CN104023068B - A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing - Google Patents
A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing Download PDFInfo
- Publication number
- CN104023068B CN104023068B CN201410263629.5A CN201410263629A CN104023068B CN 104023068 B CN104023068 B CN 104023068B CN 201410263629 A CN201410263629 A CN 201410263629A CN 104023068 B CN104023068 B CN 104023068B
- Authority
- CN
- China
- Prior art keywords
- host computer
- fictitious host
- application service
- load balancing
- service pond
- 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.)
- Active
Links
Landscapes
- Computer And Data Communications (AREA)
Abstract
The present invention provides a kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing, including:IAAS cloud computing environments are built, including:Load balancing gateway, Virtual Machine Manager center and multiple application service ponds;Load balancing gateway receives the resource request from each client, then according to default load balancing, resource request is distributed into each application service pond, while load balancing gateway is configured with logging modle, and following information is recorded by logging modle:The on-line session number of the fictitious host computer of activity, handle up flow and number of request per second;Traffic pressure model is established based on record information, if traffic pressure model reaches the critical triggering threshold values of certain state, actively sends dispatch command, flexible scheduling resources of virtual machine to Virtual Machine Manager center.Under significant increase cloud environment on the premise of the practicality and accuracy of computing resource, the burden and expense of whole system are also significantly reduced, improves the service quality of whole cloud computing platform.
Description
Technical field
The invention belongs to communication technical field, and in particular to Passive Mode elastic calculation resource is realized in a kind of load balancing
The method of scheduling.
Background technology
Under cloud computing environment, after computer hardware resource is virtualized, some virtual machines are obtained, more virtual machines group again
Into virtualization pool, calculating task is then carried out by virtualization pool.The virtual resource of virtualization pool how is made full use of, is carried
High cloud service platform parallel processing capability, it is the focus of current research.
In the prior art, the method being usually taken is:Set for every active virtual machine in each virtualization pool
Independent detection demons, by the cpu resource or memory source of the detection demons collection activity virtual machine, then count
The cpu resource or memory source collected, obtains the pressure state of the active virtual machine;Then by the pressure of each active virtual machine
Power feedback of status is divided to load-balancing device, the pressure state of each active virtual machine of the load-balancing device to receiving
Analysis, judges whether a certain virtualization pool overloads, and then decides whether into virtualization pool to increase or reduce virtually
Machine, realize the load balancing of each virtualization pool.
Although the above method can be by constantly monitoring the pressure state of each virtual machine, and then adjusts virtual resource allocation feelings
Condition, the utilization rate of virtual resource is improved on certain procedures, still, due to needing the current money of continuous collection activity virtual machine
Source situation, the burden of whole system is aggravated;In addition, independent detection demons are set for each active virtual machine,
Increase the expense of whole system.
The content of the invention
The defects of existing for prior art, the present invention, which provides, realizes that Passive Mode elastic calculation provides in a kind of load balancing
The method of source scheduling, to overcome above mentioned problem.
The technical solution adopted by the present invention is as follows:
The present invention provides a kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing, including following step
Suddenly:
S1, IAAS cloud computing environments are built, including:Load balancing gateway, Virtual Machine Manager center and multiple application services
Pond;By several member compositions, the member is fictitious host computer in each application service pond;Also, in each application service pond
The fictitious host computer that may participate in scheduling is configured with unified participation scheduling identification ID, and by participation scheduling identification ID, differentiation can be joined
With the fictitious host computer of scheduling and it is not involved in the fictitious host computer of scheduling;Fictitious host computer has two kinds of shapes in each application service pond
State:Active state and inactive state, the fictitious host computer of active state refer to the fictitious host computer being currently running;The void of inactive state
Intend main frame to further discriminate between as closed mode and gradually move back state, the fictitious host computer of closed mode refers to offline fictitious host computer, gradually
The fictitious host computer for moving back state refers to and no longer receives new client's request, continues with the client for being currently connected to itself request, etc.
