CN103744728A - Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method - Google Patents
Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method Download PDFInfo
- Publication number
- CN103744728A CN103744728A CN201410029589.8A CN201410029589A CN103744728A CN 103744728 A CN103744728 A CN 103744728A CN 201410029589 A CN201410029589 A CN 201410029589A CN 103744728 A CN103744728 A CN 103744728A
- Authority
- CN
- China
- Prior art keywords
- ple
- virtual machine
- vcpu
- parameter
- dynamic
- 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 dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method which includes monitoring running states of all virtual machines to acquire average wait time needed for waiting for the spin-lock of each virtual machine, and regulating PLE parameters dynamically according to the average wait time so as to be adaptable to the current system running state; meanwhile, when the PLE is triggered by VCPUs(virtual central processing units), on the basis of the dynamic PLE technology, upgrading all VCPUs, running in a kernel mode, of corresponding virtual machines to heads of corresponding physical CPU running queues so as to realize synchronous running of the group of VCPUs in the next scheduling period. The dynamic PLE technology and the virtual machine co-scheduling are combined so that the problems such as busy time of the VCPUs of the virtual machines due to the fact that spin-lock holders are seized are solved and integral performance of the virtual machines and the system is improved effectively.
Description
Technical field
The present invention relates to computer virtual machine performance monitoring and scheduling virtual machine field, relate in particular to a kind of virtual machine coordinated dispatching method based on dynamic PLE technology.
Background technology
The physical resource of Intel Virtualization Technology abstract machine, upwards presents virtual hardware interface, and multiple operating system can be moved on Same Physical equipment.Utilize Intel Virtualization Technology, the service assembly that originally runs on many physical equipments, to same physical server, has been reduced to the management maintenance cost of enterprise in IT infrastructure when improving physical resource utilization factor.The management to physical resource by Intel Virtualization Technology on the other hand, can carry out dynamic-configuration to resource, stronger for miscellaneous service adaptability.Simultaneously Intel Virtualization Technology can be monitored and isolate virtual server, and guaranteeing can be separate between the normal operation of virtual server and virtual server.In addition, virtual server can copy transfer more efficiently, improves availability and disaster recovery speed.Therefore, Intel Virtualization Technology becomes the indispensable part of cloud computing and data center.
Monitor of virtual machine (Virtual Machine Monitor, VMM) be the Intel Virtualization Technology of current main flow, between hardware and legacy operating system, be in charge of the physical resource of computing machine, comprise the hardware resources such as internal memory, CPU and I/O equipment, and fictionalize hardware interface for upper strata, multiple operating system can be moved on same physical equipment.
In order to promote better the performance of virtualized server, on hardware, also added virtualized support simultaneously.The VT series Intel Virtualization Technology of Intel, provides the support to Intel Virtualization Technology, and the performance of virtual machine is significantly improved.
The introducing of monitor of virtual machine, makes the problem much having solved in legacy operating system occur new variation, also makes more existing mechanism in legacy operating system occur new problem simultaneously.Spin lock in traditional operating system (Spin-Lock) is for solving the competition for shared resource between CPU.When multiple CPU need to obtain same spin lock, only have a CPU can hold this lock, other CPU always cycle detection spin lock whether be released.The CPU that holds spin lock in traditional operating system can discharge lock within the very short time, waits for that like this time of the CPU cycle detection of spin lock can't be very long.But in virtualized environment, the VCPU that holds spin lock in virtual machine may be dispatched out physical cpu by monitor of virtual machine, make to wait in this virtual machine other VCPU state in cycle detection spin lock always of this spin lock, this problem is called as LHP(Lock Holder Preemption, spin lock holder be preempted) problem, can cause the decline of virtual machine performance.
