CN104598298A - Virtual machine dispatching algorithm based on task load and current work property of virtual machine - Google Patents
Virtual machine dispatching algorithm based on task load and current work property of virtual machine Download PDFInfo
- Publication number
- CN104598298A CN104598298A CN201510058421.4A CN201510058421A CN104598298A CN 104598298 A CN104598298 A CN 104598298A CN 201510058421 A CN201510058421 A CN 201510058421A CN 104598298 A CN104598298 A CN 104598298A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- cpu
- queue
- scheduling
- module
- 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.)
- Pending
Links
Landscapes
- Debugging And Monitoring (AREA)
Abstract
The invention discloses a virtual machine dispatching algorithm based on a task load and the current work property of a virtual machine, wherein an interruption monitoring module is in charge of monitoring the current interruption frequency of the virtual machine, and is used for judging whether the work property of the virtual machine is in a CPU (central processing unit) dense type or I/O (input/output) dense type. A CPU monitoring module is in charge of monitoring the current CPU use rate of the virtual machine and is used for monitoring the task load of the virtual machine, a dispatching time segment of the virtual machine is calculated according to the current task load and the historical information in a certain time range, and in addition, a dispatcher is informed to update the time segment. A dispatcher module modifies a credit value dispatcher, so that the dispatcher interrupts the information provided by the monitoring module and the CPU monitoring module, different dispatching time segments are applied to the virtual machines with different work properties, the I/O request delay is reduced, meanwhile, the sufficient time is provided for handling the I/O request, and in addition, lower additional overhead is introduced. The modification of the three modules respectively occurs in the virtual machine monitor, so that good applicability is realized.
Description
Technical field
The present invention relates to system virtualization, virtual machine interrupts detection, virtual machine CPU usage detects and scheduling virtual machine device design field.Particularly, relate to the scheduling virtual machine algorithm based on virtual machine work at present character and task load, for solve caused by scheduling virtual machine device design defect I/O operating lag, cannot dynamically adapting virtual machine load change and the lower problem of I/O performance.
Background technology
Intel Virtualization Technology is present in more than 50 year with various different form.Abstract says, virtual is the logical expressions of various available resources, can not by the restriction of physical condition.Concrete is exactly a bit that Intel Virtualization Technology with the addition of one deck virtualization layer in systems in which, the physical resource of lower floor can be abstracted into virtual resource, for the application on upper strata.It both can mark off multiple virtual resource from single physical resource, also multiple physical resource can be integrated into a virtual resource.
Monitor of virtual machine (Virtual Machine Monitor, VMM) is a most important part in Intel Virtualization Technology.It is between hardware layer and virtual machine, operates in the highest privilege level.Its major function is abstract bottom hardware resource, makes it to become virtual resource, and the virtual machine being supplied to upper strata establishment uses.The major function of monitor of virtual machine comprises the management of virtual environment and the management of physical resource.
The monitor of virtual machine of current industry main flow comprises Xen, KVM, VirtualBox, Hyper-V, VMware etc.Wherein Xen is as the monitor of virtual machine of increasing income of main flow, occupies very consequence in practical application and academic research.The dispatching algorithm of Xen acquiescence is credit value dispatching algorithm (Credit Scheduler).Credit value dispatching algorithm is considered as credit value the time of distributing, each credit value assignment period can be all VCPU distribute credit value, VCPU often runs a period of time will deduct the credit value corresponding with its working time, does not just rerun after credit value is less than 0 within this dispatching cycle.
Due to the introducing of monitor of virtual machine, the problem originally solved in legacy operating system there will be new change, also there will be the problem that there will not be in some legacy operating systems simultaneously.One of them very important problem is exactly the problem of I/O operating lag.In legacy operating system, when an I/O request comes interim, operating system can perceive this request, and dispatches his recipient, makes it seize the process of current operation, process I/O request, thus the delay that I/O is asked diminishes.But in virtualized environment, all I/O requests can be received by monitor of virtual machine, are then transmitted to corresponding virtual machine, operating system cannot perceive the recipient which virtual machine concrete is I/O request, therefore cannot seize.When several virtual machine shares a physical cpu time, virtual machine waits for that with regard to having to other virtual machine could process oneself I/O request after using the time distributing to them.In time having a lot of virtual machine to share same physical cpu, delay can become very high, so high delay for an I/O intensity virtual machine be unacceptable.