All resources are discharged after being disposed and enter closed mode;
S2, the load balancing gateway configure following initial configuration parameters by configuration module:Elastic calculation triggering threshold values,
Active members IP that members IP that each application service pond is included, each application service pond are included, each application clothes
The minimum limit value of active members, startup or the control option for disabling elastic calculation function in business pond;
S3, when the control option of the startup or disabling elastic calculation function is arranged to starting state, the load
Balanced gateway receives the resource request from each client, please by the resource then according to default load balancing
Ask and distribute to each application service pond, while the load balancing gateway is configured with logging modle, by logging modle record with
Lower information:The on-line session number of movable fictitious host computer, handle up flow and number of request per second in each application service pond;
Meanwhile the load balancing gateway reads the configuration module, obtains the activity that each application service pond is included
Member IP, the active members quantity that each application service pond is included is obtained, for any one application service pond I, be performed both by
Operate below:
S31, the load balancing gateway is it is first determined whether need to increase a virtual master into the application service pond I
Machine, step are as follows:
S311, if fictitious host computer quantity movable under the I current states of application service pond is n;
S312, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S313, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Sch:The on-line session number threshold values upper limit that every fictitious host computer of expression standard can be supported;
Sbh:The flow threshold values upper limit that every fictitious host computer of expression standard can be supported;
Srh:The number of request threshold values upper limit per second that every fictitious host computer of expression standard can be supported;
S314, δ values are calculated by formula (one), wherein, δ indicates whether to increase the logical value of fictitious host computer, if δ values are 1,
S315 is then performed, otherwise, performs S316;
(1);
S315, the load balancing gateway is sent to the Virtual Machine Manager center increases by one into application service pond I
The dispatch command of fictitious host computer;
The Virtual Machine Manager center remote wakes up a fictitious host computer, and increases in the application service pond I, puts new
The state of increased fictitious host computer is active state, while updates the following parameter in the configuration module:Application service pond I institutes
Comprising the active members IP that is included of members IP, application service pond I;
S316, the load balancing gateway determine whether fictitious host computer number movable under the I current states of application service pond
Measure n and whether be more than the minimum limit value of active members in the application service pond I set by configuration module, if it is not greater, performing
S33;Otherwise, S32 is performed;
Secondly S32, the load balancing gateway judge whether to need to reduce a virtual master in the application service pond I
Machine, step are as follows:
S321, if fictitious host computer quantity movable under the I current states of application service pond is n;
S322, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S323, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Scl:The on-line session number lower threshold that every fictitious host computer of expression standard can be supported;
Sbl:The flow lower threshold that every fictitious host computer of expression standard can be supported;
Srl:The number of request lower threshold per second that every fictitious host computer of expression standard can be supported;
S324, γ values are calculated by formula (two), wherein, γ indicates whether to reduce the logical value of fictitious host computer, if δ values are
1, then S325 is performed, otherwise, performs S33;
(2)
S325, the load balancing gateway determine whether to be applied the specific void that current pressure is most light in service pool I
Intend main frame, the scheduling that specific fictitious host computer described in reduction application service pond I is then sent to the Virtual Machine Manager center refers to
Order;
The Virtual Machine Manager center remote closes the specific fictitious host computer, puts the state of the specific fictitious host computer as gradually
Move back state;When the specific fictitious host computer processing completes to be currently connected to the client's request of itself, then the specific void is closed
Intend main frame, meanwhile, the state for putting the specific fictitious host computer is closed mode, while updates the following ginseng in the configuration module
Number:The active members IP that members IP, the application service pond I that application service pond I is included are included;Then S33 is performed;
S33, the load balancing gateway terminate epicycle flexible scheduling, and when reaching next dispatching cycle, circulation performs S3
Step.
Preferably, in S325, the load balancing gateway current pressure in service pool I that determines whether to be applied is most light
Specific fictitious host computer, specifically include following steps:
S3251, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S3252, the initial configuration parameters set by the configuration module are read, obtain following parameter:
Δ α is the weight coefficient of on-line session number;
Δ β is flow weight coefficient;
Δ γ is that current request handles number weight coefficient;
S3253, Ri is calculated by formula (three), wherein, Ri is represented it needs to be determined that the fictitious host computer closed and discharged, as works as
The most light specific fictitious host computer of preceding pressure;
(3).
Preferably, in S33, the load balancing gateway terminates epicycle flexible scheduling, when reaching next dispatching cycle, follows
Ring performs S3 steps, is specially:
The configuration module is additionally provided with quiet period initial parameter value;
In epicycle scheduling process, after performing wake-up or closing the operation of a certain virtual machine, a quiet period is spaced, then
Perform next round scheduling.
The method that Passive Mode elastic calculation scheduling of resource is realized in load balancing provided by the invention, have following excellent
Point:The load balancing gateway of flexible computing capability is combined with virtualization technology, load balancing gateway is in distribution from visitor
During the service request at family end, directly record the on-line session number of fictitious host computer movable in each application service pond, handle up
Flow and number of request per second, and then judge whether to need to increase or decrease fictitious host computer into application service pond, in significant increase
Under cloud environment on the premise of the practicality and accuracy of computing resource, the burden and expense of whole system are also significantly reduced, is carried
The service quality of high whole cloud computing platform.