In order to solve above-mentioned LHP problem, existing technical scheme comprises the detection such as busy of the cooperative scheduling of VCPU in virtual machine and VCPU.In virtual machine the cooperative scheduling of VCPU refer to by one group of VCPU(in virtual machine can be in virtual machine all VCPU can be also part VCPU) dispatch into physical cpu simultaneously, can effectively reduce like this appearance of LHP phenomenon, but this method can cause the problems such as the fragmentation of cpu resource and the uncertainty increase of system delay.VCPU busy waits detection, refers to when detecting that VCPU is busy and wait VCPU is dispatched out to physical cpu, is mainly divided into busy grade of VCPU based on virtual machine and detects and the detection such as the VCPU based on hardware supported is busy.The detections such as the VCPU based on virtual machine hurries need to be modified to VME operating system, notify VMM while there is the busy grade of VCPU in virtual machine, and this method is because need to revise VME operating system, and the scope of application is restricted.Based on the VCPU of hardware supported is busy, wait detections that the detected parameters such as busy in virtual machine control structure need to be set, these parameters are needs variation at computing machine in the optimal value under different loads.The detection the such as busy VCPU of the cooperative scheduling of VCPU in virtual machine and dynamic state of parameters adjustment is combined and can reduce the negative effect that cooperative scheduling brings, improve the performance of virtual machine simultaneously.
The current VCPU based on hardware supported is busy, and to wait detection be mainly PLE(Pause Loop Exit) technology, the detected parameters such as wherein hurry is PLE_GAP and PLE_WINDOW.When VCPU hurries etc., can carry out Pause instruction, if two continuous Pause instructions are in the time range of PLE_GAP appointment, can think that this VCPU is busy etc. due to what wait for that same spin lock causes, when this continuous Pause instruction exceedes the time range of PLE_WINDOW appointment, think that the busy equal time of this VCPU is long, now PLE provides trigger message to the scheduler module of VMM layer, scheduler module is dispatched out physical cpu by this VCPU, from the operation queue of physical cpu, selecting afterwards next VCPU to enter in this physical cpu moves, trigger implementation as shown in Figure 1.
In the utilization of existing PLE technology, have changeless parameter static PLE technology, also with good grounds computing machine when different loads dynamic change parameter to get the dynamic PLE technology of its optimal value.
In existing technical scheme, the cooperative scheduling of VPN and dynamic PLE technology are only used separately to one of them, the former is only based on static PLE technology, and it utilizes the trigger message of PLE to carry out cooperative scheduling to the VCPU in virtual machine, does not consider the adjustment of PLE parameter in different loads situation.The latter dynamically adjusts in different loads situation PLE parameter, does not utilize the cooperative scheduling of VCPU to reduce the generation of LHP.
At present still not by the cooperative scheduling of VCPU and technical scheme that dynamically PLE combines, therefore, those skilled in the art be devoted to develop a kind of by the cooperative scheduling of VCPU and dynamically PLE combine, utilize both advantages separately, with this, further promote the method for virtual machine performance.
Summary of the invention
Because do not utilize cooperative scheduling and dynamic PLE advantage separately in prior art, they are combined together, further to promote the method for virtual machine performance.Technical matters to be solved by this invention is to provide a kind of virtual machine coordinated dispatching method based on dynamic PLE technology, and it combines virtual machine cooperative scheduling and the dynamic advantage of PLE, effectively promotes the performance of virtual machine.
For achieving the above object, method provided by the present invention, its design concept is the running status by monitoring all virtual machines, obtains each virtual machine waiting for average waiting duration required during spin lock, with this, PLE parameter is dynamically adjusted, made it more be adapted to current running status.Simultaneously, based on dynamic PLE technology, when VCPU triggers PLE, all VCPU that run on kernel mode in virtual machine under this VCPU are promoted to the head of the queue of physical cpu operation queue separately, make next dispatching cycle this group VCPU can move simultaneously.Why according to VCPU, whether in kernel mode, adjusting queue is because in operating system, process operational mode has user model and kernel mode, and the spin lock of the overwhelming majority obtains at kernel mode, and can not be switched to user model before release lock.
