CN105718310A - Virtual machine scheduling method for IO dense application under cloud platform - Google Patents
Virtual machine scheduling method for IO dense application under cloud platform Download PDFInfo
- Publication number
- CN105718310A CN105718310A CN201610021270.XA CN201610021270A CN105718310A CN 105718310 A CN105718310 A CN 105718310A CN 201610021270 A CN201610021270 A CN 201610021270A CN 105718310 A CN105718310 A CN 105718310A
- Authority
- CN
- China
- Prior art keywords
- virtual machine
- physical host
- load
- resource
- intensive type
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/46—Multiprogramming arrangements
- G06F9/48—Program initiating; Program switching, e.g. by interrupt
- G06F9/4806—Task transfer initiation or dispatching
- G06F9/4843—Task transfer initiation or dispatching by program, e.g. task dispatcher, supervisor, operating system
- G06F9/4881—Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/455—Emulation; Interpretation; Software simulation, e.g. virtualisation or emulation of application or operating system execution engines
- G06F9/45533—Hypervisors; Virtual machine monitors
- G06F9/45537—Provision of facilities of other operating environments, e.g. WINE
Abstract
The present invention provides a virtual machine scheduling method for a dense application under a cloud platform. The method comprises: S1: setting a threshold interval [M, N]; S2: setting efficiency of running an IO dense application by each physical host; S3: presetting 4 set tables T1-T4, and initializing the set tables T1-T4; S4: keeping waiting for the IO dense application, and if the IO dense application appears, entering step S5 while waiting for a new IO dense application; S5: determining whether the set table T1 is null, and if no, allocating the IO dense application to a virtual machine started by a physical host in the set table T1 and running the IO dense application, performing load determination on the physical host, and entering step S6; and S6: determining whether the set table T4 is null, and if no, allocating the IO dense application to a virtual machine started by a newly started physical host and running the IO dense application, and performing load determination on the physical host. According to the method provided by the present invention, for the IO dense application, the virtual machine can be allocated to the physical host in real time, so as to ensure the total usage number of the physical host is minimized.
Description
Technical field
The present invention relates to field of cloud calculation, particularly to be I/O intensive type (IO Bound under cloud platform
Application) dispatching method of virtual machine applied.
Background technology
The appearance of cloud computing changes the function mode of conventional data centers, and the various aspects in society all create huge
Impact.Increasing data center, provides powerful resource pool, such as CPU, Memory by Intel Virtualization Technology and (deposits
Reservoir), Bandwidth (bandwidth) etc..But, the most more efficiently promote the resource utilization of cloud computing and reduce energy
Consumption, then become one of hot fields of research at present.
Intel Virtualization Technology is to build one of indispensable key technology of cloud architecture.At present in cloud computing data
For the heart, use Intel Virtualization Technology aspect, improve utilization rate and the treatment effeciency of application of resource, mainly from the standpoint of two:
1) load balancing, the request of user with the form of virtual machine, is averagely allocated to Servers-all so that the thing of cloud data center
The load of reason server is kept in balance state, it is therefore intended that ensure the efficiency that application processes;2) virtual machine is merged, as much as possible
Virtual machine is merged and is deployed on the fewest physical server, make full use of the resource of physical server, reduce and use thing
The quantity of reason server, reduces the energy consumption of whole cloud computation data center with this.CloudStack (the one of cloud operating system
Kind) in system platform, typical resource dispatching strategy is FirstFit strategy, enters at cluster, the Host List specifying user
When row sequencing selection cluster or main frame, directly or dispose as virtual machine distribution with first of satisfactory cluster or main frame
Target, but application treatment effeciency the most relatively low.
Summary of the invention
Purpose to be solved by this invention is to provide the dispatching method of virtual machine of I/O intensive type application, pin under a kind of cloud platform
I/O intensive type is applied so that virtual machine can distribute to physical host in real time, ensure the use sum of physical host the most simultaneously
Little, promote cloud data center and process the efficiency of I/O intensive type application and reduce its process energy consumption to a certain extent.
For solving the problems referred to above, the present invention proposes the dispatching method of virtual machine of I/O intensive type application, bag under a kind of cloud platform
Include following steps:
S1: setting threshold interval [M, N], wherein N is the virtual machine number in efficiency zone of reasonableness of I/O intensive type application
Maximum critical value, the marginal value of virtual machine number when M is to maximally utilize resource;
S2: set every physical host and run efficiency E of I/O intensive type applicationIO, wherein,
kIORepresent the virtual machine number of actual motion I/O intensive type application;
S3: preset 4 collection table T1-T4, initializes collection table T1-T4, and wherein, collection table T1 is EIOThe physics master of=0
Machine set;Collection table T1 is EIOThe physical host set of=1;Collection table T1 is EIOThe physical host set of=-1;Collection table T1
For the physical host set do not opened;
S4: persistently waiting the application of I/O intensive type, I/O intensive type is applied if occurring, then waiting the same of new I/O intensive type application
Time enter step S5;
S5: judge whether collection table T1 is empty, if sky, then enters step S6;If not being empty, then by this I/O intensive
Type application is distributed to run in the virtual machine that the physical host in described collection table T1 is opened, and updates collection table T1-T4, to this
Physical host carries out load and judges, makes efficiency E in loadIOMigrate out a certain amount of virtual machine time not in preset range, enter
Enter step S6;
S6: judge whether collection table T4 is empty, if sky, then refusal distribution physical host, waits pending described new IO
Intensive applications, returns step S5;If not being empty, then start a new physical host, and update collection table T1-T4, and
The virtual machine apply the physical host distributing to this new startup to be opened this I/O intensive type running, this physical host is carried out
Load judges, makes efficiency E in loadIOA certain amount of virtual machine is migrated out time not in preset range.