Brief description of the drawings
Fig. 1 is the method flow signal that Passive Mode elastic calculation scheduling of resource is realized in load balancing provided by the invention
Figure.
Embodiment
Below in conjunction with accompanying drawing, the present invention is described in detail:
As shown in figure 1, the present invention provides a kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing,
By load-balancing technique, subsequent use service pool perception and virtual platform in schedule virtual resources flexibility it is close
It is combined together, more complete elastic calculation solution is provided for cloud computing environment.
Specifically include following steps:
S1, IAAS cloud computing environments are built, including:Load balancing gateway, Virtual Machine Manager center and multiple application services
Pond;By several member compositions, the member is fictitious host computer in each application service pond;Also, in each application service pond
The fictitious host computer that may participate in scheduling is configured with unified participation scheduling identification ID, and by participation scheduling identification ID, differentiation can be joined
With the fictitious host computer of scheduling and it is not involved in the fictitious host computer of scheduling;Fictitious host computer has two kinds of shapes in each application service pond
State:Active state and inactive state, the fictitious host computer of active state refer to the fictitious host computer being currently running;The void of inactive state
Intend main frame to further discriminate between as closed mode and gradually move back state, the fictitious host computer of closed mode refers to offline fictitious host computer, gradually
The fictitious host computer for moving back state refers to and no longer receives new client's request, continues with the client for being currently connected to itself request, etc.
All resources are discharged after being disposed and enter closed mode;In the present invention, fictitious host computer is set gradually to move back state, energy in the presence of one
Enough ensure that the influence lasting to client traffic minimizes.
The fictitious host computer that flexible resource scheduling is may participate in application service pond sets special identifier, is to load for convenience
Gateway weigh when carrying out flexible scheduling, the fictitious host computer that can allow flexible scheduling can be identified.
S2, the load balancing gateway configure following initial configuration parameters by configuration module:Elastic calculation triggering threshold values,
Active members IP that members IP that each application service pond is included, each application service pond are included, each application clothes
The minimum limit value of active members, startup or the control option for disabling elastic calculation function in business pond;
The members and active members included by application service pond are dynamic change, therefore, work as members
, it is necessary to dynamically update the two parameters in real time when being changed with active members, facilitate follow-up load balancing gateway according to this two
Individual parameter carries out flexible scheduling.
For during ensuring the business free time, application service pond state is normal and ensures that service is not interrupted, it is necessary to set using clothes
The minimum limit value for pond active members of being engaged in, will not be further according to elastometer after when application service pond, active members quantity reaches minimum value
The lowest limit threshold parameter of calculation continues to close fictitious host computer.Certainly, if the active members quantity for being not involved in scheduling meets application clothes
Be engaged in the minimum limit value of pond active members, then elastic calculation engine can close it is all participate in scheduling fictitious host computers.
S3, when the control option of the startup or disabling elastic calculation function is arranged to starting state, the load
Balanced gateway receives the resource request from each client, please by the resource then according to default load balancing
Ask and distribute to each application service pond, while the load balancing gateway is configured with logging modle, by logging modle record with
Lower information:The on-line session number of movable fictitious host computer, handle up flow and number of request per second in each application service pond;
The present invention it is most important innovation be:Load-balancing device when distributing application traffic, i.e., record in detail this three
The individual important parameter for reflecting its pressure condition:The on-line session number of movable fictitious host computer in application service pond
(Concurrent Connection), flow of handling up (Through Put) and number of request per second (Request per
Second);Then the parameter of these reflection pressure conditions is linked together with the automatic start-stop of virtual machine, realizes virtual resource
Flexible scheduling, load-balancing device overhead need not be increased completely;In addition, also without the virtual master in application service pond
Special system resource monitoring demons are installed, with specially disposing demons in fictitious host computer in the prior art in machine
Pattern is compared, and the present invention has more preferable IaaS system compatibilities.
Meanwhile the load balancing gateway reads the configuration module, obtains the activity that each application service pond is included
Member IP, the active members quantity that each application service pond is included is obtained, for any one application service pond I, setting increases
Add the fictitious host computer priority in application service pond higher than closing and discharge the default convention of the fictitious host computer in an application service pond.