The fairness of scheduling placed above the other things by traditional scheduler, I/O task and CPU task dispatching treated together, can cause I/O task inefficiency.In order to solve the problem of I/O inefficiency, academia proposes multiple optimization method, I/O task and CPU task are treated with a certain discrimination, but more extreme is divided into two classes virtual machine: pure I/O operation and pure CPU calculate, do not consider that distributing to I/O asks enough CPU time to go process, do not consider the virtual machine of mixed type, therefore can affect the efficiency of I/O task yet.On the other hand, in part research in the past, the type of virtual machine manually can only be specified when system starts, the virtual machine of an I/O intensity also may change when system cloud gray model and become CPU intensity, when the virtual machine that is not noted as I/O intensity receives a large amount of I/O request, scheduler cannot change scheduling time sheet, and the delay that I/O is responded cannot reduce.Therefore these methods have limitation in the middle of the system of reality.
In sum, the problem that I/O operating lag is larger is there is in traditional scheduler, although the method researched and proposed existing can reduce delay to a certain extent, but virtual machine load can not be adapted to change dynamically and be not thought of as I/O request and provide the sufficient time to process, equally also can affect the performance of I/O request.Therefore, those skilled in the art is devoted to develop a kind of scheduling virtual machine algorithm based on virtual machine work at present character and task load, addresses this problem.
Summary of the invention
Because the above-mentioned defect of prior art, technical matters to be solved by this invention is to provide the scheduling virtual machine algorithm based on virtual machine work at present character and task load, while guarantee I/O request can be responded in time, enough CPU time are had again to ask for the treatment of I/O, the size of balance timeslice, make I/O maximizing performance, while lifting I/O performance, again minimum overhead is brought to the virtual machine of CPU intensity.
For achieving the above object, the invention provides the scheduling virtual machine algorithm based on virtual machine work at present character and task load, this algorithm contains interruption monitoring module, cpu monitor module and these three modules of Scheduler module:
Interrupting monitoring module operates in monitor of virtual machine, obtains the interrupt request of virtual machine reception from event channel.According to the information that interrupt request is carried, identify and comprise domainU and send request, domainU receives request, domain0 sends a reply, domain0 receives reply etc. and operates in interior I/O operation, monitor the interruption frequency of all virtual machines, for judging that the job specification of each virtual machine is that CPU is intensive or I/O is intensive.
Cpu monitor module operates in monitor of virtual machine, the busy ratio of monitoring virtual machine CPU, and calculates CPU usage, calculates the scheduling time sheet needed for virtual machine according to the CPU usage of virtual machine.When sheet computing time, cpu monitor module can be mapped in 1ms to the 20ms interval of timeslice from 0% to 100% cpu busy percentage, in order to change stable of retention time sheet, reduce jolting of performance, only just can modification time sheet size when continuous several cycle CPU usage is all in certain scope.This module make use of aging algorithm when calculating CPU usage, the historical information within the scope of certain hour is attached in current information and calculates.
Scheduler module is modified on the basis of CREDIT scheduler, and the data structure of amendment scheduler enables to store interrupts monitoring module and cpu monitor module institute feedack.Two queues can be safeguarded in scheduler algorithms, be referred to as A queue and B queue.According to interruption monitoring module feedack, the virtual machine of different operating character is put in different queues.The virtual machine of CPU intensity leaves in A queue, the virtual machine of I/O intensity leaves in B queue, when starting each dispatching cycle, the virtual machine that the head of the queue element that scheduler can choose one of them queue will run as the next one, concrete choosing method is as follows:
Step 1, first choose from B queue the virtual machine that head of the queue will run as the next one, then the virtual machine of present finishing scheduling is inserted the tail of the queue of the queue corresponding with its job specification type.
