CN106973030A - A kind of cloud artificial resource dispatching method based on SLA - Google Patents

Abstract

The invention discloses a kind of cloud artificial resource dispatching method based on SLA, the step of this method, includes being based on SLA SLA, artificial resource dynamically distributes Benefit Model S1 is set up, the task reception mechanism S2 based on SLA is set up and sets up the task reception mechanism based on SLA, based on SLA resource dispatching strategies, selection and the allotment S3 of artificial resource are carried out.Technical scheme of the present invention is based on sla management, scheduling of resource scheduling theory, propose a kind of cloud simulation virtual resource regulating method constrained based on SLA, ensure the service class requirements customized according to user, the global effect of the Simulation Application base support environment that system is provided is maximized.

Description

A kind of cloud artificial resource dispatching method based on SLA

Technical field

The present invention relates to the dispatching method of cloud artificial resource, more particularly to a kind of cloud emulation money based on SLA Source dispatching method.

Background technology

Cloud emulation is the network measuring emulation technology based on cloud computing theory, and the scheduling of cloud artificial resource is referred to In cloud simulated environment, according to cloud resource service condition, in the case where meeting user task demand, to not Enter the process of Mobile state adjustment and distribution with cloud artificial resource used in user, and by this scheduling, Realize distributing rationally for cloud artificial resource.The scheduling of resource of current cloud emulation is all based on the available of resource service Property, do not carry out fine-grained scheduling for service quality.

For service quality QoS in cloud environment, of greatest concern at present is to be based on SLA (Service Level Agreements, SLA) describe.SLA is between the demander of service and the supplier of service Sign on the specific horizontal formal agreement of service.Scheduling of resource technology based on SLA is to meet The service quality rating specified in the SLA of signature and support resource provider to carry out reconfiguring money to resource Source.And according to SLA difference, the now scheduling based on SLA must distinguish use according to its priority The importance at family.High-priority users obtain quality of service guarantee prior to low priority user, while ensureing On the premise of the service quality of high-priority users is protected, the service of influence low priority user not too much Quality.

The content of the invention

The technical problem to be solved in the present invention is to provide a kind of cloud artificial resource dispatching method based on SLA, solution Certainly the artificial resource under cloud simulated environment carries out fine granularity scheduling problem according to service quality.

In order to solve the above technical problems, the present invention uses following technical proposals：

The step of a kind of cloud artificial resource dispatching method based on SLA, this method, includes

S1, based on SLA SLA, set up artificial resource dynamically distributes Benefit Model；

S2, task reception mechanism of the foundation based on SLA；

S3, task reception mechanism of the foundation based on SLA, based on SLA resource dispatching strategies, are emulated The selection and allotment of resource.

It is preferred that, the artificial resource scheduling model includes：

Task receives earnings pattern：Wherein, new task is virtual at i-th The overall use cost of in-fight service isThe expense R that completion task is paid^new, then service offer The income of business is prof^new；

It is any to receive cost model：Wherein, The reception cost of new task includes the unlatching cost of new virtual machineClose costTasks carrying CostMultiplexed transport costThe compensation needed with task promise breaking

Virtual machine opens cost model：Wherein, the unlatching cost of virtual machine depends on In the unit interval use cost P of i-th virtual machine_iWith opening time TO_i；

The closing cost model of virtual machine：Wherein, the closing of i-th virtual machine Cost depends on its long-run cost rate P_iWith shut-in time TC_i；

Tasks carrying cost model：Wherein, tasks carrying cost is depended on The expected time of taskWith the unit interval use cost P of required virtual machine_k；

Multiplexed transport cost model：Wherein, Multiplexed transport cost depends on input file size inDS^new, output file cost outDS^newWith for this The unit-sized input and output transmission cost inP of virtual machine_iAnd outP_i；

Multiplexed transport time model：Wherein,；

Task response-time model：

Targeted yield return rate model：

It is preferred that, index includes handling capacity, response time, availability and request arrival in the step S2 Rate.

It is preferred that, the size of the handling capacity is：

Wherein, n_riTotal service request number of moment i system processing is represented, t represents that handling all services asks The time asked.

It is preferred that, the response time is：

Wherein, rt_iRepresent that request that certain user at a time sends results in the response time of service, t It is the average time section to the multiple measurement of service status.

It is preferred that, the calculation formula of the availability is：

Wherein, T_availThe time that service can normally be run within a period of time is represented, T is to measure this section Total time.

It is preferred that, the calculation formula of the request arriving rate is：

Wherein, ap_iRepresent total service request number that cloud computing system is received in a period of time of measurement, AP_i Represent total service request number that all service consumers are sent in the period.

Beneficial effects of the present invention are as follows：

Technical scheme of the present invention is based on sla management, scheduling of resource scheduling theory, it is proposed that Yi Zhongji The cloud simulation virtual resource regulating method constrained in SLA, it is ensured that the service class requirements customized according to user, The global effect of Simulation Application base support environment that system is provided is maximized.