That is, it is first determined whether need to increase fictitious host computer, further judge whether to need if NO to reduce by a void
Intend main frame, it is specific to perform following operate:
S31, the load balancing gateway is it is first determined whether need to increase a virtual master into the application service pond I
Machine, step are as follows:
S311, if fictitious host computer quantity movable under the I current states of application service pond is n;
S312, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S313, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Sch:The on-line session number threshold values upper limit that every fictitious host computer of expression standard can be supported;
Sbh:The flow threshold values upper limit that every fictitious host computer of expression standard can be supported;
Srh:The number of request threshold values upper limit per second that every fictitious host computer of expression standard can be supported;
S314, δ values are calculated by formula (one), wherein, δ indicates whether to increase the logical value of fictitious host computer, if δ values are 1,
S315 is then performed, otherwise, performs S316;
(1);
S315, the load balancing gateway is sent to the Virtual Machine Manager center increases by one into application service pond I
The dispatch command of fictitious host computer;It is proof load equilibrium gateway and Virtual Machine Manager center normal communication in practical application, can be pre-
Following communications parameter first is set in load balancing gateway:Virtual Machine Manager center IP address, virtual machine store path, virtual machine pipe
Keeper's account at reason center, i.e. usemame/password.
The Virtual Machine Manager center remote wakes up a fictitious host computer, and increases in the application service pond I, puts new
The state of increased fictitious host computer is active state, while updates the following parameter in the configuration module:Application service pond I institutes
Comprising the active members IP that is included of members IP, application service pond I;
S316, the load balancing gateway determine whether fictitious host computer number movable under the I current states of application service pond
Measure n and whether be more than the minimum limit value of active members in the application service pond I set by configuration module, if it is not greater, performing
S33;Otherwise, S32 is performed;
Secondly S32, the load balancing gateway judge whether to need to reduce a virtual master in the application service pond I
Machine, step are as follows:
S321, if fictitious host computer quantity movable under the I current states of application service pond is n;
S322, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S323, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Scl:The on-line session number lower threshold that every fictitious host computer of expression standard can be supported;
Sbl:The flow lower threshold that every fictitious host computer of expression standard can be supported;
Srl:The number of request lower threshold per second that every fictitious host computer of expression standard can be supported;
S324, γ values are calculated by formula (two), wherein, γ indicates whether to reduce the logical value of fictitious host computer, if δ values are
1, then S325 is performed, otherwise, performs S33;
(2)
S325, the load balancing gateway determine whether to be applied the specific void that current pressure is most light in service pool I
Intend main frame, the scheduling that specific fictitious host computer described in reduction application service pond I is then sent to the Virtual Machine Manager center refers to
Order;
The Virtual Machine Manager center remote closes the specific fictitious host computer, puts the state of the specific fictitious host computer as gradually
Move back state;When the specific fictitious host computer processing completes to be currently connected to the client's request of itself, then the specific void is closed
Intend main frame, meanwhile, the state for putting the specific fictitious host computer is closed mode, while updates the following ginseng in the configuration module
Number:The active members IP that members IP, the application service pond I that application service pond I is included are included;Then S33 is performed;
In this step, the load balancing gateway current pressure in service pool I that can judge to be applied by the following method is most light
Specific fictitious host computer:
S3251, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S3252, the initial configuration parameters set by the configuration module are read, obtain following parameter:
Δ α is the weight coefficient of on-line session number;
Δ β is flow weight coefficient;
Δ γ is that current request handles number weight coefficient;
S3253, Ri is calculated by formula (three), wherein, Ri is represented it needs to be determined that the fictitious host computer closed and discharged, as works as
The most light specific fictitious host computer of preceding pressure;
(3).
S33, the load balancing gateway terminate epicycle flexible scheduling, and when reaching next dispatching cycle, circulation performs S3
Step.
Configuration module is additionally provided with quiet period initial parameter value;In epicycle scheduling process, when execution wakes up or closes certain
After the operation of one virtual machine, a quiet period is spaced, then performs next round scheduling.
Due to the dynamic characteristic of application system, to prevent threshold parameter from shaking extremely, cause fictitious host computer unlatching/closed
In frequent, therefore, in the present invention, configuration module is additionally provided with quiet period initial parameter value;In epicycle scheduling process, work as execution
After waking up or closing the operation of a certain virtual machine, a quiet period is spaced, then performs next round scheduling.
Silent period parameters can be changed, and system default quiet period configuration parameter can be as described below:
(a) quiet period policy depiction:A data are recorded every 3 seconds logging modles, the record of a decision-making period continues
For 3 minutes (60 times), threshold values data of the average of threshold values record in this cycle as scheduling of resource are calculated;It is empty increasing or closing
After intending Host behavior generation, after system 2 record periods of silence, continue record and carry out threshold values judgement.
