CN101859256B - I/O virtualization method for simplifying mode switch in hardware virtual machine - Google Patents
I/O virtualization method for simplifying mode switch in hardware virtual machine Download PDFInfo
- Publication number
- CN101859256B CN101859256B CN 200910237917 CN200910237917A CN101859256B CN 101859256 B CN101859256 B CN 101859256B CN 200910237917 CN200910237917 CN 200910237917 CN 200910237917 A CN200910237917 A CN 200910237917A CN 101859256 B CN101859256 B CN 101859256B
- Authority
- CN
- China
- Prior art keywords
- communication
- virtual machine
- task
- communication port
- queuing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Computer And Data Communications (AREA)
Abstract
The invention discloses an I/O virtualization method for simplifying mode switch in a hardware virtual machine. n communication channels are established between a privileged virtual machine which operates in a privileged mode and a hardware virtual machine which operates in a non-privileged mode, wherein each channel comprises two shared memory spaces used for communications in two directions, and an event channel is used for transmitting an event signal. A queuing record and an output record are set at two ends of each communication channel and are used for calculating the arrival rate and output rate of the communication tasks at the port of the communication channel. An n-order M/M/1 queuing model is established in the hardware virtual machine and the privileged virtual machine and is used for evaluating the communication delay of each communication channel so as to select the most suitable communication channel for communication. Therefore, the maximum reduction of the communication delay in the sustainable range of the application is guaranteed. Due to the mode switch times caused by I/O communications, the I/O performance of the hardware virtual machine is improved.
Description
Technical field
The present invention relates to the virtual field of hardware virtual machine I/O, particularly relate to the I/O virtual method that a kind of simplifying mode switches.
Background technology
Virtual machine (Virtual Machine:VM) is to support a kind of system of multiple operating system parallel running on the single physical server, and can providing more effectively, bottom hardware uses.In virtual machine, central processor core marks off one section storage area from system's other parts, and operating system and application program operate under " protected mode " environment.Virtual machine monitor (VirtualMachine Monitor:VMM) is a software layer that is used for coordinated management virtual machine in same host, it usually the side by hardware or software part organize to strengthen independence between the virtual machine.It is highest that Xen monitor (Hypervisor) operates in system privilege, has a believable franchise virtual machine to be called domain 0 other virtual machines that are in non-level of privilege of assisting management and be called domain U or guest virtual machine.
Because the virtual machine technique vast potential for future development, it is not poor singly not to be that various virtualization software layer go out; Since 2004, the hardware virtualization technology also began to be listed in project by each large IT circles giant.At present, had the processor hardware technology of supporting the Full-virtualization technology to occur, more representational is the Vanderpool technology of Intel and the Pacifica technology of AMD.On January 20th, 2005, Intel announced the Intel Virtualization Technology that code name is Vanderpool (being called for short VT) to the external world; It is the Intel Virtualization Technology of Pacifica that AMD has also outwards announced code name March 30 the same year, has remedied the virtual leak of X86-based.
VT divides VT-i (supporting the Itanium framework) and VT-x (support X86-based), and influences is based on the VT-x of more common X86-based here.VT-x has introduced a kind of new processor operations, is called VMX (Virtual Machine Extensions).The VMX operation can be in root state (root) or the lower execution of non-state (non-root), and virtual machine monitor (VMM) is carried out under the root state usually, and client software then operates in non-state.VMM can make processor enter non-state of VMX by virtual machine entrance (VM entries), on the contrary, can get back to VMX root state by virtual machine outlet (VM exits).Under the VT technical support, client computer may operate in that its priority (being priority 0) of originally wishing operation goes up and VMM still can catch client computer to the access of particular system resource.
The VMX operation has newly defined virtual machine control structure (Virtual Machine Control Structure, VMCS).Comprise client state (Guest-state area) among the VMCS, Host Status (Host-state area), virtual machine is carried out control domain (VM-execution control fields), virtual machine withdraws from control domain (VM-exit control fields), and virtual machine entrance control domain (VM-entry control fields) and virtual machine withdraw from information field (VM-exit informationfields).When processor control entered non-state of VMX and is the virtual machine entrance from VMX root state, Host Status was stored among the VMCS, and client state and virtual machine withdraw from the control strategy processor of packing into from VMCS simultaneously.Be virtual machine when withdrawing from when processor enters VMX root state from non-state of VMX on the contrary, client state is stored among the VMCS Host Status processor of then packing into from VMCS.VMM can arrange respectively the VMCS of client computer to realize different virtualization policy.