According to one embodiment of present invention, in step sl, define a five-tuple threshold determination model T, comprehensively weigh
Amount I/O load, so that it is determined that physical host can bear the threshold interval of the virtual machine number of I/O intensive type application,
T={loadavg, run, avgqu, util, contexts},
Loadavg represents that CPU is currently running the statistical information of the process number run with waiting for CPU, run within a period of time
Representing the length of instantaneous operation queue or wait the process number run, avgqu represents average I/O queue, IOQ length, and util represents one
For the percentage of time of I/O operation in Miao, contexts represents context handover overhead;
Each load is exceeded the percentage ratio loaded most and is set to value when efficiency reduces by described five-tuple threshold determination model T,
Each load is set to value when resource utilization reduces less than the percentage ratio of maximum load, sets the load of each load less than maximum
The percentage ratio of load, load exceed the interval between the percentage ratio of maximum load and load as embodying each of I/O load parameter
Excellent threshold interval, thus the threshold interval of 5 virtual machine numbers of correspondence is determined according to optimal threshold interval;
Using the common factor of the threshold interval of 5 virtual machine numbers as described threshold interval [M, N],
[M, N]=[Mloadavg, Nloadavg]∩[Mrun, Nrun]∩[Mavgqu, Navgqu]∩[Mutil, Nutil]∩[Mcontext,
Ncontext]
According to one embodiment of present invention, set the load of each load less than the percentage ratio of maximum load as 50%, negative
It is 70% that load exceedes the percentage ratio of maximum load, and optimal threshold interval is 50%~70%.
According to one embodiment of present invention, virtual machine number k of described actual motion I/O intensive type applicationIOThere is a limit
Scope processed, the determination mode of described restriction scope includes:
Definition physical host resource type, a given physical host, then this physical host has multiple different physics
Resource, uses RnRepresenting, n represents the numbering of the different resource type of physical host, is expressed as with resource collection R:
R={R1, R2…Rn…RNN ∈ 1,2 ..., N} (1);
Definition type of virtual machine, virtual machine then also can be divided into different virtual machines according to the difference of application resource requirement
Type Vi, each type of virtual machine ViRequired every resource is different, rI, nRepresent ViThe resource characteristic that type Virtual machine consumes,
It is corresponding with the resource type of physical host,
V={V1, V2…Vi…VmI ∈ 1,2 ..., m} (2)
rI, n={ rI, 1, rI, 2…rI, n…rI, NN ∈ { 1,2 ... N} (3);
Definition deploying virtual machine number restrictive condition, a physical host can only be disposed a number of within the same time
Virtual machine ki, i ∈ 1,2 ..., N}, kiRepresent ViThe virtual machine number of type Virtual machine, therefore deploying virtual machine number restrictive condition
It is indicated by equation below:
∑i(ki*Vi)≤αiRi, 0 < α < 1 (4)
Substitution, derivation according to definition type of virtual machine then have
Wherein, αiIt is then correlation coefficient, is used for determining the utilization rate limit of related resource;
Definition only considers the I/O type virtual machine number restrictive condition under I/O resource consumption,
rIO, nRepresenting the resource characteristic of I/O type virtual machine, n difference represents the parameter difference that I/O resource consumes, kIORepresent real
The virtual machine number of I/O intensive type application is run on border;Represent the virtual machine that physical host can be opened
I/O resource;αIORIORepresent the I/O resource of physical host;
Virtual machine number k of actual motion I/O intensive type applicationIORestriction scope determine according to formula (4), (5), (6),
In the range of described restriction, then this physical host can receive new I/O intensive type application always, otherwise, then cannot be the most newly-built virtual
Machine, i.e. on this physical host, virtual machine unlatching quantity reaches maximum.
According to one embodiment of present invention, the virtual machine number in efficiency zone of reasonableness of described I/O intensive type application
Maximum critical value N is the upper limit of described restriction scope.
According to one embodiment of present invention, in described step S1, described threshold interval [M, N] is according to the change of I/O load
Change and the time of I/O intensive type application process sets.
According to one embodiment of present invention, physical host carries out load judgement comprise the following steps:
A1: physical host receives the application of I/O intensive type and runs on the virtual machine opened;
A2: judge EIOWhether >=0 set up;
A3: if setting up, entering step A4, otherwise migrating out a number of virtual machine, enter step A4;
A4: judge EIOWhether=1 set up;
A5: if being false, migrates out a number of virtual machine and returns to step A4, does not the most migrate virtual machine.