Brief description of the drawings

The embodiment to the present invention is described in further detail below in conjunction with the accompanying drawings；

Fig. 1 shows cloud artificial resource dispatching method schematic diagram of the present invention.

Embodiment

In order to illustrate more clearly of the present invention, the present invention is done further with reference to preferred embodiments and drawings Explanation.Similar part is indicated with identical reference in accompanying drawing.Those skilled in the art should Work as understanding, below specifically described content be illustrative and be not restrictive, should not limit this with this The protection domain of invention.

The invention discloses a kind of cloud artificial resource dispatching method based on SLA SLA, this method Specific steps include：

The first step：Artificial resource dynamically distributes Benefit Model is set up based on SLA

The target of resource regulating method scheduling of the present invention is that resource is carried under the requirement for meeting customer sla Maximum return is obtained for business.

For the definition of evaluating in SLA contracts, Resource dynamic allocation consider SLA models mainly have with Under several indexs：

Deadline, represents task latest finishing time specified in SLA contracts；

Reward, represents that user specified in SLA contracts needs the expense paid；

Input File Size, represent that user submits task file data size；

- individual new user have submitted a new task between certain moment, and be provided in SLA, completes task and pays The expense R gone out^new, finally complete moment DL^new。

According to above index, Resource dynamic allocation Benefit Model is set up：

Define 1：Task receives income.New task is in the overall use cost of i-th virtual in-fight serviceThe expense R that completion task is paid^new, then the income of service provider is prof^new：

Define 2 tasks and receive cost.The reception cost of new task includes the unlatching cost of new virtual machine Close costTasks carrying costMultiplexed transport costBeing broken a contract with task needs Compensation

Define 3 virtual machines and open cost.When the unlatching cost of virtual machine depends on the unit of i-th virtual machine Between use cost P_iWith opening time TO_i:

Define the closing cost of 4 virtual machines.The closing cost of i-th virtual machine depend on its unit interval into This P_iWith shut-in time TC_i:

Define 5 tasks carrying costs.Tasks carrying cost depends on the expected time of task With the unit interval use cost P of required virtual machine_k：

Define 6 multiplexed transport costs.Multiplexed transport cost depends on input file size inDS^new, output text Part cost outDS^newWith the unit-sized input and output transmission cost inP for the virtual machine_iAnd outP_i：

Defined for 7 multiplexed transport times.The file transmission total time that task needs：

Define 8 task response-times.Task is able to carry out the time completed on a virtual machineFor：

Define 9 targeted yields.New task needs expectation profit return when receivingThen income is returned Report rate u is：

Second step：Set up the task reception mechanism based on SLA

Mainly there are two kinds of situations for the reception of new task：

(1) when being connected to new task request, SLA parameters is primarily based on, are consulted with user, and according to upper The Benefit Model of foundation is stated, directly calculates and the profit return that the task is is performed on the artificial resource newly opened Whether rate u is more than 1, if it is, receiving the task, otherwise, turns (2) and judges；

(2) judge the task whether hot job, continue with, if it is, directly opening new emulation Resource performs operation, otherwise, and refusal receives an assignment, and renegotiates.

3rd step：Based on SLA resource dispatching strategies

Cloud emulation platform should monitor SLA index service scenarios in service operation, and virtual machine money is monitored again The running situation in source is simultaneously stored in information bank.In order to ensure the normal operation of virtual machine, two references are set first Threshold value：Resource utilization upper limit threshold and the offline threshold value of resource utilization, should ensure that the money of each virtual machine Source utilization rate is maintained in the threshold range.

When receiving new task, virtual machine service is selected.Dispatch situation mainly has following several：

A) when the resource utilization for detecting virtual machine exceedes upper limit threshold, then no longer added to the virtual machine Any new task, untill its resource utilization drops in normal range (NR)；

B) when detecting the resource utilization of virtual machine between upper limit threshold and lower threshold, i.e. normal range (NR) When in value, if then judging whether virtual machine addition new task can cause resource utilization to exceed upper limit threshold, If will not occur, the task distributes to this virtual machine；Otherwise, next virtual machine is selected to judge；

C) when the resource utilization for detecting virtual machine is less than lower threshold, then detect that its brotgher of node is virtual Whether machine resource utilization is in normal range (NR), and now father node is it is determined that new task can not be received by crossing.Sentence If the task of breaking distributes to whether its brotgher of node can cause resource utilization to exceed upper limit threshold, if more than upper Threshold value is limited, then this virtual machine performs new task；If no more than upper limit threshold, new task distributes to brother Node virtual machine is performed, and the virtual machine enters holding state after having performed the task of wait, and its child node is virtual Machine, which has been performed, enters off-mode after the task of wait, until the node resource utilization rate is in normal condition, Its child node virtual machine recovers holding state and waits new task；

In summary, technical scheme of the present invention is carried based on sla management, scheduling of resource scheduling theory A kind of cloud simulation virtual resource regulating method constrained based on SLA is gone out, it is ensured that the clothes customized according to user Business class requirement, the global effect of Simulation Application base support environment that system is provided is maximized.