Virtual machine coordinated dispatching method based on dynamic PLE technology comprises the steps:
Step 1, monitoring module are periodically monitored the running status of all virtual machines, obtain the spin lock average waiting duration of each virtual machine and import PLE parameter management module into;
Further, the spin lock average waiting duration of each virtual machine equals cpu cycle of taking during spin lock function operation in each virtual machine kernel that monitoring module extracts call number divided by spin lock function.
Step 2, PLE parameter management module are responsible for the PLE parameter of upgrading and preserving all virtual machines, and are responsible for providing to PLE administration module the PLE parameter of all virtual machines;
Further, PLE parameter management module newly-built Parameter File corresponding with this virtual machine when virtual machine is created; When virtual machine is destroyed, delete the Parameter File of this virtual machine correspondence in PLE parameter management module;
Further, PLE parameter comprises PLE_GAP and PLE_WINDOW, and PLE_WINDOW does dynamic adjustment according to the monitored results of monitoring module, and PLE_GAP adopts the default value of hardware PLE;
Further, the value of PLE parameter management module using the spin lock average waiting duration of each virtual machine as PLE_WINDOW;
Further, PLE parameter management module is by PLE parameter read-in Parameter File.
Step 3, PLE administration module are responsible for arranging the PLE parameter of all virtual machines in hardware PLE, and send schedule information to cooperative scheduling module;
Further, when the VCPU of virtual machine is scheduled into physical cpu, from PLE parameter management module, read the PLE parameter corresponding with virtual machine, and write in the VCPU control structure of hardware PLE; When VCPU triggers PLE, to cooperative scheduling module, send schedule information, the information that schedule information comprises the VCPU that triggers PLE.
Step 4, cooperative scheduling module, according to schedule information, are carried out cooperative scheduling for the virtual machine at the VCPU place of triggering PLE;
Further, cooperative scheduling module receives after schedule information, carries out following operation: judge whether VCPU runs on kernel mode, if so, VCPU is inserted into the head of the queue of physical cpu operation queue triggering all VCPU in the virtual machine at VCPU place of PLE; If not, do not need to change the position of VCPU in physical cpu operation queue.
Virtual machine coordinated dispatching method based on dynamic PLE technology of the present invention has following useful technique effect:
(1) can monitor the running status of all virtual machines, according to current each virtual machine running status separately, the PLE parameter in virtual machine control structure is set, guarantee the PLE parameter of each virtual machine dynamically to adjust, strengthen the performance boost that PLE brings.
(2) the busy VCPU of grade arriving based on PLE technology for detection, carries out cooperative scheduling by the VCPU group that runs on kernel mode in virtual machine, reduces the generation of LHP, effectively reduces VCPU and is wasted in the time on busy grade, promotes thus the performance of virtual machine.
(3) than solution in the past, the present invention combines dynamic PLE technology and virtual machine coordinated dispatching method, by virtual machine is monitored, dynamically adjusts PLE parameter, makes PLE more be adapted to each virtual machine.By virtual machine cooperative scheduling, effectively reduce the generation of LHP simultaneously, promote on the whole the performance of virtual machine.
(4) the present invention does not modify to the scheduling time sheet of virtual machine, do not increase the context switching frequency of VCPU or introduce other VCPU competition mechanism yet, in spin lock problem, less for the impact of system script function under effective solution virtual environment.
Below with reference to accompanying drawing, the technique effect of design of the present invention, concrete structure and generation is described further, to understand fully object of the present invention, feature and effect.
Accompanying drawing explanation
Fig. 1 is the scheduling virtual machine schematic diagram that adopts conventional P LE technology;
Fig. 2 is the schematic diagram that adopts the virtual machine coordinated dispatching method based on dynamic PLE technology of the present invention;
Fig. 3 is the PLE parameter management module diagram while the present invention is based on KVM;
Fig. 4 is virtual machine cooperative scheduling module diagram in the present invention.
Embodiment
For the software virtual machine KVM based on linux kernel, introduce specific embodiments of the invention below.