According to one embodiment of present invention, in step S5 and S6, this physical host is carried out load and judges, work as EIO
When=-1, virtual machine is moved out.
After using technique scheme, the present invention has the advantages that compared to existing technology by physical host
Load situation of change migrates virtual machine, the quantity of the I/O intensive type application on conservative control physical server, and then improves cloud ring
Under border, physical host processes the efficiency of I/O intensive type application, and reduces energy consumption to a certain extent, by true cloud environment
Under great many of experiments analysis and Experimental comparison, the present invention is always held at most preferably on the treatment effeciency that I/O intensive type is applied, thing
Within the I/O load of reason server also maintains essentially in normal range, it is to avoid when the application of I/O intensive type increases, I/O intensive type
The phenomenon that application treatment effeciency reduces and physical server I/O load is overweight, enters towards dissimilar application under nowadays cloud environment
The specific resource dispatching strategy of row has good realistic meaning.
Accompanying drawing explanation
Fig. 1 be one embodiment of the invention cloud platform under I/O intensive type application dispatching method of virtual machine flow process signal
Figure;
Fig. 2 is that the physical host of one embodiment of the invention carries out the schematic flow sheet that load judges;
Fig. 3 is the avgqu relation with the virtual machine number of actual motion I/O intensive type application of one embodiment of the invention
Schematic diagram;
Fig. 4 is that the relation of the virtual machine number that the util of one embodiment of the invention applies with actual motion I/O intensive type is shown
It is intended to;
Fig. 5 is the contexts pass with the virtual machine number of actual motion I/O intensive type application of one embodiment of the invention
It it is schematic diagram;
Fig. 6 is loadavg, run virtual machine number with the application of actual motion I/O intensive type of one embodiment of the invention
Relation schematic diagram;
Fig. 7 is when virtual machine=8, the contrast schematic diagram that one embodiment of the invention is tactful with FirstFit;
Fig. 8 is when virtual machine=9, the contrast schematic diagram that one embodiment of the invention is tactful with FirstFit.
Detailed description of the invention
Understandable, below in conjunction with the accompanying drawings to the present invention for enabling the above-mentioned purpose of the present invention, feature and advantage to become apparent from
Detailed description of the invention be described in detail.
Elaborate a lot of detail in the following description so that fully understanding the present invention.But the present invention can be with
Much being different from alternate manner described here to implement, those skilled in the art can be in the situation without prejudice to intension of the present invention
Under do similar popularization, therefore the present invention is not limited by following public being embodied as.
Referring to Fig. 1, the dispatching method of virtual machine of I/O intensive type application under the cloud platform of the present embodiment, comprise the following steps:
S1: setting threshold interval [M, N], wherein N is the virtual machine number in efficiency zone of reasonableness of I/O intensive type application
Maximum critical value, the marginal value of virtual machine number when M is to maximally utilize resource;
S2: set every physical host and run efficiency E of I/O intensive type applicationIO, wherein,
kIORepresent the virtual machine number of actual motion I/O intensive type application;
S3: preset 4 collection table T1-T4, initializes collection table T1-T4, and wherein, collection table T1 is EIOThe physics master of=0
Machine set;Collection table T1 is EIOThe physical host set of=1;Collection table T1 is EIOThe physical host set of=-1;Collection table T1
For the physical host set do not opened;
S4: persistently waiting the application of I/O intensive type, I/O intensive type is applied if occurring, then waiting the same of new I/O intensive type application
Time enter step S5;
S5: judge whether collection table T1 is empty, if sky, then enters step S6;If not being empty, then by this I/O intensive
Type application is distributed to run in the virtual machine that the physical host in described collection table T1 is opened, and updates collection table T1-T4, to this
Physical host carries out load and judges, makes efficiency E in loadIOMigrate out a certain amount of virtual machine time not in preset range, enter
Enter step S6;
S6: judge whether collection table T4 is empty, if sky, then refusal distribution physical host, waits pending described new IO
Intensive applications, returns step S5;If not being empty, then start a new physical host, and update collection table T1-T4, and
The virtual machine apply the physical host distributing to this new startup to be opened this I/O intensive type running, this physical host is carried out
Load judges, makes efficiency E in loadIOA certain amount of virtual machine is migrated out time not in preset range.
Below the dispatching method of virtual machine of I/O intensive type application under the cloud platform of the present embodiment is described in detail.
First, determining that physical host processes the limit of I/O intensive type application, this limit refers to do not affecting SLA (seeervice level
Other agreement) in the case of service regulation, physical host can carry how many I/O intensive type virtual machines (how many I/O intensive simultaneously
Type application task).In one embodiment, to the resource of physical host, type of virtual machine, deploying virtual machine number restrictive condition
It is analyzed, is defined below.