Although the hardware virtualization technology can solve the half virtual problem that needs to revise client operating system nucleus, has also brought new problem.Operate in hardware virtual machine under the non-root mode and want the access privileges resource or carry out privileged instruction such as IO access etc., all need to switch to and carry out under the root mode, need to switch back non-root mode after complete, as shown in Figure 1.Hardware virtual machine switches (pattern that is referred to as is switched) and causes a large amount of overheads between root mode and non-root mode, comprise the preservation of each register of record CPU Run-time scenario and loading etc., it is very little really to account for the time scale that whole pattern switches for the treatment of time of communication request in pattern is switched, therefore there is a large amount of overheads, and then causes the guest virtual machine performance low.
Can know that by above-mentioned analysis all will carry out pattern in the I/O on current hardware virtual machine access switches under the privileged mode and just can carry out, this characteristic directly has influence on the performance of hardware virtual machine.In order to improve the I/O performance of hardware virtual machine, be necessary the characteristic of hardware virtual machine is analyzed, combined with hardware virtual machine and domain 0 operate in respectively nonprivileged mode and privileged mode but share the new I/O virtual method of characteristics design of physical memory.Therefore, the pattern switching times of required generation when the present invention proposes a kind of new I/O virtual method based on shared drive and reducing in a large number hardware virtual machine generation I/O and ask, thereby reduce because I/O accesses the overhead that brings, and then improve the performance of hardware virtual machine.Simultaneously, proposed n rank M/M/1 queuing model based on the M/M/1 queuing model and improved the high problem of delay of simplifying mode handover mechanism, thereby improved the overall performance of hardware virtual machine.
Summary of the invention
The present invention take shared drive as core concept, take waiting line theory as theoretical foundation, solves the low problem of hardware virtual machine I/O access performance to reduce hardware virtual machine I/O access required mode switching times as target.Specifically comprise:
1. at the physical resource layer, take shared drive as the basis, designed efficient two-way hardware virtual machine communication port, be used between privileged mode and nonprivileged mode, transmitting the I/O request of access, thereby so that hardware virtual machine can reduce a large amount of pattern switchings.
2. in application layer, encapsulated the communication interface of bottom, this interface can be packaged into the I/O task with upper strata I/O request of access, is transferred to franchise virtual machine by the shared drive communication mechanism and processes.
3. set up a plurality of shared drive communication ports not of uniform size, be used for the application of different delay requirements, the communication port that shared drive is large has less pattern switching times, but communication delay is high.
4. based on the M/M/1 queuing model of classics, propose n rank M/M/1 queuing model for assessment of all shared drive communication ports, thereby selected wherein to meet the communication port of communication delay demand.
5. be provided with log at hardware virtual machine and domain 0 two ends for each communication port, log recording time of arrival and the time departure of task, be used for the evaluation and test that this communication port is postponed.
Compared with prior art, innovation of the present invention is: caught hardware virtual machine and domain 0 to operate in respectively under nonprivileged mode and the privileged mode, by the shared drive communication mechanism the franchise I/O services request under the non-privilege being sent to franchise vitual stage processes, simultaneously by setting up the optimal communication port of Model Selection, thereby satisfy the demand of using the I/O communication delay.Be embodied in:
1. simplifying mode switches communication mechanism, set up communication port between franchise virtual machine and the non-franchise virtual machine by shared drive, thereby a large amount of patterns that having reduced in a large number non-franchise virtual machine needs to occur because of I/O communication are switched, thereby have improved hardware virtual machine I/O communication performance.
2. set up a plurality of communication delays communication port not of uniform size, thereby between less pattern switching times and lower communication delay, find equilibrium for communications applications by setting up model.