Step 2 was until virtual machines all in B queue all ran once or B queue is sky, the virtual machine that the head of the queue selecting A queue will run as the next one.
Step 3, repetition step 1 and 2, until all virtual machines have used up the timeslice that they distributed in this cycle.
Step 4, redistribute credit value and timeslice, get back to step 1.
After obtaining the virtual machine that the next one will run, scheduling function can take out the information of interrupting monitoring module and cpu monitor module and upgrading from domain structure, and the information according to interrupting monitoring module applies different scheduling strategies to dissimilar virtual machine.According to the information of cpu monitor module, obtain different timeslice sizes, and timer is set by this timeslice, carry out the scheduling of virtual machine.
Scheduling virtual machine algorithm based on virtual machine work at present character and task load has following useful technique effect:
(1) interrupt all events in monitoring module meeting monitor event passage, intercept and capture the communication event between domain0 and domainU, add up the interruption frequency of each virtual machine, be used as the foundation distinguishing virtual machine type.This module provides the information of Virtual Machine Worker character, virtual machine is divided into I/O intensity and intensive two types of CPU, the dispatching algorithm that scheduler can be different according to different type application.The virtual machine of CPU intensity adopts the Credit dispatching algorithm of acquiescence, and the virtual machine of I/O intensity adopts dispatching algorithm provided by the invention.
(2) cpu monitor module can add up the busy number of times of virtual machine in the certain hour cycle, by the CPU usage of busy ratio between number of times and total activation number of times as virtual machine.The timeslice needed for the dispatching algorithm that the intensive virtual machine of I/O is adopted is calculated according to the CPU usage of monitoring out.In the calculating of timeslice, have employed linear mapping, avoid jolting of timeslice, additionally use aging algorithm, allow the CPU usage within the scope of certain hour calculate the timeslice in current period and information is provided, make calculating more accurate.
(3) Scheduler module is modified on original Credit scheduler, within a dispatching cycle, allow the intensive virtual machine of CPU run larger timeslice, the intensive virtual machine of I/O is allowed to run multiple less timeslice, the intensive virtual machine of priority scheduling I/O, reduces the delay of I/O response as far as possible simultaneously.Timeslice needed for the virtual machine of I/O intensity is calculated by cpu monitor module and gets, and guarantee has sufficient time to process I/O request as far as possible, promotes I/O performance further.On the other hand, the timeslice that all virtual machines obtain within a dispatching cycle is equal, ensure that the fairness of scheduling.
(4) be all realize in monitor of virtual machine with upper module, do not need to revise the code of client computer, have nothing to do with concrete Client OS, therefore go for all Client OSs, there is extremely strong adaptive.
(5) state in realization in the process of module, shorter timeslice is adopted for the intensive virtual machine of I/O, and the intensive virtual machine of CPU is adopted to the timeslice of Credit scheduler acquiescence, the impact caused for the performance of the virtual machine of CPU intensity is less, and the overhead that system is introduced is less.
Be described further below with reference to the technique effect of accompanying drawing to design of the present invention, concrete structure and generation, to understand object of the present invention, characteristic sum effect fully.
Accompanying drawing explanation
Fig. 1 is the scheduling model framework of the method for the invention;
Fig. 2 is the flow process that Xen process is interrupted;
Fig. 3 is the embodiment of dispatching method of the present invention.
Embodiment
Below in conjunction with accompanying drawing, embodiments of the invention are elaborated: the present embodiment is implemented under premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
Fig. 1 is scheduling model framework of the present invention.The present invention adds two new modules in original scheduling model, is respectively to interrupt monitoring module and cpu monitor module.Interrupt monitoring module and can intercept and capture the information transmitted in event channel, add up the interruption frequency of each virtual machine according to the id of interrupt number and virtual machine.Cpu monitor module can monitor the schedule information of all virtual machines, counts the CPU usage of virtual machine in some cycles.These two modules can monitor the current running status of virtual machine, provide operation information to scheduler, enable scheduler change scheduling strategy dynamically.