Definition 1:(physical machine resource type) a given physical host, then this physical host has multiple different physics
Resource, including cpu resource, memory source etc., uses RnRepresenting (i.e. representing different disposal abilities), n represents physical host not
Same resource type.Such as R1Represent cpu resource, R2Represent memory source etc..It is expressed as with set R:
R={R1, R2…Rn…RNN ∈ 1,2 ..., N} (1)
Definition 2:(type of virtual machine), for specific application, cloud data center distributes corresponding virtual machine.And virtual machine
Then according to the difference of application resource requirement, and different type of virtual machine V can be divided intoi(i.e. the resource characteristic of virtual machine consumption divides
Class).And each type of virtual machine ViRequired every resource is again different, rI, nRepresent the money of the consumption of i type Virtual machine
Source feature, corresponding with the resource type of physical host.ViType Virtual machine can be that I/O intensive type, CPU intensive type, internal memory are close
These several types such as collection type or network bandwidth intensity.
V={V1, V2…Vi…VmI ∈ 1,2 ..., m} (2)
rI, n={ rI, 1, rI, 2…rI, n…rI, N}n∈{1,2…N} (3)
Definition 3:(deploying virtual machine number restrictive condition) because Intel Virtualization Technology portion can be passed through on a physical host
Administration's multiple stage virtual machine, but be not that unlimited amount is disposed, because physical host configuration is limited, a physical host is in the same time
In can only carry a number of virtual machine ki, i ∈ 1,2 ..., N}, kiRepresent ViThe virtual machine number of type Virtual machine, thus empty
Plan machine is disposed number restrictive condition and can be indicated by equation below:
∑i(ki*Vi)≤αiRi, 0 < α < 1 (4)
And by definition 2, derive and then have
Wherein, αiBe then correlation coefficient, every physical host due to the impact of the factor such as system, hardware, system resource
Utilization rate is not up to 100%, so according to the situation of its reality, having parameter alphaiDetermine the utilization rate pole of related resource
Limit.
The present invention be directed to the application of I/O intensive type and carry out scheduling virtual machine.Therefore only need to pay close attention to the most main in terms of resource limit
The resource consumption wanted, i.e. I/O resource, for other resources, will not become physical host and process the bottleneck of I/O intensive application.Therefore have
Equation below is set up:
rIO, nRepresenting the resource characteristic of I/O type virtual machine, i.e. consume the major parameter of I/O resource, n is different, represents IO money
The parameter that source consumes is different, kIORepresent the virtual machine number running the application of I/O intensive type;Represent institute's energy
The virtual machine I/O resource opened;αIORIORepresent physical host I/O resource.Therefore the virtual machine I/O resource opened not can exceed that place thing
I/O resource on reason main frame.
According to above-mentioned definition, when cloud data center processes one group of I/O intensive type application, if the reality on physical host
Run virtual machine number k of I/O intensive type applicationIOCoincidence formula (4), (5), (6), can answer as actual motion I/O intensive type
Virtual machine number kIORestriction scope, then this physical host can receive the application of new I/O intensive type always.Otherwise, then because
Physical machine inadequate resource, it is impossible to the most newly-built virtual machine, i.e. on this physical host, virtual machine unlatching quantity reaches maximum.
In step sl, for I/O intensive type apply, herein set I/O intensive type application at efficiency zone of reasonableness interior-deficiency
The maximum critical value of plan machine number is N, and when maximally utilizing resource, the marginal value of virtual machine number is M, setting threshold interval [M,
N], it is E that physical host processes the efficiency of I/O intensive type applicationIO。
Efficiency E that I/O intensive type is applied is processed weighing physical hostIOTime, can be directly from the process of I/O intensive type application
Time change draws.But it is when processing application under cloud environment, it is impossible to carry out scheduling virtual machine by running the deadline, this
Cold migration only can increase system load, reduces resource treatment effeciency and increases energy consumption.So, hold on physical host to find out
Carry the maximum critical value N of virtual machine number in efficiency zone of reasonableness of I/O intensive type application, and maximally utilize during resource empty
Marginal value M of plan machine number, can take react the most representational several parameters of I/O load situation, comprehensively weigh
Threshold interval and the situation of I/O load of the virtual machine number of I/O intensive type application can be born on one physical host.
In one embodiment, in step sl, define a five-tuple threshold determination model T, comprehensively weigh I/O load,
So that it is determined that physical host can bear the threshold interval of the virtual machine number of I/O intensive type application,
T={loadavg, run, avgqu, util, contexts},
Loadavg represents that CPU is currently running the statistical information of the process number run with waiting for CPU within a period of time, enters
The average statistical value of number of passes, i.e. cpu load, the statistical information of the length of namely CPU use queue.General loadavg is being
In system, maximum is 2 processes.The too high typically performance of system loadavg is problematic, and system loadavg reflects cpu often
One important performance characteristic of load, i.e. system loadavg height then represents that CPU capability problems or quantity are inadequate.Appointing in queue
Pragmatic border is probably consumption CPU's, it is also possible to consumption I/O's, the present invention is directed to the application of I/O intensive type and uses physical host resource
The impact of situation, ignores the task of consumption CPU.Referring to Fig. 6, (load the most) in the case of virtual machine number is the most loadavg's
Number is the most, thus can be using loadavg parameter as the standard weighing main frame IO performance.