The simplifying mode that the present invention proposes in a kind of hardware virtual machine switches the I/O virtual method, the method realizes in the system with following structure: this system has the Xen monitor, this Xen monitor is supervised other as virtual machine monitor and is operated in virtual machine on the actual physics machine, this system has franchise virtual machine, this privilege virtual machine has all authorities, can directly access physical resource, this system's support hardware is virtual, support Intel VT technology, has hardware virtual machine, hardware virtual machine need to switch to carry out I/O communication by pattern, the method is characterized in that to comprise the steps:
1) the Xen monitor be set up between hardware virtual machine and the franchise virtual machine many to physics shared drive not of uniform size, authorize both by authorization list the shared drive communication port is had the read-write operation authority, be that every pair of physics shared drive creates an event channel, the event channel two ends connect respectively and operate in the hardware virtual machine under the nonprivileged mode and operate in domain 0 under the privileged mode;
2) every pair of shared drive and corresponding event channel consist of a communication port, set up two records at each communication port two ends, record is comprised of annular chain meter, be used for being recorded in all communication task time of arrival of the sensible end of communication and finishing time departure the arrival rate (being the quantity that p.s., task arrived) by calculating communication task these writing times and the completion rate (being the quantity of finishing the work p.s.) of task;
3) set up n rank M/M/1 queuing model in hardware virtual machine and domain 0, the average communication that this model utilizes task arrival rate in the step (2) and task completion rate to calculate each shared drive communication port in this system postpones;
4) the I/O communication request is arranged or in domain 0, have when needing with request that hardware virtual machine I/O communicates by letter when hardware virtual machine, the delay of at first calculating each communication port by the model in the step (3), selection has the communication port that proper communication postpones, the I/O communication request is packaged into communication task, send to the opposing party by this communication port, the opposing party resolves the I/O communication task and processes;
Three pairs of shared drives are arranged between hardware virtual machine and domain 0, every pair comprises the equally large shared drive of size for the communication on the both direction, three pairs of shared drive sizes are respectively: 2 shared drive page table sizes, 4 shared drive page table sizes, (each shared drive page table size is 4096 bytes to 8 shared drive page table sizes, the page table size is fixing, is determined by system performance, and the size of the shared drive of communication usefulness can be regulated).
Log is divided into two classes: arrive task record and the record of finishing the work, each record is the fifo queue that contains 50 nodes, comprise two attribute: task_ID, time, task_ID in the data structure of each node and be ID number of each task of unique identification, time is time of arrival or the deadline of this task, and this formation is upgraded by constantly inserting new node toward tail of the queue.
Description of drawings
Fig. 1 Xen traditional mode switches I/O communication architecture figure
Fig. 2 simplifying mode switching architectures figure
Fig. 3 I/O communication task encapsulating structure figure
Fig. 4 n rank M/M/1 queuing model block diagram
The M/M/1 queuing model constitutional diagram of Fig. 5 n rank
Fig. 6 communication channel is selected modular design figure
Fig. 7 interrecord structure figure
Embodiment
Consult Fig. 2, the present invention is by the comprehensive Design of shared drive and event channel, sets up a plurality of direct communication ports operating in the domain 0 under the privileged mode and operate between the hardware virtual machine under the nonprivileged mode.Based on this, by in hardware virtual machine and domain 0 encapsulation I/O communication request, set up the optimal communication port of Model Selection and communicate, thus the pattern blocked operation number of times that causes in the time of can reducing in a large number traditional hardware virtual machine I/O communication.Have three pairs of shared drives not of uniform size between hardware virtual machine and domain 0, every pair of shared drive comprises two unidirectional shared drives for the communication on the both direction.Be that every pair of shared drive creates an event channel simultaneously, event channel is used for transmitting the signal event of informing and occurs between hardware virtual machine and domain 0, the every pair of shared drive and an event channel jointly form simplifying mode and switch a communication channel in communicating by letter.The parameter that communication selects module to be used for providing according to the model of setting up when communication request arrives is selected suitable communication channel.The communication package module is used for the I/O communication request is packaged into communication task, and all I/O communication request send to communication objective ground with the form of communication task in simplifying mode switching communication system.The communications module is used for packaged communication task is sent to communication objective ground by communication port.It is to set up shared drive and event channel between hardware virtual machine and the domain 0 that communication management module in domain 0 is used for when initial.
Consult Fig. 3, the present invention is packaged into communication task with the I/O communication request when receiving communication request, and communication task data packet head length is 8 bytes (64).Be successively:
A) Type: account for 8 bits, represent the type of this I/O communication request, such as network, IO access etc.;
B) Domain ID: account for 8 bits, represent the destination, source of this communication data, each virtual machine on the Domain ID energy unique identification physical machine;
C) Totallen: account for 16 bits, represent the total length (take byte as unit) of this task data bag;
D) Datalen: account for 16 bits, represent the total length (take byte as unit) of data in this task data bag;
E) Timestamp: account for 16 bits, represent the time that this task data bag creates;
F) Data: maximum 4088 bytes, depositing the data of communication;
The I/O communication request can be packaged into one or more communication task the communication initiator, selects a communication port to send to communication objective ground, obtains the I/O communication request and then processes thereby communication objective side adopts same form to resolve communication task.