Fig. 2 is the flow process that Xen process is interrupted.First need before breaking in processes interrupt number and event channel binding.When a virtual machine is created, first it can call get_free_port function and send to monitor of virtual machine the event channel port that a hypercalls applies for the free time, then calls the page table of grant_table_create function some free time of application for I/O ring.The page table of these free time can be shared by the mode of authorization list and host, and each authorization list can have a mandate to quote.After virtual machine activation, the port numbers of the free time of application and the mandate of I/O ring are quoted and can be passed to domain0 by XenStore.XenStore is mainly used to make can communicate before setting up event channel and authorization list between each domain.After domain0 receives port numbers that domainU sends and mandate is quoted, its can port and interrupt request (Interrupt Request, IRQ) binding, and such Xen just knows which domainU this interruption is dealt in.Then domain0 can be mapped to the page table shared in the address space of oneself, completes the attended operation that front and back end drives.When one interrupts arriving, first Xen can enter the public entrance of interrupt processing of a common_interrupt by name.After common_interrupt preserves interrupt spot, call do_IRQ and interruption is specifically processed.For the interruption of client registration, do_IRQ can call _ _ do_guest_IRQ process._ _ do_guest_IRQ finally can call send_guest_pirq to all this domainU interrupted that have registered, and arranges the pending position of the event channel of corresponding VCPU wherein
Fig. 3 is an example being used for describing dispatching algorithm.In fig. 3 a, assuming that the VM1 intensive virtual machine that is I/O, in Credit scheduler, one-period it can only be scheduled once, wait for after running 30ms and distribute timeslice next time, I/O request at this time postpones the highlyest can reach 90ms.In scheduler of the present invention, we are divided into several less timeslices this 30ms, and partitioning standards is from cpu monitor module (for 10ms in figure).In fig 3b, VM1 and VM2 is the virtual machine of I/O intensity, and first algorithm can allow VM1 run 10ms when scheduling, then insert back scheduling queue, then VM2 runs 10ms, and inserts back queue, continue scheduling VM3, circulate until run complete dispatching cycle so successively.Can be found out by schematic diagram, what within a dispatching cycle, the intensive virtual machine of I/O ran does not become T.T., but the frequency of scheduling adds, and therefore under the prerequisite guaranteed fairness, the response for I/O request becomes faster.
The particular content that the present invention is based on the scheduling virtual machine algorithm of virtual machine work at present character and task load comprises: interrupt monitoring module, cpu monitor module and Scheduler module.
Interrupting monitoring module operates in monitor of virtual machine, obtains the interrupt request of virtual machine reception from event channel.According to the information that interrupt request is carried, identify and comprise domainU and send request, domainU receives request, domain0 sends a reply, domain0 receives reply etc. and operates in interior I/O operation, monitor the interruption frequency of all virtual machines, for judging that the job specification of each virtual machine is that CPU is intensive or I/O is intensive.In the middle of Xen, when there being I/O event to arrive, front-end driven and rear end driving can notify the other side by event channel.The transmission of I/O data is then realized by I/O ring.Being used for the structure of a description virtual machine in Xen is struct domain, and primary structure is as follows:
The wherein structure of struct evtchn presentation of events passage, a domain can have at most 1024 event channels.Each event channel has a pending position, is used for indicating that interruption needs process.When a virtual machine is created, it can call get_free_port function and send to monitor of virtual machine the event channel port that hypercalls applies for the free time, then calls the page table of grant_table_create function some free time of application for I/O ring.The page table of these free time can be shared by the mode of authorization list and host, and each authorization list can have a mandate to quote.