Obviously, the above embodiment of the present invention is only intended to clearly illustrate example of the present invention, and simultaneously Non- is the restriction to embodiments of the present invention, for those of ordinary skill in the field, above-mentioned It can also be made other changes in different forms on the basis of explanation, here can not be to all implementation Mode is exhaustive, every to belong to the obvious changes or variations that technical scheme is extended out Still in the row of protection scope of the present invention.

Claims (7)

1. a kind of cloud artificial resource dispatching method based on SLA, it is characterised in that wrap the step of this method Include

S1, based on SLA SLA, set up artificial resource dynamically distributes Benefit Model；

S2, task reception mechanism of the foundation based on SLA；

S3, task reception mechanism of the foundation based on SLA, based on SLA resource dispatching strategies, are emulated The selection and allotment of resource.

2. cloud artificial resource dispatching method according to claim 1, it is characterised in that the emulation money Source scheduling model includes：

Task receives earnings pattern：Wherein, new task is virtual at i-th The overall use cost of in-fight service isThe expense R that completion task is paid^new, then service offer The income of business is prof^new；

It is any to receive cost model： ${\mathrm{cost}}_i^{new}={\mathrm{IC}}_i^{new}+{\mathrm{CC}}_i^{new}+{\mathrm{PC}}_i^{new}+{\mathrm{DTC}}_i^{new}+{\mathrm{PDC}}_i^{new},$ Wherein, The reception cost of new task includes the unlatching cost of new virtual machineClose costTasks carrying CostMultiplexed transport costThe compensation needed with task promise breaking

Virtual machine opens cost model：Wherein, the unlatching cost of virtual machine depends on In the unit interval use cost P of i-th virtual machine_iWith opening time TO_i；

The closing cost model of virtual machine：Wherein, the closing of i-th virtual machine Cost depends on its long-run cost rate P_iWith shut-in time TC_i；

Tasks carrying cost model：Wherein, tasks carrying cost is depended on The expected time of taskWith the unit interval use cost P of required virtual machine_k；

Multiplexed transport cost model： ${\mathrm{DTC}}_i^{new}={\mathrm{outDS}}^{new}*{\mathrm{inP}}_i+{\mathrm{outDS}}^{new}*{\mathrm{outP}}_i,$ Wherein, Multiplexed transport cost depends on input file size inDS^new, output file cost outDS^newWith for this The unit-sized input and output transmission cost inP of virtual machine_iAnd outP_i；

Multiplexed transport time model： ${\mathrm{DTT}}_i^{new}={\mathrm{inDT}}_i^{new}+{\mathrm{outDT}}_i^{new},$ Wherein,；

Task response-time model： ${\mathrm{RT}}_i^{new}=\left\{\begin{array}{c}{\Σ}_{m=1}^M{\mathrm{procT}}_i^m+{\mathrm{procT}}_i^{new}\\ {\mathrm{procT}}_i^{new}+{\mathrm{DTT}}_i^{new}+{\mathrm{TO}}_i\end{array}\right.;$

Targeted yield return rate model：

3. cloud artificial resource dispatching method according to claim 1, it is characterised in that the step Index includes handling capacity, response time, availability and request arriving rate in S2.

4. cloud artificial resource dispatching method according to claim 3, it is characterised in that described to handle up The size of amount is：

$TP=\frac{\underset{0<i<t}{\&Sigma;}{n}_{ri}}{t}$

Wherein, n_riTotal service request number of moment i system processing is represented, t represents that handling all services asks The time asked.

5. cloud artificial resource dispatching method according to claim 3, it is characterised in that the response Time is：

$MRT=\frac{\underset{0<i<t}{\&Sigma;}{\mathrm{rt}}_i}{t}$

Wherein, rt_iRepresent that request that certain user at a time sends results in the response time of service, t It is the average time section to the multiple measurement of service status.

6. cloud artificial resource dispatching method according to claim 3, it is characterised in that described to use The calculation formula of property is：

$AvailRatio=\frac{\underset{0<i<T}{\&Sigma;}{T}_{avail}}{T}$

Wherein, T_availThe time that service can normally be run within a period of time is represented, T is to measure this section Total time.

7. cloud artificial resource dispatching method according to claim 3, it is characterised in that the request The calculation formula of arrival rate is：

$AR=\frac{\underset{O<i<t}{\&Sigma;}{\mathrm{ap}}_i}{\underset{O<i<t}{\&Sigma;}{\mathrm{AP}}_i}$

Wherein, ap_iRepresent total service request number that cloud computing system is received in a period of time of measurement, AP_i Represent total service request number that all service consumers are sent in the period.