Run represents the length of instantaneous operation queue or waits the maximum of the process number run, this parameter and loadavg parameter
Value is the same can be 2, is used equally to weigh the situation of cpu load, and the place that this run parameter is different from loadavg is, run joins
What number reflected is the situation of change of this physical host momentary load, and that loadavg then reflects is past 1min, 5min, 15min
Deng the average load situation in a period of time, therefore this value has instantaneous reference meaning, referring to Fig. 6, the feelings that virtual machine number is the most
The number (loading) under condition run the most is the most, thus can be using run parameter as the standard weighing main frame IO performance.
Avgqu represents average I/O queue, IOQ length, and queue length also can work as I/O as the index of measurement system I/O load
When queue length becomes larger, illustrating that this system exists I/O bottleneck, general maximum selection rule is 1.Referring to Fig. 3, virtual machine number
The number (loading) in the case of the most avgqu the most is the most, thus can be using avgqu parameter as weighing main frame IO performance
One standard.
Util represents the percentage of time in a second for I/O operation, i.e. one second has percent how many time for
I/O operation, has in one second in other words how long I/O queue, IOQ is non-NULL, and maximum is 100%, and this parameter can reflect disk
Busy degree, referring to Fig. 4, the number (loading) in the case of virtual machine number is the most util the most is the most, thus util ginseng
Number can be used to weigh a standard of main frame I/O load.
Contexts represents that context handover overhead, too much context switching can cause the expense that system is the biggest, for
Applying in I/O intensive type, the timeslice that CPU processes needed for application runs is the shortest, needs a lot of contexts to switch, at context
Maximum in switching is generally 15000 times, referring to Fig. 5, (loads) contexts in the case of virtual machine number is the most the most
Number the most, can using this contexts parameter as weigh I/O load a standard.
Individually parameter can not reflect the behavior pattern of system I/O load and physical host exactly, and the present embodiment is combined
Close the most representational I/O load parameter considering to reflect physics host computer system I/O load situation, be specifically based on I/O intensive type
The virtual machine (vm) migration threshold determination model of application is as shown in table 1 below:
Table 1
In the optimal threshold interval of table 1, each load is exceeded the percentage ratio loaded most and sets by five-tuple threshold determination model T
Value when reducing for efficiency, is set to value when resource utilization reduces by each load less than the percentage ratio of maximum load, sets each
The interval loading the percentage ratio less than maximum load, loading between the percentage ratio exceeding maximum load of load is born as embodying IO
The optimal threshold of each load carrying parameter is interval, thus determines the threshold of 5 virtual machine numbers of correspondence according to optimal threshold interval
Value interval.
It is also preferred that the left set the load of each load percentage ratio less than maximum load as 50%, load exceed maximum load
Percentage ratio is 70%, and optimal threshold interval is 50%~70%.Because either CPU or I/O resource, when itself CPU or IO bears
Carry and exceed the 70% of maximum and i.e. think that efficiency reduces, and when load is less than the 50% of maximum load, i.e. think this utilization of resources
Rate is low.Therefore can be interval by the optimal threshold of five load parameters in above-mentioned model, draw corresponding five void
The threshold interval of plan machine number, the relation of load and virtual machine number can be referring to aforementioned formula (1)~(6).
Consider several parameter, analyze the situation of change of physical host I/O resource utilization rate, by 5 virtual machine numbers
The common factor of threshold interval is as described threshold interval [M, N] so that between minimizing 50%-70%, threshold value is error, draws
Whole optimum interval [M, N] (the most now EIO=1), shown in equation below (9), same procedure can operate with EIO=-1 and EIO=0 institute
Corresponding virtual machine threshold zone,
[M,N]=[Mloadavg, Nloadavg]∩[Mrun, Nrun]∩[Mavgqu, Navgqu]∩[Mutil, Nutil]∩[Mcontext,
Ncontext] (9)
kIORepresent the number of the virtual machine running the application of I/O intensive type.Work as kIODuring≤N, then EIOWithin zone of reasonableness;And
Work as kIODuring > N, then EIOEfficiency then the most drastically declines, and does not meets the requirement of user.By imitating that I/O intensive type is applied
The maximum critical value N of virtual machine number in rate zone of reasonableness, it can be determined that EIOWhether within rational efficiency range, (load is closed
Reason).If it was not then the application to I/O intensive type carries out load migration process.It is divided into again in the range of rational efficiency
Bigization utilizes scope of resource and the most fully utilized scope of resource, and the threshold value of virtual machine when maximally utilizing resource is set as M, when
kIO, i.e. there is resources idle in < M, then it is assumed that the most fully utilized resource of this physical machine.
In step sl, threshold interval [M, N] can according to the change of I/O load and I/O intensive type application process time
Between set.
In step s 2, according to the threshold interval [M, N] obtained in step S1, set every physical host and run I/O intensive type
Efficiency E of applicationIO, thus it is defined on the application treatment effeciency in load change procedure, wherein,
kIORepresent the virtual machine number of actual motion I/O intensive type application.
In step s3, preset 4 collection table T1-T4, the physical host under the corresponding efficiency of labelling, initialize set
Table T1-T4, wherein, collection table T1 is EIOThe physical host set of=0;Collection table T1 is EIOThe physical host set of=1;Collection
Conjunction table T1 is EIOThe physical host set of=-1;Collection table T1 is the physical host set do not opened.