(b) strategy of fictitious host computer is closed:To ensure the integrality of applied business, it is necessary to when closing fictitious host computer, first will
Fictitious host computer is identified as and gradually moves back state, and now the fictitious host computer is not as animated virtual main frame.Only when the fictitious host computer is without work
During dynamic connection, the fictitious host computer just can be formally closed.
The method that Passive Mode elastic calculation scheduling of resource is realized in load balancing provided by the invention, load balancing gateway
The situation of change of each fictitious host computer flow load is recorded in real time, and application service pond elasticity is used as by artificial default triggering threshold values
The trigger of scheduling, linkage Virtual Machine Manager center, the member for remotely arousing or closing subsequent use service pool, has following spy
Point:
(1) load balancing gateway is linked by API and vCenter or XenCenter Virtual Machine Managers center, is realized virtual
Main frame is turned on and off;
(2) by being linked with Virtual Machine Manager center, load balancing gateway can understand fictitious host computer deployment scenario automatically, from
Member of the dynamic selection fictitious host computer as application service pond;
(3) load balancing gateway is that application service pond creates the configuration that elastic calculation triggers threshold parameter, passes through parameter group
Close structure traffic pressure model, i.e. formula (one) or formula (two);After traffic pressure model enables, if reaching certain state
Critical triggering threshold values, then load balancing gateway actively to Virtual Machine Manager center send dispatch command, flexible scheduling virtual machine
Resource;
(4) load balancing gateway can monitor member's running status in each application service pond, including number of nodes in real time
The health status of (Pool member) and node (applies health examination), and foundation is provided for flexible scheduling;
(5) by configuring elastic calculation ability in load balancing gateway, it is possible to achieve the meter in IAAS cloud computing environments
Resource automatic governing is calculated, so as to optimize the utilization rate of computing resource, has the function that the cloud computing IAAS platforms of efficient energy-saving.
The method that Passive Mode elastic calculation scheduling of resource is realized in load balancing provided by the invention, in multiple virtual machines
Under the IAAS platform environments of realization, multiple application service ponds are disposed, can by configuring elastic calculation function in load balancing gateway
Easily to discharge computing resource (i.e. virtual machine) superfluous in some application service pond, and dynamic adaptation increases meter to needs
Calculate in the application service pond belonging to the application system of resource, so as to maximally utilize computing resource, realize intensive IAAS clouds
Computing environment.The burden and expense of whole system are also significantly reduced, improves the service quality of whole cloud computing platform.
Described above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
For member, under the premise without departing from the principles of the invention, some improvements and modifications can also be made, these improvements and modifications also should
Depending on protection scope of the present invention.
Claims (1)
1. the method for Passive Mode elastic calculation scheduling of resource is realized in a kind of load balancing, it is characterised in that by load balancing
Technology, subsequent use service pool perception and virtual platform in the flexibilities of schedule virtual resources combine closely one
Rise, more complete elastic calculation solution is provided for cloud computing environment;Comprise the following steps:
S1, IAAS cloud computing environments are built, including:Load balancing gateway, Virtual Machine Manager center and multiple application service ponds;Often
By several member compositions, the member is fictitious host computer in one application service pond;Also, participate in adjusting in each application service pond
The fictitious host computer of degree is configured with unified participation scheduling identification ID, by participation scheduling identification ID, distinguishes the void for participating in scheduling
Intend main frame and be not involved in the fictitious host computer of scheduling;Fictitious host computer has two states in each application service pond:Active state
And inactive state, the fictitious host computer of active state refer to the fictitious host computer being currently running;The fictitious host computer of inactive state enters one
Step divides into closed mode and gradually moves back state, and the fictitious host computer of closed mode refers to offline fictitious host computer, gradually moves back the void of state
Plan main frame, which refers to, no longer to be received new client's request, continues with the client for being currently connected to itself request, after waiting processing
Discharge all resources and enter closed mode;
S2, the load balancing gateway configure following initial configuration parameters by configuration module:Elastic calculation triggering threshold values, each
Active members IP that members IP that application service pond is included, each application service pond are included, each application service pond