As shown in Figure 2, than traditional virtual machine coordinated dispatching method, the situation for each virtual machine in current system is dynamically adjusted PLE parameter, has added monitoring module, PLE parameter management module, PLE administration module and cooperative scheduling module.Monitoring module detects the running status of existing all virtual machines in system, obtains kernel function call number of times and the kernel function of virtual machine and moves shared cpu cycle.PLE parameter management module, for all virtual machine PLE parameters of system are managed, is responsible for creating the proc file that this virtual machine is corresponding during virtual machine creating, this file is for recording the PLE parameter of optimizing according to the monitored results of monitoring module; And when destroying, deletes virtual machine corresponding proc file; Need provides all virtual machines current PLE parameter to PLE administration module simultaneously.When PLE administration module is responsible for VCPU between virtual machine and is switched, according to the PLE parameter that obtains the corresponding virtual machine of VCPU that is scheduled in physical cpu from PLE parameter management module, and by the control structure of this parameter read-in hardware PLE.The information such as the VCPU of cooperative scheduling module based on PLE feedback is busy, when VCPU triggers PLE, carry out cooperative scheduling to one group of VCPU that may hold spin lock in virtual machine under VCPU.
Virtual machine coordinated dispatching method based on dynamic PLE technology in the present embodiment comprises the following steps:
Step 1, monitoring module are responsible for periodically monitoring the running status of all virtual machines, the cpu cycle taking while extracting the call number of spin lock function in each virtual machine kernel and spin lock function operation, the cpu cycle taking during spin lock function operation is exactly to each virtual machine spin lock average waiting duration separately divided by the value of the call number of spin lock function, the result of calculation that this monitoring is obtained is imported PLE parameter management module into, for PLE parameter is set.
Step 2, PLE parameter management module, be responsible for the PLE parameter of all virtual machines to manage, and be responsible for providing to PLE administration module the PLE parameter of all virtual machines.While creating virtual machine, in PLE parameter management module, create the Parameter File corresponding with virtual machine, by this Parameter File of PLE parameter read-in after optimizing; When destroying, deletes virtual machine the Parameter File of this virtual machine correspondence in PLE parameter management module.The spin lock average waiting duration of each virtual machine that PLE parameter management module provides monitoring module, as its current optimized PLE_WONDOW parameter, and PLE_GAP adopts the default value of hardware PLE.
Step 3, PLE administration module, be responsible for dynamically arranging and sending schedule information to cooperative scheduling module of PLE parameter.When the VCPU of virtual machine is scheduled into physical cpu, the interface that utilizes PLE parameter management module to provide, obtains the PLE parameter that this virtual machine is corresponding, writes in the VCPU control structure of hardware PLE, thereby the PLE parameter of this virtual machine is set.When VCPU triggers PLE, be responsible for sending schedule information to cooperative scheduling module.
Step 4, cooperative scheduling module, be responsible for virtual machine to carry out cooperative scheduling.When cooperative scheduling module is received after described schedule information, obtain the virtual machine information at the VCPU place of triggering PLE, for VCPU all in this virtual machine, carry out following operation: judge whether VCPU runs on kernel mode, if so, this VCPU is inserted into the head of the queue of affiliated physical cpu operation queue; If not, do not need to change the position of this VCPU in physical cpu operation queue.
Related realization for KVM is mainly divided into two parts, and the dynamic PLE technical step of Part I is as follows:
Step 1, utilize perf under Linux and trace-cmd instrument to carry out periodicity monitoring to all virtual machines.The monitored results of each virtual machine is write to the proc file of each virtual machine correspondence under PLE parameter management module root directory when each monitoring period finishes.The flow process of each monitoring period is as follows:
Step 2, import monitored results into PLE parameter management module, the PLE parameter to virtual machine in PLE parameter management module manages.In PLE parameter management module, store the PLE parameter of each virtual machine and PLE parameter read-in interface is provided to upper strata, the query interface of each virtual machine PLE parameter is provided downwards.As shown in Figure 3, system starts, and at KVM, starts during initialization, creates the root directory of PLE parameter management module under/proc catalogue, and PLE parameter is carried out to initialization.When creating virtual machine, according to the pid of virtual machine, create proc file, the PLE parameter of initialization virtual machine, and when virtual machine is destroyed, corresponding proc file is deleted.The mode of shining upon by Hash when this module realizes manages the PLE parameter of virtual machine, be convenient to virtual machine PLE parameter newly-built, delete and inquiry.
