similar resource polymerization module retrieves the calculating node of all idle conditions from computing array resource status storehouse, distribute the idle node that calculates according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, calculate node according to the order assignment of level from higher level to the subordinate of energy consumption control module.
Whether the resource allocation request of the 5th step, inspection operation is satisfied.If the resource allocation request of operation is satisfied, changeed for the 7th step.
The 6th step, similar resource polymerization module retrieves the calculating node of all dormant states from computing array resource status storehouse, distribute dormancy to calculate node according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, calculate node according to the order assignment of level from higher level to the subordinate of energy consumption control module.
In the 7th step, the power consumption state control module is waken the energy consumption control module that is in dormancy or closed condition that distributes operation up according to the order of rack energy consumption control module, blade energy consumption control module, node energy consumption control module.
The 8th step, power consumption state control module are detected the state of energy consumption control module every time interval T, and the ACPI control interface of call operation system calculates node to Idle state and carries out sleep operation.High performance computing system Control Software execution blade shutoff operation is called in all dormancy of all node unit in the blade unit.All blade units are buttoned-up in the rack unit, call the high performance computing system Control Software and carry out the rack shutoff operation.
Claims (2)
1. computing array energy optimization method based on similar resource polymerization is characterized in that may further comprise the steps:
The first step, according to energy consumption control module in system's component units definition computing array of computing array, the energy consumption control module comprises from the subordinate to higher level: node energy consumption control module, blade energy consumption control module, rack energy consumption control module; For the higher level's energy consumption control module that comprises subordinate's energy consumption control module, only all be in consistent dormancy or during the low power consumpting state of closing, higher level's energy consumption control module just can place closed condition when its subordinate's energy consumption control module;
In second step, structure computing array energy optimization system, energy optimization system are deployed in the software package on the server in the front end services array of high performance computing system, are made up of similar resource polymerization module and power consumption state control module; Similar resource polymerization module is carried out the energy optimization decision-making, assembles the energy consumption control module with similar power consumption state in system's operational process, for the power consumption state control module provides the energy consumption control module that places dormancy or closed condition as much as possible; The power consumption state control module is implemented energy optimization control, detects the state of energy consumption control module, closes the energy consumption control module of inner power consumption state unanimity, opens and has distributed the energy consumption control module of operation;
In the 3rd step, similar resource polymerization module receives the operation that client is submitted to, and it is M that the node demand is calculated in this operation, and M is a positive integer;
The 4th step, similar resource polymerization module retrieves the calculating node of all idle conditions from computing array resource status storehouse, distribute the idle node that calculates according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, the order assignment of level from higher level to the subordinate according to the energy consumption control module calculated node, and method is:
4.1 initialization operation calculating node quantity to be allocated is k, k=M;
Be formation jqueue 4.2 all rack unit are sorted from high to low according to the quantity of idle node, the identical rack unit of idle node quantity sorts from low to high according to the average of the free time of all idle nodes in the rack unit; Other resource selection of rack level is from the jqueue queue heads; The initial value of current rack unit pointer j is the rack unit of jqueue queue heads, and the idle node quantity of rack unit j is jtmp;
If 4.3 k 〉=jtmp and jtmp〉0, changeed for 4.4 steps, otherwise changeed for 4.5 steps;
4.4 give current operation with the whole idle node that calculates of current rack unit j, k=k-jtmp towards rack unit of jqueue rear of queue direction reach, changeed for 4.3 steps with the j pointer;