Minimum limit value, startup or the control option for disabling elastic calculation function of middle active members;
S3, when the control option of the startup or disabling elastic calculation function is arranged to starting state, the load balancing
Gateway receives the resource request from each client, then according to default load balancing, by the resource request point
The each application service pond of dispensing, while the load balancing gateway is configured with logging modle, and following letter is recorded by logging modle
Breath:The on-line session number of movable fictitious host computer, handle up flow and number of request per second in each application service pond;
Meanwhile the load balancing gateway reads the configuration module, obtains the active members that each application service pond is included
IP, the active members quantity that each application service pond is included is obtained, for any one application service pond I, be performed both by following
Operation:
S31, the load balancing gateway it is first determined whether need into the application service pond I increase a fictitious host computer,
Step is as follows:
S311, if fictitious host computer quantity movable under the I current states of application service pond is n;
S312, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S313, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Sch:The on-line session number threshold values upper limit that every fictitious host computer of expression standard is supported;
Sbh:The flow threshold values upper limit that every fictitious host computer of expression standard is supported;
Srh:The number of request threshold values upper limit per second that every fictitious host computer of expression standard is supported;
S314, δ values are calculated by formula (one), wherein, δ indicates whether to increase the logical value of fictitious host computer, if δ values are 1, holds
Row S315, otherwise, perform S316;
S315, the load balancing gateway is sent to the Virtual Machine Manager center increases by one into application service pond I virtually
The dispatch command of main frame;
The Virtual Machine Manager center remote wakes up a fictitious host computer, and increases in the application service pond I, puts and newly increases
The state of fictitious host computer be active state, while update the following parameter in the configuration module:Application service pond I is included
The active members IP that is included of members IP, application service pond I;
S316, the load balancing gateway determine whether that fictitious host computer quantity n movable under the I current states of application service pond is
The minimum limit value of active members in the no application service pond I more than set by configuration module, if it is not greater, performing S33;It is no
Then, S32 is performed;
Secondly S32, the load balancing gateway judge whether to need to reduce a fictitious host computer in the application service pond I, walked
It is rapid as follows:
S321, if fictitious host computer quantity movable under the I current states of application service pond is n;
S322, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S323, the elastic calculation triggering threshold values in the configuration module is read, specifically obtains following parameter:
Scl:The on-line session number lower threshold that every fictitious host computer of expression standard is supported;
Sbl:The flow lower threshold that every fictitious host computer of expression standard is supported;
Srl:The number of request lower threshold per second that every fictitious host computer of expression standard is supported;
S324, γ values are calculated by formula (two), wherein, γ indicates whether to reduce the logical value of fictitious host computer, if γ values are 1,
S325 is then performed, otherwise, performs S33;
S325, the load balancing gateway determine whether to be applied most light specific virtual main of current pressure in service pool I
Machine, the dispatch command for reducing specific fictitious host computer described in application service pond I is then sent to the Virtual Machine Manager center;
The Virtual Machine Manager center remote closes the specific fictitious host computer, and the state for putting the specific fictitious host computer is gradually to move back shape
State;When the specific fictitious host computer processing completes to be currently connected to the client's request of itself, then close described specific virtual main
Machine, meanwhile, the state for putting the specific fictitious host computer is closed mode, while updates the following parameter in the configuration module:
The active members IP that members IP, the application service pond I that application service pond I is included are included;Then S33 is performed;
S33, the load balancing gateway terminate epicycle flexible scheduling, and when reaching next dispatching cycle, circulation performs S3 steps;
Wherein, in S325, the load balancing gateway determines whether to be applied most light specific of current pressure in service pool I
Fictitious host computer, specifically include following steps:
S3251, the logging modle is read, obtain following parameter:
Ci:Represent fictitious host computer i movable in application service pond I on-line session number;
Bi:Represent fictitious host computer i movable in application service pond I flow of handling up;