The I/O communication request is admitted to the suitable communication channel of selection in the M/M/1 queuing model of n rank after being packaged into communication task, as shown in Figure 4.The pacing items that meets the M/M/1 queuing model is:
A) input process of system be N (t), t 〉=0}, for parameter is the Poisson process of λ, namely time of arrival intervening sequence { J
k, k 〉=0} is sequence of random variables, and I
1～Γ (1, λ).
B) only has one-server, each client's sequence (B service time
k, k 〉=1} is sequence of random variables, and B
1～Γ (1, μ).And establish { B
k, k 〉=1} and (J
k, k 〉=0} is independent.
The probability that does not have client in the system, namely client arrives does not need the probability P waited for
0=1-ρ, wherein ρ=λ/μ is the utilization factor of service.N rank M/M/1 queuing model forms by n M/M/1 queuing model is parallel, and n is that 3, n can arrange in the present invention.The service rate of the information desk in these models is different.In this system, the I/O communication request that is arranged in hardware virtual machine is the client of waiting line theory, and it is λ according to parameter
iPoisson distribution arrive, during as input system, at first calculate the stand-by period of each instantaneous communication port, select to have the communication port that is fit to the stand-by period and enter, have eggs on the spit such as the service window of this communication port, then wait in line service.The information desk of each communication port was negative exponent and distributes for time of each customer service, remembered that (mean service rate of 1≤i≤n) is μ to i communication port
i, the arrival rate of customers that enters the i sub-systems is λ
iBy the n rank M/M/1 queuing model constitutional diagram that forms arranged side by side of the individual independently M/M/1 queuing of n submodel as shown in Figure 5, state is 00 during system initialization, and namely all communication ports all are empty, without client.After arriving, client select a suitable communication port to enter by comparing current average latency and the demand for services of self of each communication port, thus so that this communication port redirect between different states.
When having an I/O communication request to arrive, defining its communication delay patience is T
Bear(namely the patient maximum communication of this I/O signal post postpones, follow-up will select the communicator passage with this).The average retardation that communication channel in the simplifying mode switching I/O virtual method selects module to calculate each communication port according to the Delay computing module is selected which communication channel is communication request put into, as shown in Figure 6.The Delay computing module adds by the average latency with communication port and obtains the average retardation of this communication port the service time of this communication port:
Obtained the average retardation T of each communication channel by said method
0, T
1, T
2... T
i... T
n, 0≤i≤n wherein.For the average retardation of the communication channel of the selection of satisfying communication request must be less than the delay patience T of this communication request
BearBecome in the communication channel that satisfies above-mentioned condition
S＝{T
j≤T
bear|0≤j≤n}
On above-mentioned collection basis, should from the communication channel that satisfies condition, select larger better of shared drive.Therefore, we select to satisfy k communication channel of following condition:
T
k=max{T
0, T
1... T
p... T wherein
p∈ S
Observation formula (1) calculates the average retardation of each communication port as can be known need to know arrival rate of customers and the service rate of this communication port in this time.The different communication channel has different arrival rate of customers and service rate, simultaneously, same communication channel also has different arrival rate of customers and service rate constantly in difference, in order to obtain comparatively accurately arrival rate of customers and service rate, the present invention is provided with the queuing record and in output place the output record is set at the queue queue place of each communication channel, is mainly used in the time of log history client arrival and the time of leaving system.Both record client's arrival with the list structure that 50 nodes are arranged and leave, and each nodes records a client's corresponding information, and structure comprises in the Wait_Queue structure of queuing end as shown in Figure 7:
A) task_ID: be used for unique identification each enter the communication task of this communication port;
B) time: be used for recording the time that enters this communication port queue queue;
In the Output_Queue structure of queuing end, comprise:
A) task_ID: be used for each communication task from the output of this communication port of unique identification;
B) time: be used for recording the time from this communication port output;
At the waiting list end, the annular chain meter that the present invention is 50 nodes with a length records client's arrival situation as fifo queue, whenever a new communication request enters subsystem, current time and task ID number are inserted in the tail of the queue node Wait_Queue data structure of this formation.Thereby use annular chain meter to avoid dynamic creation and deletion of node, new node adds only needs the old nodal information of covering just passable.Suppose that when t m node arranged in the waiting list, be t writing time in each node
1, t
2, t
3... t
m, 0≤m≤50 wherein.Then the communication arrival rate average at this i sub-systems in this moment is:
At the output queue end, same, the annular chain meter that the present invention is 50 nodes with a length records client's output situation as fifo queue, whenever processed finishing of new communication request, current time and task ID number are inserted in the tail of the queue node Output_Queue data structure of this formation.Thereby use annular chain meter to avoid dynamic creation and deletion of node, new node adds only needs the old nodal information of covering just passable.Suppose that when t m node arranged in the output queue, be t writing time in each node
1, t
2, t
3... t
m, 0≤m≤50 wherein.Then the communication output rating average at this i sub-systems in this moment is:
Obtain the rate that reaches of each communication port and the communication delay that output rating can calculate each communication port, thereby select suitable communication port to communicate according to this communication delay.