4.5 sorting from high to low according to the quantity of idle node, the blade unit that rack unit j is comprised is the dqueue of formation, the identical blade unit of idle node quantity sorts from low to high according to the average of the free time of all idle nodes in the blade unit, the initial value of current blade unit pointer d is the blade unit of dqueue queue heads, and the idle node quantity of blade unit d is dtmp;
If 4.6 k 〉=dtmp and dtmp〉0, changeed for 4.7 steps, otherwise changeed for 4.8 steps;
4.7 give current operation with the whole idle node that calculates of current blade unit d, k=k-dtmp moves a blade unit with the d pointer towards dqueue rear of queue direction, changes for 4.6 steps;
Be formation nqueue 4.8 the node unit that blade unit d is comprised sorts from low to high according to the size of free time, the node unit of queue heads is designated as n, and active node element number is ntmp;
If 4.9 k〉0 and k＜ntmp, changeed for 4.10 steps, otherwise changeed for the 5th step;
4.10 give current operation, k=0 with preceding k the idle node of blade unit d;
In the 5th step, check whether the resource allocation request of operation is satisfied, if k=0 changeed for the 7th step;
The 6th step, similar resource polymerization module retrieves the calculating node of all dormant states from computing array resource status storehouse, distribute dormancy to calculate node according to the principle that keeps the similar resource of power consumption state to be aggregated in same energy consumption control module, the order assignment of level from higher level to the subordinate according to the energy consumption control module calculated node, and method is:
Be formation jsqueue 6.1 all rack unit are sorted from low to high according to the quantity of dormancy node, the identical rack unit of dormancy node quantity sorts from high to low according to the average of the dormancy time of all dormancy nodes in the rack unit, and other resource selection of rack level is from the jsqueue queue heads; The initial value of current rack unit pointer js is the rack unit of jsqueue queue heads, and the dormancy node quantity of rack unit js is jstmp;
If 6.2 k 〉=jstmp and jstmp〉0, then changeed for 6.3 steps, otherwise changeed for 6.4 steps;
Give current operation 6.3 node is calculated in whole dormancy of current rack unit js, k=k-jstmp towards rack unit of jsqueue rear of queue direction reach, changeed for 6.2 steps with the js pointer;
6.4 sorting from low to high according to the quantity of dormancy node, the blade unit that rack unit js is comprised is the dsqueue of formation, the identical blade unit of dormancy node quantity sorts from high to low according to the average of the dormancy time of all dormancy nodes in the blade unit, the initial value of current blade unit pointer ds is the blade unit of dsqueue queue heads, and the dormancy node quantity of blade unit ds is dstmp;
If 6.5 k 〉=dstmp and dstmp〉0, changeed for 6.6 steps, otherwise changeed for 6.7 steps;
Give current operation 6.6 node is calculated in whole dormancy of current blade unit ds, k=k-dstmp moves a blade unit with the d pointer towards dsqueue rear of queue direction, changes for 6.5 steps;
Be formation nsqueue 6.7 the node unit that blade unit ds is comprised sorts from high to low according to the size of dormancy time, the node unit of queue heads is designated as ns, and the node element number of dormancy is nstmp;
If 6.8 k〉0 and k＜nstmp, changeed for 6.9 steps, otherwise changeed for the 7th step;
6.9 give current operation, k=0 with preceding k the dormancy node of blade unit ds;
In the 7th step, the power consumption state control module is waken the energy consumption control module that is in dormancy or closed condition that distributes operation up according to the order of rack energy consumption control module, blade energy consumption control module, node energy consumption control module; If distributed the rack unit at the calculating node place of operation to be in closed condition, called the high performance computing system Control Software and wake the rack unit up; If distributed the blade unit at the calculating node place of operation to be in closed condition, called the high performance computing system Control Software and wake blade unit up; If distributed the calculating node of operation to be in dormant state, the ACPI control interface that the call operation system provides wakes the dormant state that distributes operation up and calculates node; The high performance computing system Control Software is the application software that is used to open or close the energy consumption unit in the high performance computing system, is to be deployed in the software package on the server in the front end services array of high performance computing system;
In the 8th step, the power consumption state control module is every the state of time interval T detection energy consumption control module, and the ACPI control interface of call operation system calculates node to Idle state and carries out sleep operation; High performance computing system Control Software execution blade shutoff operation is called in all dormancy of all node unit in the blade unit; All blade units are buttoned-up in the rack unit, call the high performance computing system Control Software and carry out the rack shutoff operation.