As shown in Figure 2, the virtual machine that finally can call send_guest_pirq function notice corresponding carrys out handling interrupt requests to concrete interruption process of transmitting.First interruption monitoring module of the present invention have recorded the port numbers of the event channel that it uses when virtual machine activation.When virtual machine runs, monitor whole event passage, obtain concrete interrupting information, comprise domainU and send request, domain0 receives request, and domain0 sends a reply, and domainU receives the information of replying and waiting operation.Then in evtchn _ send function, made some amendments, carry out the statistics of interrupting when communicating between each domain, concrete algorithm is as shown in following false code:
Cpu monitor module operates in monitor of virtual machine, the busy ratio of monitoring virtual machine CPU, and calculates CPU usage, calculates the scheduling time sheet needed for virtual machine according to the CPU usage of virtual machine.When sheet computing time, cpu monitor module can be mapped in 1ms to the 20ms interval of timeslice from 0% to 100% cpu busy percentage, in order to change stable of retention time sheet, reduce jolting of performance, only just can modification time sheet size when continuous several cycle CPU usage is all in certain scope.This module make use of aging algorithm when calculating CPU usage, the historical information within the scope of certain hour is attached in current information and calculates.
When can utilize that in a period of time, VCPU is scheduled when calculating CPU usage, the number of times of non-idle and the ratio of the total degree be scheduled are as CPU usage.Scheduling times measurement period due to VCPU is 10ms, within some 10ms cycles, if the VCPU free time, can think that it does not run in this 10ms, same, if not the free time then can think that VCPU is running in this cycle always, so this ratio can think the utilization rate of CPU in theory, that is the utilization rate of CPU can be embodied by the ratio of the number of times of idle non-when VCPU is scheduled in a period of time with the total degree be scheduled.Concrete statistic algorithm is by shown in following false code:
The timeslice calculating dispatching office need according to CPU usage is needed after obtaining CPU usage.30ms under the timeslice default situations of Xen scheduling, this timeslice for CPU intensity task be an all well and good timeslice, the demand of cpu resource can be met.But for I/O intensity virtual machine and improper, the delay that it can cause I/O to respond.The timeslice of 1ms can well meet the requirement of I/O response speed, but due to the minimizing of timeslice, the consumption of contextual processing also can uprise, and CPU treatment effeciency can be caused to be deteriorated, and cannot process some I/O needing computational resource in time and ask.Present invention employs the method for dynamic adjustments timeslice, according to the real time load amendment scheduling time sheet of virtual machine.The present invention is mapped to cpu busy percentage in 1ms to the 20ms interval of timeslice from 0% to 100%.The size ceaselessly changing timeslice finds in actual experiment process, although CPU usage is relatively stable within a period of time, can ceaselessly change in adjacent several measurement periods, if also can cause jolting of performance.So only just can modification time sheet size when continuous several cycle CPU usage is all in certain scope.On the other hand, record is CPU usage in some sense cycle in cpu monitor module, in the middle of actual motion, the CPU usage for the previous period of this sense cycle also should include limit of consideration in, can reduce CPU usage like this and to suddenly change the error brought.Be used herein aging algorithm to obtain CPU usage.Concrete computing method are as shown in following false code:
More than describe preferred embodiment of the present invention in detail.Should be appreciated that the ordinary skill of this area just design according to the present invention can make many modifications and variations without the need to creative work.Therefore, all technician in the art, all should by the determined protection domain of claims under this invention's idea on the basis of existing technology by the available technical scheme of logical analysis, reasoning, or a limited experiment.