Claims (2)
1. the I/O virtual method that simplifying mode switches in the hardware virtual machine, the method realizes in the system with following structure: this system has the Xen monitor, this Xen monitor is supervised other as virtual machine monitor and is operated in virtual machine on the actual physics machine, this system has franchise virtual machine, this privilege virtual machine has all authorities, operate under the privileged mode, can directly access physical resource, this system also has the hardware virtual machine that operates under the nonprivileged mode, hardware virtual machine under the nonprivileged mode need to carry out pattern and switch, carry out I/O communication under the privileged mode to switch to, the method is characterized in that to comprise the steps:
1) the Xen monitor is to distribute n to shared drive not of uniform size between franchise virtual machine and the hardware virtual machine, comprises two equirotal shared drives in the every pair of shared drive and is respectively applied to I/O communication on the both direction; The Xen monitor is that every pair of shared drive distributes an event channel, is used for pass-along message between hardware virtual machine and franchise virtual machine; Every pair of shared drive and corresponding event channel form a communication port;
2) set up n rank M/M/1 queuing model at hardware virtual machine and franchise virtual machine, this queuing model is made of side by side n M/M/1 queuing submodel; The corresponding communication port of each M/M/1 queuing submodel is for the communication delay that calculates this communication port;
3) respectively set up two records for each communication port two ends: queuing record and output record; Queuing record be used for record enters queuing from this communication port port communication task with and enter time of communication port, the output record be used for record from the communication task of this communication port output with and time of output;
4) when having the I/O communication request to arrive, at first in application layer the I/O communication request is encapsulated, it is packaged into one or more communication task; Then calculate the communication task arrival rate of this communication port according to described queuing record and output recording gauge, described arrival rate is the number that p.s., communication task arrived, and the output rating of communication task, described output rating is the number of communication task output p.s., thereby the n rank M/M/1 queuing model of arrival rate and output rating parameter being passed to described foundation calculates each communication port communication delay, and then select a communication port to communicate, described selection communication port is according to following rule: select the communication port of the communication delay that satisfies communication task from all communication ports, namely the communication delay of communication port is not more than the communication delay demand of communication task; And then the communication port of selecting to have the maximum communication delay in the communication port of selecting communicates;
Wherein n is 3; The exponent number of queuing model and the logarithm of shared drive are consistent;
The communication delay that calculates i communication port uses formula
Calculate, wherein λ
iBe the communication task arrival rate of this communication port, μ
iCommunication task output rating for this communication port.