2. the computing array energy optimization method based on similar resource polymerization as claimed in claim 1, the span that it is characterized in that time interval T is 1 minute to 3 minutes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010281682XA CN101943943B (en) | 2010-09-15 | 2010-09-15 | Method for optimizing energy consumption of computation array based on similar resource aggregation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201010281682XA CN101943943B (en) | 2010-09-15 | 2010-09-15 | Method for optimizing energy consumption of computation array based on similar resource aggregation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101943943A true CN101943943A (en) | 2011-01-12 |
| CN101943943B CN101943943B (en) | 2011-11-30 |
Family
ID=43435967
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201010281682XA Expired - Fee Related CN101943943B (en) | 2010-09-15 | 2010-09-15 | Method for optimizing energy consumption of computation array based on similar resource aggregation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101943943B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016058386A1 (en) * | 2014-10-17 | 2016-04-21 | 深圳市中兴微电子技术有限公司 | Power consumption management method and device and computer storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100153634A1 (en) * | 2008-12-12 | 2010-06-17 | Datadirect Networks, Inc. | System and method for data migration between computer cluster architecture and data storage devices |
-
2010
- 2010-09-15 CN CN201010281682XA patent/CN101943943B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100153634A1 (en) * | 2008-12-12 | 2010-06-17 | Datadirect Networks, Inc. | System and method for data migration between computer cluster architecture and data storage devices |
Non-Patent Citations (3)
| Title |
|---|
| 《计算机工程》 20100831 董晶等 基于资源限制的高性能计算系统功耗管理 276-277，280 1-2 第36卷, 第16期 2 * |
| 《计算机工程与科学》 20091231 戴永涌，杨树军 基于资源调度的集群节能系统的设计与实现 176-178，213 1-2 第31卷, 第A1期 2 * |
| 《计算机研究与发展》 20091231 卢宇彤，杨学军 面向分布对象存储结构的高性能计算系统资源管理方法 64-70 1-2 , 2 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2016058386A1 (en) * | 2014-10-17 | 2016-04-21 | 深圳市中兴微电子技术有限公司 | Power consumption management method and device and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101943943B (en) | 2011-11-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102929720B (en) | A kind of energy-conservation job scheduling system | |
| CA2741088C (en) | Methods of achieving cognizant power management | |
| CN102111337B (en) | Method and system for task scheduling | |
| US9274585B2 (en) | Combined dynamic and static power and performance optimization on data centers | |
| Gu et al. | Energy efficient scheduling of servers with multi-sleep modes for cloud data center | |
| CN105868004B (en) | Scheduling method and scheduling device of service system based on cloud computing | |
| CN104572307A (en) | Method for flexibly scheduling virtual resources | |
| CN103488539A (en) | Data center energy saving method based on central processing unit (CPU) dynamic frequency modulation technology | |
| CN113672383A (en) | Cloud computing resource scheduling method, system, terminal and storage medium | |
| Jin et al. | Energy-efficient task scheduling for CPU-intensive streaming jobs on Hadoop | |
| CN101819459B (en) | Heterogeneous object memory system-based power consumption control method | |
| CN103428102A (en) | Dynamic load balance achieving method and system in distributed network | |
| CN102339233A (en) | Cloud computing centralized management platform | |
| CN103023802B (en) | A kind of low energy consumption dispatching patcher towards web cluster and method | |
| CN103414784B (en) | Support the cloud computing resource scheduling method of contingency mode | |
| US9195514B2 (en) | System and method for managing P-states and C-states of a system | |
| Jin et al. | A virtual machine scheduling strategy with a speed switch and a multi-sleep mode in cloud data centers | |
| CN110825212B (en) | Energy-saving scheduling method and device and computer storage medium | |
| CN101943943B (en) | Method for optimizing energy consumption of computation array based on similar resource aggregation | |
| Sakamoto et al. | Analyzing resource trade-offs in hardware overprovisioned supercomputers | |
| CN101943944B (en) | Idle history information-based energy-saving method for computing array | |
| Zhang et al. | An energy-aware task scheduling algorithm for a heterogeneous data center | |
| Li et al. | SLA‐Aware and Energy‐Efficient VM Consolidation in Cloud Data Centers Using Host State 3rd‐Order Markov Chain Model | |
| CN101847044B (en) | Low-power consumption management method for petascale computer cluster | |
| Liu et al. | Towards optimal cpu frequency and different workload for multi-objective vm allocation |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20111130 Termination date: 20150915 |
|
| EXPY | Termination of patent right or utility model |