Claims (5)
1. based on a scheduling virtual machine algorithm for virtual machine work at present character and task load, it is characterized in that, comprise and interrupt monitoring module, cpu monitor module and Scheduler module,
Described interruption monitoring module operates in monitor of virtual machine, obtains the interrupt request of virtual machine reception from event channel; And according to described interrupt request, identify domainU and send request, domainU receives request, and domain0 sends a reply, and domain0 receives the I/O operation of replying; Monitor the interruption frequency of all virtual machines; Judge that the job specification of each virtual machine is that CPU is intensive or I/O is intensive;
Described cpu monitor module operates in monitor of virtual machine, the busy ratio of monitoring virtual machine CPU, and calculates CPU usage; The scheduling time sheet needed for virtual machine is calculated according to described CPU usage;
Described Scheduler module can store described interruption monitoring module and described cpu monitor module institute feedack; The virtual machine of different operating character, according to described interruption monitoring module feedack, is put in different queues by described Scheduler module; When starting each dispatching cycle, the virtual machine that the head of the queue element that described Scheduler module can choose one of them queue will run as the next one, then takes out the information of described interruption monitoring module and the renewal of described cpu monitor module from domain structure; Information according to described interruption monitoring module applies different scheduling strategies to dissimilar virtual machine; According to the information of described cpu monitor module, obtain different described scheduling time sheet, and timer is set with described scheduling time sheet.
2. scheduling virtual machine algorithm as claimed in claim 1, is characterized in that, when calculating described scheduling time sheet, described cpu monitor module is mapped to cpu busy percentage in 1ms to the 20ms interval of described scheduling time sheet from 0% to 100%.
3. scheduling virtual machine algorithm as claimed in claim 2, is characterized in that, when CPU usage is all in preset range described in continuous three cycles, revises the size of described scheduling time sheet.
4. scheduling virtual machine algorithm as claimed in claim 1, is characterized in that, described cpu monitor module utilizes aging algorithm when calculating described CPU usage.
5. scheduling virtual machine algorithm as claimed in claim 1, it is characterized in that, the virtual machine of CPU intensity leaves in A queue, and the virtual machine of I/O intensity leaves in B queue, and the virtual machine choosing method that the described next one will run is as follows,
Step 1, first choose from B queue the virtual machine that head of the queue will run as the next one, then the virtual machine of present finishing scheduling is inserted the tail of the queue of the queue corresponding with its job specification type;
Step 2 was until virtual machines all in B queue all ran once or B queue is sky, the virtual machine that the head of the queue selecting A queue will run as the next one;
Step 3, repeating said steps 1 and described step 2, until all virtual machines have used up the timeslice of distributing in this cycle;
Step 4, redistribute credit value and described timeslice, get back to described step 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510058421.4A CN104598298A (en) | 2015-02-04 | 2015-02-04 | Virtual machine dispatching algorithm based on task load and current work property of virtual machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510058421.4A CN104598298A (en) | 2015-02-04 | 2015-02-04 | Virtual machine dispatching algorithm based on task load and current work property of virtual machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104598298A true CN104598298A (en) | 2015-05-06 |
Family
ID=53124116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510058421.4A Pending CN104598298A (en) | 2015-02-04 | 2015-02-04 | Virtual machine dispatching algorithm based on task load and current work property of virtual machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104598298A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105045667A (en) * | 2015-07-13 | 2015-11-11 | 中国科学院计算技术研究所 | Resource pool management method for vCPU scheduling of virtual machines |