Ri:Represent fictitious host computer i movable in application service pond I number of request per second;
S3252, the initial configuration parameters set by the configuration module are read, obtain following parameter:
Δ α is the weight coefficient of on-line session number;
Δ β is flow weight coefficient;
Δ γ is that current request handles number weight coefficient;
S3253, Ri is calculated by formula (three), wherein, Ri is represented it needs to be determined that the fictitious host computer closed and discharged, is current pressure
The most light specific fictitious host computer of power;
Ri=min i | Ci × Δ α+Bi × Δ β+Ri × Δ γ } (i=1,2 ..., n) (three);
Wherein, in S33, the load balancing gateway terminates epicycle flexible scheduling, and when reaching next dispatching cycle, circulation performs
S3 steps, it is specially:
The configuration module is additionally provided with quiet period initial parameter value;
In epicycle scheduling process, after performing wake-up or closing the operation of a certain virtual machine, a quiet period is spaced, then perform
Next round is dispatched;
System default quiet period configuration parameter is:
(a) quiet period policy depiction:A data are recorded every 3 seconds logging modles, the record of a decision-making period is continuously 3 points
Clock, calculate threshold values data of the average of threshold values record in this cycle as scheduling of resource;Increasing or closing fictitious host computer behavior
After generation, after system 2 record periods of silence, continue record and carry out threshold values judgement;
(b) strategy of fictitious host computer is closed:To ensure the integrality of applied business, it is necessary to when closing fictitious host computer, first incite somebody to action virtual
Host identification is gradually moves back state, and now the fictitious host computer is not as animated virtual main frame;Only when the fictitious host computer non-activity connects
When connecing, the fictitious host computer just can be formally closed;
The method that Passive Mode elastic calculation scheduling of resource is realized in load balancing provided by the invention, load balancing gateway are real-time
The situation of change of each fictitious host computer flow load is recorded, application service pond flexible scheduling is used as by artificial default triggering threshold values
Trigger, linkage Virtual Machine Manager center, the member for remotely arousing or closing subsequent use service pool, specifically include:
(1) load balancing gateway is linked by API and vCenter or XenCenter Virtual Machine Managers center, realizes fictitious host computer
Be turned on and off;
(2) by being linked with Virtual Machine Manager center, load balancing gateway understands fictitious host computer deployment scenario, automatically selected automatically
Member of the fictitious host computer as application service pond;
(3) load balancing gateway is that application service pond creates the configuration that elastic calculation triggers threshold parameter, passes through parameter combination structure
Build traffic pressure model, i.e. formula (one) or formula (two);After traffic pressure model enables, if reaching facing for certain state
Boundary triggers threshold values, then load balancing gateway actively sends dispatch command, flexible scheduling resources of virtual machine to Virtual Machine Manager center;
(4) load balancing gateway monitors member's running status in each application service pond in real time, including number of members and member
Health status, foundation is provided for flexible scheduling;
(5) by configuring elastic calculation ability in load balancing gateway, realize computing resource in IAAS cloud computing environments from
Dynamic allotment, so as to optimize the utilization rate of computing resource, has the function that the cloud computing IAAS platforms of efficient energy-saving.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410263629.5A CN104023068B (en) | 2014-06-13 | 2014-06-13 | A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410263629.5A CN104023068B (en) | 2014-06-13 | 2014-06-13 | A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104023068A CN104023068A (en) | 2014-09-03 |
CN104023068B true CN104023068B (en) | 2017-12-15 |
Family
ID=51439637
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410263629.5A Active CN104023068B (en) | 2014-06-13 | 2014-06-13 | A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104023068B (en) |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105208054A (en) * | 2014-06-16 | 2015-12-30 | 中兴通讯股份有限公司 | Cloud scheduling management method, management center and management system |
CN106254408A (en) * | 2015-06-12 | 2016-12-21 | 财团法人工业技术研究院 | Control method, network system and service platform for mobile edge calculation |
CN105450728B (en) * | 2015-11-04 | 2017-03-08 | 合一网络技术(北京)有限公司 | A kind of method and device realizing the scheduling of virtual machine real equipment |
CN106845746A (en) * | 2016-06-15 | 2017-06-13 | 曹大海 | A kind of cloud Workflow Management System for supporting extensive example intensive applications |
CN111357257B (en) * | 2017-08-31 | 2022-06-14 | 格林伊登美国控股有限责任公司 | System and method for load balancing media server instances |
CN107645731B (en) * | 2017-09-21 | 2019-10-08 | 北京邮电大学 | Load-balancing method based on self-organizing resource allocation in a kind of NOMA system |
CN108170520A (en) * | 2018-01-29 | 2018-06-15 | 北京搜狐新媒体信息技术有限公司 | A kind of cloud computing resources management method and device |