The interface that PLE parameter management module provides is as follows, needs to call corresponding interface in kvm_init, kvm_create_vm, kvm_destroy_vm and kvm_exit in KVM:
void?proc_mykpm_create(int?nr);
void?proc_mykpm_destroy(int?nr);
int?proc_mykpm_get(int?nr);
void?proc_mykpm_init(void);
void?proc_mykpm_deinit(void);
The specific implementation of the corresponding virtual machine PLE of virtual machine monitoring module read-write PLE parameter management module parameter:
Step 3, when VCPU is scheduled in physical cpu, inquire about the PLE parameter of virtual machine under this VCPU, write in the VMCS of virtual machine, thereby the PLE parameter of virtual machine be set according to the virtual machine running status monitoring.In vcpu_put, need to call int proc_mykpm_get (int nr) interface of PLE parameter management module, and write PLE parameter by vmcs_write.The PLE parameter of inquiry virtual machine needs the pid of virtual machine, therefore needs to revise struct kvm, adds nr variable for record the pid of virtual machine when virtual machine is newly-built, and after vcpu_put in for inquiring about the PLE parameter of virtual machine.
The VCPU cooperative scheduling of Part II virtual machine, as shown in Figure 4, concrete steps are as follows:
When step 1, VCPU trigger PLE, PLE administration module notice cooperative scheduling module has VCPU to trigger PLE, and imports the VCPU information that triggers PLE into.The interface of PLE administration module notice cooperative scheduling module is handle_pause (struct kvm_vcpu*me), and the kvm_vcpu_on_spin for step (structkvm_vcpu*me) that goes forward side by side completes processing logic in this function.
Struct cfs_rq in step 2, modification linux code, adds struct sched_entity*ple, for recording the VCPU information of virtual machine.Revise the interface of pick_next_entity function newly-built modification struct sched_entity*ple simultaneously.
EXPORT_SYMBOL_GPL(set_task_cfs_rq_ple);
Step 3, by the virtual machine information under the VCPU acquisition of information VCPU importing into, for VCPU all in this virtual machine, carry out following operation: if VCPU be can move and in kernel mode, the head of the queue of physical cpu operation queue under this VCPU being inserted into; If not, use acquiescence scheduler to dispatch this VCPU.Through above operation, this group VCPU that may hold spin lock will be dispatched to physical cpu in next dispatching cycle simultaneously.
void?kvm_vcpu_on_spin(struct?kvm_vcpu*me)
More than describe preferred embodiment of the present invention in detail, the design concept of applying this discovery can realize object of the present invention too on other software virtual machines (as Xen).The ordinary skill that should be appreciated that this area just can design according to the present invention be made many modifications and variations without creative work.Therefore, all technician in the art, all should be in 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 (10)
1. the virtual machine coordinated dispatching method based on dynamic PLE technology, is characterized in that, comprises the following steps:
Step 1, monitoring module are periodically monitored the running status of all virtual machines, obtain the spin lock average waiting duration of each described virtual machine and import PLE parameter management module into;
Step 2, described PLE parameter management module are responsible for the PLE parameter of upgrading and preserving all described virtual machines, and are responsible for providing to PLE administration module the described PLE parameter of all described virtual machines;
Step 3, described PLE administration module are responsible for arranging the described PLE parameter of all described virtual machines in hardware PLE, and send schedule information to cooperative scheduling module;
Step 4, described cooperative scheduling module, according to described schedule information, are carried out cooperative scheduling for the described virtual machine at the VCPU place of triggering PLE.
2. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 1, it is characterized in that, in described step 1, the cpu cycle that the described spin lock average waiting duration of each described virtual machine takies while equaling spin lock function operation in each described virtual machine kernel that described monitoring module extracts is divided by the call number of spin lock function.
3. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 2, is characterized in that, in described step 2, described PLE parameter management module is newly-built and Parameter File that described virtual machine is corresponding when described virtual machine is created; When described virtual machine is destroyed, delete the described Parameter File of described virtual machine correspondence in described PLE parameter management module.
4. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 3, it is characterized in that, in described step 2, described PLE parameter comprises PLE_GAP and PLE_WINDOW, described PLE_WINDOW does dynamic adjustment according to the monitored results of described monitoring module, and described PLE_GAP adopts the default value of described hardware PLE.
5. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 4, it is characterized in that, in described step 2, the value of described PLE parameter management module using the described spin lock average waiting duration of each described virtual machine as described PLE_WINDOW.
6. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 5, is characterized in that, in described step 2, described PLE parameter management module is by Parameter File described in described PLE parameter read-in.
7. the virtual machine coordinated dispatching method based on dynamic PLE technology as described in claim 1-6 any one, it is characterized in that, in described step 3, when the described VCPU of described virtual machine is scheduled into physical cpu, from described PLE parameter management module, read the described PLE parameter corresponding with described virtual machine, and write in the VCPU control structure of described hardware PLE; When described VCPU triggers PLE, to described cooperative scheduling module, send described schedule information, the information that described schedule information comprises the described VCPU that triggers PLE.
8. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 7, it is characterized in that, in described step 4, described cooperative scheduling module is received after described schedule information, in the described virtual machine at the described VCPU place to triggering PLE, all described VCPU carry out following operation: judge whether described VCPU runs on kernel mode, if so, described VCPU is inserted into the head of the queue of described physical cpu operation queue; If not, do not need to change the position of described VCPU in described physical cpu operation queue.
9. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 8, is characterized in that, the software virtual machine that described method is used is the software virtual machine KVM based on linux kernel.
10. the virtual machine coordinated dispatching method based on dynamic PLE technology as claimed in claim 9, is characterized in that, in described step 3, described Parameter File is the proc file under the described PLE parameter management module root directory being positioned under system/proc catalogue.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410029589.8A CN103744728B (en) | 2014-01-22 | 2014-01-22 | Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410029589.8A CN103744728B (en) | 2014-01-22 | 2014-01-22 | Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103744728A true CN103744728A (en) | 2014-04-23 |
| CN103744728B CN103744728B (en) | 2017-02-08 |
Family
ID=50501748
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410029589.8A Active CN103744728B (en) | 2014-01-22 | 2014-01-22 | Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103744728B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104866370A (en) * | 2015-05-06 | 2015-08-26 | 华中科技大学 | Dynamic time slice dispatching method and system for parallel application under cloud computing environment |
| CN105955796A (en) * | 2016-04-22 | 2016-09-21 | 华中科技大学 | Method and system for dispatching high-performance parallel application on cloud platform |
| CN110795231A (en) * | 2018-08-03 | 2020-02-14 | 阿里巴巴集团控股有限公司 | Processing method and device of virtual CPU |
| CN111209079A (en) * | 2019-12-27 | 2020-05-29 | 山东乾云启创信息科技股份有限公司 | Scheduling method, device and medium based on Roc processor |