In step s 4, persistently waiting the application of I/O intensive type, if occurring, I/O intensive type is applied, and system can produce accordingly, then
Step S5 is entered, it is achieved the physical machine of this I/O intensive type application is distributed and processes while waiting the application of new I/O intensive type.
In step s 5, after the application of I/O intensive type occurs, it is judged that whether collection table T1 is empty, if sky, then enters step
Rapid S6;If not being empty, then that is opened by this I/O intensive type application physical host distributed in described collection table T1 is virtual
Machine runs, the virtual machine that the physical host that the application of this I/O intensive type sets up in this T1 is opened is run, updates set
Table T1-T4, carries out load and judges, make efficiency E in load this physical hostIOMigrate out a certain amount of time not in preset range
Virtual machine, enter step S6.
In step s 6, it is judged that whether collection table T4 is empty, if sky, then refusal distribution physical host, waits pending
Described new I/O intensive type is applied, and returns step S5;If not being empty, then start a new physical host, and update collection table
T1-T4, and the virtual machine applying the physical host distributing to this new startup to be opened on this I/O intensive type run, to this physics
Main frame carries out load and judges, makes efficiency E in loadIOA certain amount of virtual machine is migrated out time not in preset range.
Referring to Fig. 2, in one embodiment, physical host is carried out load and judges, and make the most virtual when necessary
The mode that machine migrates comprises the following steps:
A1: physical host receives the application of I/O intensive type and runs on the virtual machine opened;
A2: judge EIOWhether >=0 set up;
A3: if setting up, entering step A4, otherwise migrating out a number of virtual machine, enter step A4;
A4: judge EIOWhether=1 set up;
A5: if being false, migrates out a number of virtual machine and returns to step A4, does not the most migrate virtual machine.
In another embodiment, in step S5 and S6, this physical host is carried out load and judges, work as EIOWill when=-1
Virtual machine is moved out.
By experiment, show that through T model the virtual machine of the I/O intensive type application run on a physical host is optimal
Number, guarantees physical server optimization process efficiency.Below the present invention with under cloud platform original scheduling strategy compare,
Verify feasibility and the high efficiency of the present invention.In CloudStack system platform, typical resource dispatching strategy is
FirstFit strategy.The present invention is contrasted with FirstFit strategy.In the figure 7, VM (virtual machine)=8, in fig. 8, VM
(virtual machine)=9, for the present invention is relative to FirstFit strategy, the process total time of physical host (physical server) and application
Process average time the most less.
The present invention passes through to migrate virtual machine in good time, the quantity of the I/O intensive type application on conservative control physical server, carries
Under high cloud environment, physical host processes the efficiency of I/O intensive type application, and reduces energy consumption to a certain extent.Meanwhile, by
Great many of experiments analysis under true cloud environment and Experimental comparison, the checking present invention can make the treatment effeciency one that I/O intensive type is applied
Directly it is maintained at optimal, within the I/O load of physical server also maintains essentially in normal range, it is to avoid when I/O intensive type is applied
When increasing, the phenomenon that I/O intensive type application treatment effeciency reduces and physical server I/O load is overweight, below nowadays cloud environment
Carry out specific resource dispatching strategy to dissimilar application and there is good realistic meaning.
Although the present invention is open as above with preferred embodiment, but it is not for limiting claim, any this area
Technical staff without departing from the spirit and scope of the present invention, can make possible variation and amendment, therefore the present invention
Protection domain should be defined in the range of standard with the claims in the present invention.
Claims (8)
1. the dispatching method of virtual machine of I/O intensive type application under a cloud platform, it is characterised in that comprise the following steps:
S1: set threshold interval [M, N], wherein N be the virtual machine number in efficiency zone of reasonableness of I/O intensive type application
Big marginal value, the marginal value of virtual machine number when M is to maximally utilize resource;
S2: set every physical host and run efficiency E of I/O intensive type applicationIO, wherein,
kIORepresent the virtual machine number of actual motion I/O intensive type application;
S3: preset 4 collection table T1-T4, initializes collection table T1-T4, and wherein, collection table T1 is EIOThe physical host collection of=0
Close;Collection table T1 is EIOThe physical host set of=1;Collection table T1 is EIOThe physical host set of=-1;Collection table T1 is not
The physical host set opened;
S4: persistently waiting the application of I/O intensive type, I/O intensive type is applied if occurring, then enter while waiting the application of new I/O intensive type
Enter step S5;
S5: judge whether collection table T1 is empty, if sky, then enters step S6;If not being empty, then should by this I/O intensive type
The virtual machine opened with the physical host distributed in described collection table T1 runs, updates collection table T1-T4, to this physics
Main frame carries out load and judges, makes efficiency E in loadIOMigrate out a certain amount of virtual machine time not in preset range, enter step
Rapid S6;
S6: judge whether collection table T4 is empty, if sky, then refusal distribution physical host, waits pending described new I/O intensive
Type is applied, and returns step S5;If not being empty, then start a new physical host, and update collection table T1-T4, and should
I/O intensive type is applied in the virtual machine that the physical host distributing to this new startup is opened and is run, and loads this physical host
Judge, make efficiency E in loadIOA certain amount of virtual machine is migrated out time not in preset range.
2. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 1, it is characterised in that in step
In S1, define a five-tuple threshold determination model T, comprehensively weigh I/O load, so that it is determined that physical host can to bear IO close
The threshold interval of the virtual machine number of collection type application,
T={loadavg, run, avgqu, util, contexts},
Loadavg represents that CPU is currently running the statistical information of the process number run with waiting for CPU within a period of time, and run represents
The process number that the length of instantaneous operation queue or wait run, avgqu represents average I/O queue, IOQ length, and util represented in one second
For the percentage of time of I/O operation, contexts represents context handover overhead;
Each load is exceeded the percentage ratio loaded most and is set to value when efficiency reduces by described five-tuple threshold determination model T, will be each
Load is set to value when resource utilization reduces less than the percentage ratio of maximum load, sets the load of each load less than maximum load
Percentage ratio, load exceed the optimum threshold as each load embodying I/O load parameter of the interval between the percentage ratio of maximum load
Value interval, thus the threshold interval of 5 virtual machine numbers of correspondence is determined according to optimal threshold interval;
Using the common factor of the threshold interval of 5 virtual machine numbers as described threshold interval [M, N],
[M, N]=[Mloadavg, Nloadavg]∩[Mrun, Nrun]∩[Mavgqu, Navgqu]∩[Mutil, Nutil]∩[Mcontext,
Ncontext]。
3. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 2, it is characterised in that set each
The load of load less than the percentage ratio of maximum load be 50%, to exceed the percentage ratio of maximum load be 70% in load, optimal threshold
Interval is 50%~70%.
4. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 1, it is characterised in that described reality
Virtual machine number k of I/O intensive type application is run on borderIOHaving a restriction scope, the determination mode of described restriction scope includes:
Definition physical host resource type, a given physical host, then this physical host has multiple different physical resource,
Use RnRepresenting, n represents the numbering of the different resource type of physical host, is expressed as with resource collection R:
R={R1, R2...Rn...RNN ∈ 1,2 ..., N} (1);
Definition type of virtual machine, virtual machine then also can be divided into different type of virtual machine V according to the difference of application resource requirementi
Each type of virtual machine ViRequired every resource is different, rI, nRepresent ViThe resource characteristic that type Virtual machine consumes, with physics
The resource type of main frame is corresponding,
V={V1, V2...Vi...VmI ∈ 1,2 ..., m} (2)
rI, n={ rI, 1, rI, 2...rI, n...rI, N} n∈{1,2...N} (3);
Definition deploying virtual machine number restrictive condition, a physical host can only be disposed a number of virtual within the same time
Machine ki, i ∈ 1,2 ..., N}, kiRepresent ViThe virtual machine number of type Virtual machine, therefore deploying virtual machine number restrictive condition is used
Equation below is indicated:
Σi(ki*Vi)≤αiRi, 0 < α < 1 (4)
Substitution, derivation according to definition type of virtual machine then have
Wherein, αiIt is then correlation coefficient, is used for determining the utilization rate limit of related resource;
Definition only considers the I/O type virtual machine number restrictive condition under I/O resource consumption,
rIO, nRepresenting the resource characteristic of I/O type virtual machine, n difference represents the parameter difference that I/O resource consumes, kIORepresent reality fortune
The virtual machine number of row I/O intensive type application;Represent the virtual machine IO money that physical host can be opened
Source;αIORIORepresent the I/O resource of physical host;
Virtual machine number k of actual motion I/O intensive type applicationIORestriction scope determine according to formula (4), (5), (6), in institute
State in the range of restriction then this physical host and can receive new I/O intensive type application always, otherwise, then cannot the most newly-built virtual machine, i.e.
On this physical host, virtual machine unlatching quantity reaches maximum.
5. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 4, it is characterised in that described IO
The maximum critical value N of virtual machine number in efficiency zone of reasonableness of intensive applications is the upper limit of described restriction scope.
6. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 1, it is characterised in that described
In step S1, the time that described threshold interval [M, N] processes according to change and the I/O intensive type application of I/O load sets.
7. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 1, it is characterised in that to physics
Main frame carries out loading judgement and comprises the following steps:
A1: physical host receives the application of I/O intensive type and runs on the virtual machine opened;
A2: judge EIOWhether >=0 set up;
A3: if setting up, entering step A4, otherwise migrating out a number of virtual machine, enter step A4;
A4: judge EIOWhether=1 set up;
A5: if being false, migrates out a number of virtual machine and returns to step A4, does not the most migrate virtual machine.