CN108769271A (en) | 2018-08-20 | 2018-11-06 | 北京百度网讯科技有限公司 | Method, apparatus, storage medium and the terminal device of load balancing |
CN110166435B (en) * | 2019-04-18 | 2021-06-22 | 杭州电子科技大学 | Mimicry Web gateway system and method for dynamic scheduling by adopting load balancing |
CN114598656B (en) * | 2022-05-09 | 2022-07-29 | 张家港保税科技集团电子商务有限公司 | Load balancing method and system applied to delivery service platform |
CN117076057B (en) * | 2023-10-18 | 2024-01-26 | 成都阿加犀智能科技有限公司 | AI service request scheduling method, device, equipment and medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719081A (en) * | 2009-12-01 | 2010-06-02 | 北京大学 | Method for scheduling virtual machines |
CN102244685A (en) * | 2011-08-11 | 2011-11-16 | 中国科学院软件研究所 | Distributed type dynamic cache expanding method and system supporting load balancing |
CN102427475A (en) * | 2011-12-08 | 2012-04-25 | 曙光信息产业(北京)有限公司 | Load balance scheduling system in cloud computing environment |
CN102932413A (en) * | 2012-09-26 | 2013-02-13 | 华为软件技术有限公司 | Computing resource allocation method, cloud management platform node and computing resource cluster |
CN103051564A (en) * | 2013-01-07 | 2013-04-17 | 杭州华三通信技术有限公司 | Dynamic resource allocation method and device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7099350B2 (en) * | 2001-04-24 | 2006-08-29 | Atitania, Ltd. | Method and apparatus for converting data between two dissimilar systems |
-
2014
- 2014-06-13 CN CN201410263629.5A patent/CN104023068B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101719081A (en) * | 2009-12-01 | 2010-06-02 | 北京大学 | Method for scheduling virtual machines |
CN102244685A (en) * | 2011-08-11 | 2011-11-16 | 中国科学院软件研究所 | Distributed type dynamic cache expanding method and system supporting load balancing |
CN102427475A (en) * | 2011-12-08 | 2012-04-25 | 曙光信息产业(北京)有限公司 | Load balance scheduling system in cloud computing environment |
CN102932413A (en) * | 2012-09-26 | 2013-02-13 | 华为软件技术有限公司 | Computing resource allocation method, cloud management platform node and computing resource cluster |
CN103051564A (en) * | 2013-01-07 | 2013-04-17 | 杭州华三通信技术有限公司 | Dynamic resource allocation method and device |
Also Published As
Publication number | Publication date |
---|---|
CN104023068A (en) | 2014-09-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104023068B (en) | A kind of method that Passive Mode elastic calculation scheduling of resource is realized in load balancing | |
Zhang et al. | High-performance isolation computing technology for smart IoT healthcare in cloud environments | |
CN104572307B (en) | The method that a kind of pair of virtual resource carries out flexible scheduling | |
CN101488098B (en) | Multi-core computing resource management system based on virtual computing technology | |
CN109918198A (en) | A kind of emulation cloud platform load dispatch system and method based on user characteristics prediction | |
Guazzone et al. | Exploiting VM migration for the automated power and performance management of green cloud computing systems | |
US11206193B2 (en) | Method and system for provisioning resources in cloud computing | |
CN107018175A (en) | The dispatching method and device of mobile cloud computing platform | |
CN109714431A (en) | A kind of edge calculations method and apparatus of Internet of Things intelligence sensor | |
CN108924221A (en) | The method and apparatus for distributing resource | |
CN106415498A (en) | Virtual machine power management | |
CN102662750A (en) | Virtual machine resource optimal control method and control system based on elastic virtual machine pool | |
CN107659609A (en) | A kind of deep learning support platform and deep learning training method based on cloud computing | |
CN104503832B (en) | A kind of scheduling virtual machine system and method for fair and efficiency balance | |
CN105607952B (en) | Method and device for scheduling virtualized resources | |
Ismaeel et al. | Using ELM techniques to predict data centre VM requests | |
CN107203255A (en) | Power-economizing method and device are migrated in a kind of network function virtualized environment | |
CN110460490A (en) | Server cluster based on Internet of Things monitors system and method | |
CN105975047A (en) | Method and system for adjusting power consumption of cloud data centre | |
WO2021204227A1 (en) | Processing resource allocation method and apparatus for neural network training and intelligent analysis | |
KR20220030050A (en) | Method and server for adjusting the allocation of computing resources to a plurality of vnf(virtualized network function) | |
CN106330558A (en) | Controller load prediction system and method applied to software defined network | |
Li et al. | RLDRM: Closed loop dynamic cache allocation with deep reinforcement learning for network function virtualization | |
WO2023213097A1 (en) | Online monitoring and analyzing system for energy consumption of elevator | |
CN105302641A (en) | Node scheduling method and apparatus in virtual cluster |
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 | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20220111 Address after: 100176 2403, floor 24, building 1, yard 8, KEGU 1st Street, Beijing Economic and Technological Development Zone, Daxing District, Beijing (Yizhuang group, high-end industrial area of Beijing Pilot Free Trade Zone) Patentee after: Tongming Zhiyun (Beijing) Technology Co.,Ltd. Address before: 101111 Room 302, 3 / F, building 4, yard 29, Jinghai Second Road, Beijing Economic and Technological Development Zone, Daxing District, Beijing Patentee before: SINOGRID SOFTWARE SYSTEMS LTD. |