| US11194615B2 (en) | 2019-12-05 | 2021-12-07 | Red Hat, Inc. | Dynamic pause exiting |
| CN115145688A (en) * | 2022-06-29 | 2022-10-04 | 科东(广州)软件科技有限公司 | User state virtual machine virtual core suspension method and device |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130007322A1 (en) * | 2011-06-29 | 2013-01-03 | International Business Machines Corporation | Hardware Enabled Lock Mediation |
-
2014
- 2014-01-22 CN CN201410029589.8A patent/CN103744728B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130007322A1 (en) * | 2011-06-29 | 2013-01-03 | International Business Machines Corporation | Hardware Enabled Lock Mediation |
Non-Patent Citations (2)
| Title |
|---|
| 时光等: "Xen中VCPU调度算法分析", 《计算机工程与设计》 * |
| 耿玮: "基于锁感知的多处理器VCPU调度系统的设计", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104866370A (en) * | 2015-05-06 | 2015-08-26 | 华中科技大学 | Dynamic time slice dispatching method and system for parallel application under cloud computing environment |
| CN104866370B (en) * | 2015-05-06 | 2018-02-23 | 华中科技大学 | Towards the dynamic time piece dispatching method and system of Parallel application under a kind of cloud computing environment |
| CN105955796A (en) * | 2016-04-22 | 2016-09-21 | 华中科技大学 | Method and system for dispatching high-performance parallel application on cloud platform |
| CN105955796B (en) * | 2016-04-22 | 2018-12-07 | 华中科技大学 | The dispatching method and system of high performance parallel application in a kind of cloud platform |
| CN110795231A (en) * | 2018-08-03 | 2020-02-14 | 阿里巴巴集团控股有限公司 | Processing method and device of virtual CPU |
| CN110795231B (en) * | 2018-08-03 | 2023-12-22 | 阿里巴巴集团控股有限公司 | Virtual CPU processing method and device |
| US11194615B2 (en) | 2019-12-05 | 2021-12-07 | Red Hat, Inc. | Dynamic pause exiting |
| CN111209079A (en) * | 2019-12-27 | 2020-05-29 | 山东乾云启创信息科技股份有限公司 | Scheduling method, device and medium based on Roc processor |
| CN115145688A (en) * | 2022-06-29 | 2022-10-04 | 科东(广州)软件科技有限公司 | User state virtual machine virtual core suspension method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103744728B (en) | 2017-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Marty et al. | Snap: A microkernel approach to host networking | |
| CN103744728A (en) | Dynamic PLE (pause loop exit) technology based virtual machine co-scheduling method | |
| US10733032B2 (en) | Migrating operating system interference events to a second set of logical processors along with a set of timers that are synchronized using a global clock | |
| CN107025205B (en) | Method and equipment for training model in distributed system | |
| US9507631B2 (en) | Migrating a running, preempted workload in a grid computing system | |
| Vaishnav et al. | Resource elastic virtualization for FPGAs using OpenCL | |
| US8832688B2 (en) | Kernel bus system with a hyberbus and method therefor | |
| AU2014202257B2 (en) | Dynamic virtual machine sizing | |
| CN102096596B (en) | Cloud computing service Cache system based on internal memory template of virtual machine | |
| EP3073373B1 (en) | Method for interruption affinity binding of virtual network interface card, and computer device | |
| WO2014015725A1 (en) | Resource scheduling system and method in graphics card virtualization and based on instant feedback of application effect | |
| CN103942178A (en) | Communication method between real-time operating system and non-real-time operating system on multi-core processor | |
| CN104598294A (en) | Efficient and safe virtualization method for mobile equipment and equipment thereof | |
| Lai et al. | Sol: Fast distributed computation over slow networks | |
| CN103744716A (en) | Dynamic interrupt balanced mapping method based on current virtual central processing unit (VCPU) scheduling state | |
| US20090133099A1 (en) | Methods and systems for transparent software license suspension | |
| CN106250217A (en) | Synchronous dispatching method between a kind of many virtual processors and dispatching patcher thereof | |
| WO2018040845A1 (en) | Method and apparatus for scheduling computing resource | |
| US8977752B2 (en) | Event-based dynamic resource provisioning | |
| CN106815067A (en) | The online moving method of virtual machine, device with I/O virtualizations | |
| CN107528871A (en) | Data analysis in storage system | |
| CN105718320A (en) | Clock task processing method, device and facility | |
| US11556371B2 (en) | Hypervisor task execution management for virtual machines | |
| CN104123174A (en) | Dynamic real-time CPU scheduling system under virtualization environment | |
| CN107479945B (en) | Virtual machine resource scheduling method and device |
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 |