| CN106095578A (en) * | 2016-06-14 | 2016-11-09 | 上海交通大学 | Method is submitted in directly interruption based on hardware ancillary technique and virtual cpu running status |
| CN106293881A (en) * | 2016-08-11 | 2017-01-04 | 上海交通大学 | A kind of Network Performance Monitor and monitoring method thereof accessing framework based on nonuniformity I/O |
| CN106874102A (en) * | 2015-12-18 | 2017-06-20 | 北京奇虎科技有限公司 | Resource regulating method and device based on container work property |
| CN108027752A (en) * | 2015-09-16 | 2018-05-11 | 佳能株式会社 | Information processor, control method and program for information processor |
| CN109309726A (en) * | 2018-10-25 | 2019-02-05 | 平安科技（深圳）有限公司 | Document generating method and system based on mass data |
| CN111198755A (en) * | 2019-12-23 | 2020-05-26 | 曙光信息产业(北京)有限公司 | SLURM job scheduling system-based pre-charging device and method |
| WO2020220436A1 (en) * | 2019-04-29 | 2020-11-05 | 东北大学 | Virtual machine work queue and redundant queue update method for different aging scenes |
| CN114691374A (en) * | 2022-05-31 | 2022-07-01 | 苏州浪潮智能科技有限公司 | Request distribution method of virtual channel and related device |
| WO2023165485A1 (en) * | 2022-03-04 | 2023-09-07 | 阿里巴巴（中国）有限公司 | Scheduling method and computer system |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101101562A (en) * | 2007-07-10 | 2008-01-09 | 北京大学 | Dummy machine external storage on-line migration method |
| CN101436966A (en) * | 2008-12-23 | 2009-05-20 | 北京航空航天大学 | Network monitoring and analysis system under virtual machine circumstance |
| CN101465863A (en) * | 2009-01-14 | 2009-06-24 | 北京航空航天大学 | Method for implementing high-efficiency network I/O in kernel virtual machine circumstance |
| CN101706742A (en) * | 2009-11-20 | 2010-05-12 | 北京航空航天大学 | Method for dispatching I/O of asymmetry virtual machine based on multi-core dynamic partitioning |
| US20100131669A1 (en) * | 2008-11-26 | 2010-05-27 | Microsoft Corporation | Hardware acceleration for remote desktop protocol |
| US20110022818A1 (en) * | 2009-07-24 | 2011-01-27 | Kegel Andrew G | Iommu using two-level address translation for i/o and computation offload devices on a peripheral interconnect |
-
2015
- 2015-02-04 CN CN201510058421.4A patent/CN104598298A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101101562A (en) * | 2007-07-10 | 2008-01-09 | 北京大学 | Dummy machine external storage on-line migration method |
| US20100131669A1 (en) * | 2008-11-26 | 2010-05-27 | Microsoft Corporation | Hardware acceleration for remote desktop protocol |
| CN101436966A (en) * | 2008-12-23 | 2009-05-20 | 北京航空航天大学 | Network monitoring and analysis system under virtual machine circumstance |
| CN101465863A (en) * | 2009-01-14 | 2009-06-24 | 北京航空航天大学 | Method for implementing high-efficiency network I/O in kernel virtual machine circumstance |
| US20110022818A1 (en) * | 2009-07-24 | 2011-01-27 | Kegel Andrew G | Iommu using two-level address translation for i/o and computation offload devices on a peripheral interconnect |
| CN101706742A (en) * | 2009-11-20 | 2010-05-12 | 北京航空航天大学 | Method for dispatching I/O of asymmetry virtual machine based on multi-core dynamic partitioning |
Non-Patent Citations (1)
| Title |
|---|
| 常郅博: "基于高性能网络虚拟化的可调节信用值调度算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105045667A (en) * | 2015-07-13 | 2015-11-11 | 中国科学院计算技术研究所 | Resource pool management method for vCPU scheduling of virtual machines |
| CN105045667B (en) * | 2015-07-13 | 2018-11-30 | 中国科学院计算技术研究所 | A kind of resource pool management method for virtual machine vCPU scheduling |
| US11397603B2 (en) | 2015-09-16 | 2022-07-26 | Canon Kabushiki Kaisha | Information processing apparatus, control method therefor, and program |
| CN108027752A (en) * | 2015-09-16 | 2018-05-11 | 佳能株式会社 | Information processor, control method and program for information processor |
| CN108027752B (en) * | 2015-09-16 | 2021-10-29 | 佳能株式会社 | Information processing apparatus, control method for information processing apparatus, and program |