8. the dispatching method of virtual machine of I/O intensive type application under cloud platform as claimed in claim 1, it is characterised in that in step
In S5 and S6, this physical host is carried out load and judges, work as EIOWhen=-1, virtual machine is moved out.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610021270.XA CN105718310B (en) | 2016-01-13 | 2016-01-13 | The dispatching method of virtual machine that I/O intensive type is applied under a kind of cloud platform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610021270.XA CN105718310B (en) | 2016-01-13 | 2016-01-13 | The dispatching method of virtual machine that I/O intensive type is applied under a kind of cloud platform |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105718310A true CN105718310A (en) | 2016-06-29 |
CN105718310B CN105718310B (en) | 2018-09-18 |
Family
ID=56147860
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610021270.XA Expired - Fee Related CN105718310B (en) | 2016-01-13 | 2016-01-13 | The dispatching method of virtual machine that I/O intensive type is applied under a kind of cloud platform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105718310B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110928683A (en) * | 2019-11-14 | 2020-03-27 | 湖南工业大学 | Edge computing resource allocation method based on two types of intensive virtual machines |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070204265A1 (en) * | 2006-02-28 | 2007-08-30 | Microsoft Corporation | Migrating a virtual machine that owns a resource such as a hardware device |
CN103353853A (en) * | 2013-07-26 | 2013-10-16 | 浪潮电子信息产业股份有限公司 | Automatic load adjusting method |
CN103473139A (en) * | 2013-09-26 | 2013-12-25 | 四川中电启明星信息技术有限公司 | Virtual machine cluster resource allocation and scheduling method |
CN104102523A (en) * | 2013-04-03 | 2014-10-15 | 华为技术有限公司 | Method for migrating virtual machine and resource scheduling platform |
CN104375897A (en) * | 2014-10-27 | 2015-02-25 | 西安工程大学 | Cloud computing resource scheduling method based on minimum relative load imbalance degree |
-
2016
- 2016-01-13 CN CN201610021270.XA patent/CN105718310B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070204265A1 (en) * | 2006-02-28 | 2007-08-30 | Microsoft Corporation | Migrating a virtual machine that owns a resource such as a hardware device |
CN104102523A (en) * | 2013-04-03 | 2014-10-15 | 华为技术有限公司 | Method for migrating virtual machine and resource scheduling platform |
CN103353853A (en) * | 2013-07-26 | 2013-10-16 | 浪潮电子信息产业股份有限公司 | Automatic load adjusting method |
CN103473139A (en) * | 2013-09-26 | 2013-12-25 | 四川中电启明星信息技术有限公司 | Virtual machine cluster resource allocation and scheduling method |
CN104375897A (en) * | 2014-10-27 | 2015-02-25 | 西安工程大学 | Cloud computing resource scheduling method based on minimum relative load imbalance degree |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110928683A (en) * | 2019-11-14 | 2020-03-27 | 湖南工业大学 | Edge computing resource allocation method based on two types of intensive virtual machines |
CN110928683B (en) * | 2019-11-14 | 2022-01-28 | 湖南工业大学 | Edge computing resource allocation method based on two types of intensive virtual machines |
Also Published As
Publication number | Publication date |
---|---|
CN105718310B (en) | 2018-09-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Zhang et al. | Energy-aware virtual machine allocation for cloud with resource reservation | |
CN100511151C (en) | Multiple-path multiple-core server and CPU virtualization processing method thereof | |
Kc et al. | Scheduling hadoop jobs to meet deadlines | |
US9405585B2 (en) | Management of heterogeneous workloads | |
Polo et al. | Deadline-based MapReduce workload management | |
Lin et al. | Energy-efficient task scheduling for multi-core platforms with per-core DVFS | |
CN103701886A (en) | Hierarchic scheduling method for service and resources in cloud computation environment | |
CN105912406A (en) | Low-energy independent task scheduling and resource configuration method | |
CN104023042B (en) | Cloud platform resource scheduling method | |
Sharma et al. | An improved task allocation strategy in cloud using modified k-means clustering technique | |
Gandomi et al. | HybSMRP: a hybrid scheduling algorithm in Hadoop MapReduce framework | |
CN104793993A (en) | Cloud computing task scheduling method of artificial bee colony particle swarm algorithm based on Levy flight | |
CN112181620A (en) | Big data workflow scheduling method for sensing service capability of virtual machine in cloud environment | |
Jiang et al. | An energy-aware virtual machine migration strategy based on three-way decisions | |
CN105912383A (en) | High-reliability dependent task scheduling and resource configuration method | |
Emara et al. | Genetic-Based Multi-objective Task Scheduling Algorithm in Cloud Computing Environment. | |
Joshi et al. | Evaluating the performance of load balancing algorithm for heterogeneous cloudlets using HDDB algorithm | |
Biswas et al. | A novel resource aware scheduling with multi-criteria for heterogeneous computing systems | |
Gouasmi et al. | Exact and heuristic MapReduce scheduling algorithms for cloud federation | |
CN107589980A (en) | A kind of dispatching method of cloud computing resources | |
CN105718310A (en) | Virtual machine scheduling method for IO dense application under cloud platform | |
Wang | A dynamic resource management in mobile agent by artificial neural network | |
Qu et al. | Improving the energy efficiency and performance of data-intensive workflows in virtualized clouds | |
Filippini et al. | Hierarchical Scheduling in on-demand GPU-as-a-Service Systems | |
Manekar et al. | Optimizing cost and maximizing profit for multi-cloud-based big data computing by deadline-aware optimize resource allocation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180918 Termination date: 20210113 |