2. method according to claim 1, it is characterized in that: in the step 4), the I/O communication request is packaged into communication task, the packet header length of communication task is 8 bytes: the type that accounts for 1 byte is used for representing communication task types, the Domain ID that accounts for 1 byte represents the destination, source of this communication data, the Totallen that accounts for 2 bytes represents the total length of this task data bag, the Datalen that accounts for 2 bytes represents the total length of data in this task data bag, and the Timestamp that accounts for 2 bytes represents the time that this task data bag creates.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910237917 CN101859256B (en) | 2009-11-25 | 2009-11-25 | I/O virtualization method for simplifying mode switch in hardware virtual machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 200910237917 CN101859256B (en) | 2009-11-25 | 2009-11-25 | I/O virtualization method for simplifying mode switch in hardware virtual machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101859256A CN101859256A (en) | 2010-10-13 |
| CN101859256B true CN101859256B (en) | 2013-03-06 |
Family
ID=42945180
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 200910237917 Expired - Fee Related CN101859256B (en) | 2009-11-25 | 2009-11-25 | I/O virtualization method for simplifying mode switch in hardware virtual machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101859256B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102034046B (en) * | 2010-12-10 | 2012-10-03 | 北京世纪互联工程技术服务有限公司 | Data underlay encryption method based on disk drive in cloud computing environment |
| US9122508B2 (en) * | 2012-06-15 | 2015-09-01 | International Business Machines Corporation | Real time measurement of I/O interrupt delay times by hypervisor by selectively starting and/or stopping corresponding LPARs |
| DE102014019240A1 (en) * | 2014-12-19 | 2016-07-07 | Unify Gmbh & Co. Kg | Telecommunication system and method for flexible control of the telecommunication system by a issued by an application to a platform switching order |
| KR102549540B1 (en) * | 2017-09-22 | 2023-06-29 | 삼성전자주식회사 | Storage device and method of operating the same |
| CN107832117A (en) * | 2017-11-15 | 2018-03-23 | 北京东土科技股份有限公司 | A kind of virtual machine state information synchronous method and electronic equipment |
| CN109597671B (en) * | 2018-12-11 | 2022-06-10 | 国云科技股份有限公司 | Desktop cloud implementation method compatible with centralized and distributed operation |
| CN115309511B (en) * | 2022-09-28 | 2023-04-04 | 亿咖通(湖北)技术有限公司 | Xen-based data interaction method and device, storage medium and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101419535A (en) * | 2008-11-19 | 2009-04-29 | 北京航空航天大学 | Distributed virtual magnetic disc system of virtual machine |
| CN101425046A (en) * | 2008-10-28 | 2009-05-06 | 北京航空航天大学 | Method for implementing distributed I/O resource virtualizing technique |
-
2009
- 2009-11-25 CN CN 200910237917 patent/CN101859256B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101425046A (en) * | 2008-10-28 | 2009-05-06 | 北京航空航天大学 | Method for implementing distributed I/O resource virtualizing technique |
| CN101419535A (en) * | 2008-11-19 | 2009-04-29 | 北京航空航天大学 | Distributed virtual magnetic disc system of virtual machine |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101859256A (en) | 2010-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101859256B (en) | I/O virtualization method for simplifying mode switch in hardware virtual machine | |
| US10922269B2 (en) | Proactive optimizations at multi-tier file systems | |
| US10447806B1 (en) | Workload scheduling across heterogeneous resource environments | |
| US8832688B2 (en) | Kernel bus system with a hyberbus and method therefor | |
| CN106502792B (en) | A kind of multi-tenant priority scheduling of resource method towards different type load | |
| CN1794177B (en) | Systems and methods for exposing processor topology for virtual machines | |
| US9747136B2 (en) | Methods and systems that allocate cost of cluster resources in virtual data centers | |
| CN102609298B (en) | Based on network interface card virtualization system and the method thereof of hardware queue expansion | |
| EP3242440B1 (en) | Fault tolerant method, apparatus and system for virtual machine | |
| US11347558B2 (en) | Security-aware scheduling of virtual machines in a multi-tenant infrastructure | |
| US11579908B2 (en) | Containerized workload scheduling | |
| CN101706742A (en) | Method for dispatching I/O of asymmetry virtual machine based on multi-core dynamic partitioning | |
| US10852990B2 (en) | Hybrid framework of NVMe-based storage system in cloud computing environment | |
| US9389910B2 (en) | Paravirtualized migration counter for migrating a virtual CPU to a different physical CPU | |
| US10860352B2 (en) | Host system and method for managing data consumption rate in a virtual data processing environment | |
| CN104335180A (en) | Real time measurement of virtualization I/O processing delays | |
| CN105630575B (en) | For the performance estimating method of KVM virtualization server | |
| Pongsakorn et al. | Container rebalancing: Towards proactive linux containers placement optimization in a data center | |
| CN107220125A (en) | A kind of cloud resource dispatching method and device | |
| Clemente-Castelló et al. | Enabling big data analytics in the hybrid cloud using iterative mapreduce | |
| Moody et al. | Reconfigurable network testbed for evaluation of datacenter topologies | |
| Tadesse et al. | Characterizing the power cost of virtualization environments | |
| US9710296B2 (en) | Allocating cost of disk usage to a linked clone virtual machine | |
| Pretto et al. | Boosting HPC applications in the cloud through JIT traffic-aware path provisioning | |
| Lukashin et al. | Resource scheduler based on multi-agent model and intelligent control system for OpenStack |
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: 20130306 Termination date: 20141125 |
|
| EXPY | Termination of patent right or utility model |