| CN106874102A (en) * | 2015-12-18 | 2017-06-20 | 北京奇虎科技有限公司 | Resource regulating method and device based on container work property |
| CN106095578A (en) * | 2016-06-14 | 2016-11-09 | 上海交通大学 | Method is submitted in directly interruption based on hardware ancillary technique and virtual cpu running status |
| CN106095578B (en) * | 2016-06-14 | 2019-04-09 | 上海交通大学 | Method is submitted in direct interruption based on hardware ancillary technique and virtual cpu operating status |
| CN106293881B (en) * | 2016-08-11 | 2020-02-07 | 上海交通大学 | Performance monitor based on non-uniform I/O access framework and monitoring method thereof |
| CN106293881A (en) * | 2016-08-11 | 2017-01-04 | 上海交通大学 | A kind of Network Performance Monitor and monitoring method thereof accessing framework based on nonuniformity I/O |
| CN109309726A (en) * | 2018-10-25 | 2019-02-05 | 平安科技（深圳）有限公司 | Document generating method and system based on mass data |
| WO2020220436A1 (en) * | 2019-04-29 | 2020-11-05 | 东北大学 | Virtual machine work queue and redundant queue update method for different aging scenes |
| CN111198755A (en) * | 2019-12-23 | 2020-05-26 | 曙光信息产业(北京)有限公司 | SLURM job scheduling system-based pre-charging device and method |
| WO2023165485A1 (en) * | 2022-03-04 | 2023-09-07 | 阿里巴巴（中国）有限公司 | Scheduling method and computer system |
| CN114691374A (en) * | 2022-05-31 | 2022-07-01 | 苏州浪潮智能科技有限公司 | Request distribution method of virtual channel and related device |
| CN114691374B (en) * | 2022-05-31 | 2022-10-28 | 苏州浪潮智能科技有限公司 | Request distribution method of virtual channel and related device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104598298A (en) | Virtual machine dispatching algorithm based on task load and current work property of virtual machine | |
| US11221880B2 (en) | Adaptive computing resource allocation approach for virtual network functions | |
| CN108701046B (en) | Method, system, device, apparatus, and medium for virtual CPU merging to avoid physical CPU contention between virtual machines | |
| EP2799990A2 (en) | Dynamic virtual machine sizing | |
| CN102253857B (en) | Xen virtual machine scheduling control method in multi-core environment | |
| CN109726005B (en) | Method, server system and computer readable medium for managing resources | |
| US9104491B2 (en) | Batch scheduler management of speculative and non-speculative tasks based on conditions of tasks and compute resources | |
| US10733022B2 (en) | Method of managing dedicated processing resources, server system and computer program product | |
| EP2772854B1 (en) | Regulation method and regulation device for i/o channels in virtualization platform | |
| CN103744716A (en) | Dynamic interrupt balanced mapping method based on current virtual central processing unit (VCPU) scheduling state | |
| CN105550040A (en) | KVM platform based virtual machine CPU resource reservation algorithm | |
| CN103049314A (en) | Real-time scheduling method for open system | |
| CN113821308A (en) | System on chip, virtual machine task processing method and device, and storage medium | |
| CN104917839A (en) | Load balancing method for use in cloud computing environment | |
| CN105242954A (en) | Mapping method between virtual CPUs (Central Processing Unit) and physical CPUs, and electronic equipment | |
| CN111597044A (en) | Task scheduling method and device, storage medium and electronic equipment | |
| Moulik et al. | COST: A cluster-oriented scheduling technique for heterogeneous multi-cores | |
| CN116820784B (en) | GPU real-time scheduling method and system for reasoning task QoS | |
| WO2020166423A1 (en) | Resource management device and resource management method | |
| KR101330609B1 (en) | Method For Scheduling of Mobile Multi-Core Virtualization System To Guarantee Real Time Process | |
| CN109189581B (en) | Job scheduling method and device | |
| CN109656716B (en) | Slurm job scheduling method and system | |
| KR20140125893A (en) | Virtualized many-core environment job distribution system and method, and recording medium thereof | |
| CN106020980A (en) | Virtual desktop oriented VCPU (Virtual Central Processing Unit) scheduling method | |
| Dehsangi et al. | cCluster: a core clustering mechanism for workload-aware virtual machine scheduling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150506 |
|
| RJ01 | Rejection of invention patent application after publication |