CN106020982A - Method for simulating resource consumption of software component - Google Patents

Abstract

The invention discloses a method for simulating resource consumption of a software component. The method comprises the following steps of (1) determining mathematical distribution of a CPU (Central Processing Unit) resource consumption rate of the component and mathematical distribution of the memory resource consumption according to component running parameter, and establishing a CPU resource consumption model and a memory resource consumption model; (2) respectively constructing a CPU resource simulation component and a memory resource consumption simulation component according to the CPU resource consumption model and the memory resource consumption model; (3) deploying and running the simulation component according to the scale of a to-be-evaluated software architecture, and obtaining real-time CPU resource consumption status data and memory resource consumption status data of the component; and (4) writing the real-time resource consumption status data into a corresponding evaluation database. According to the method, the performance of the system architecture can be evaluated at the early development stage of the system, and the effects of reducing the development cost and reducing the research and development risk are realized.

Description

A kind of component software resource consumption analogy method

Technical field

The invention belongs to the Software Architecture Evaluation field of computer, particularly to a kind of component software resource consumption mould Plan method.

Background technology

For software architecture, the major issue paid close attention to is exactly quality, especially to large-scale multiple Miscellaneous software system is the most so.So a software architecture is estimated, to guaranteeing that the quality of final system is to closing weight Want.One software system structure is estimated, is the quality in order to predict it before being fabricated in system, it is not absolutely required to The most accurate assessment result, by analyzing the related data main impact for mass of system of system structure, and then proposes Improve.The estimation of component software resource consumption is a core in Software Architecture Evaluation and key, mainly includes software group Cpu resource consumption and memory source when part runs consume two.Further, the consumption simulation of component resources also relies on and comments every time The concrete deployment scale estimated.

Modern software is increasingly using component technology and develops, and uses component technology exploitation software to improve code Reusability and the efficiency of software development.Component technology has had become as the basis of Distributed Calculation and Web service, advises greatly The exploitation of modular system depends on the assembling of assembly.Research to assembly at present is concentrated mainly on the research to concrete component technology, Such as COM+ assembly .NET technology and JavaBeans etc., or how research uses component technology to realize specific software system System, i.e. framework technology.But, there are no report about the correlational study of resource consumption evaluation method during assembly operating, thus lead Cause developer and the performance of system cannot be estimated in early days in the exploitation of system.

Before system development, the performance of needs assessment prognoses system, need resource consumption state during assembly operating is entered Row estimation.Traditional method is directly to assume based on experience or set up performance model, and some performance models and program are used Numerical algorithm is closely related, lacks objectivity, and process of setting up needs to spend substantial amounts of expert's manpower, and model calculates the time also Longer, it is impossible to realize automatization；Instrument that other performance models are used and modeling method are based on certain certain types of meter Calculate platform, there is platform dependence, and only use its minority strategic alliance is open.

If the design of system architecture lacks corresponding Performance Evaluation, then it is unfavorable for its iteration optimization, it will directly It is related to the quality of system.When therefore, using this resource consumption evaluation method objective effectively component software to be run Resource consumption is simulated, and then is objective effective assessment and the iteration optimization design foundation thereof of whole host system architecture Basis, and reduce the cost and risk of system research and development.

Summary of the invention

Goal of the invention: in order to overcome the deficiencies in the prior art, the present invention provides the one can in system development early System architecture performance is estimated by the phase, and the component software of the effect reduce development cost, reducing R&D risk provides Source consumes analogy method.

Technical scheme: for achieving the above object, the present invention provides a kind of component software resource consumption analogy method, specifically wraps Include following steps:

Step one: mathematical distribution and memory source according to the cpu resource consumption rate of assembly operating parameter determination assembly disappear The mathematical distribution of consumption, sets up cpu resource consumption models and memory source consumption models；

Step 2: respectively according to cpu resource consumption models and memory source consumption models, constructs cpu resource simulated assembly Simulated assembly is consumed with memory source；

Step 3: according to the scale of software architecture to be assessed, disposes and runs simulated assembly, securing component real-time Cpu resource consumption state data and memory source consumption state data；

Step 4: data base is assessed in the write of real time resources consumption state data accordingly.

Further, in described step one, the computing formula of cpu resource consumption rate is,

$cpu\_usage=\frac{\cos t\_time}{ST*core\_num}$

Wherein core_num is the check figure of CPU, and ST is the sampling period, and cost_time is the CPU time consumed in ST, Cpu_usage is CPU consumption rate.

Further, in described step one, the mathematical distribution of cpu resource consumption rate specifically determines that method is: disappear according to resource Consumption rate and assembly are in each sampling period ST relation i.e. calculating of cpu resource consumption rate to the elapsed time cost_time of CPU It is exponential that formula obtains the mathematical distribution of cpu resource consumption rate.

Further, in described step one, the mathematical distribution of memory source consumption specifically determines that method is: disappear according to assembly It is normal distribution that the variation relation of dynamically change and the memory consumption rate of consumption internal memory obtains the mathematical distribution of memory source consumption.

Further, described step one sets up cpu resource consumption models method particularly includes: provide according to the CPU of assembly Source consumption rate and assembly be relation to the holding time of CPU in each sampling period, determines and provides as CPU using exponential e (λ) Source consumption models, meanwhile, using 1/ (0.4*M) as the value of λ, M is that cpu resource consumes maximum；Wherein λ is a parameter, clothes From exponential e (λ) that parameter is λ, the mathematic expectaion of exponential e (λ) is 1/ λ.

Further, described step one sets up memory source consumption models method particularly includes: with normal distribution N (μ, σ²) as memory source consumption models, meanwhile, using M'/2 as the value of average value mu, using M'/6 as the value of standard deviation sigma, its In, M' is the maximum that memory source consumes.

Further, the method building cpu resource simulated assembly in described step 2 is: for cpu resource consumption rate, clothes From exponential, set U as obeying being uniformly distributed on (0,1), then the execution logic of the simulated assembly built is: x=-λ * log(1-U)。

Further, the method building memory source simulated assembly in described step 2 is: Normal Distribution, simulation group The structure of part needs to firstly generate two randoms number, and the quadratic sum of random number have to be larger than 0 less than 1；Then utilize the two with The quadratic sum of machine number and random number determines σ in normal distribution²Coefficient, specifically determine that step is as follows:

Step 1: take u1=random number rand ()/RAND_MAX；U2=random number rand ()/RAND_MAX；Rand () uses In producing a system random number, RAND_MAX is then the maximum of random number；

Step 2: take v1=2*u1-1, v2=2*u2-1；

Step 3: take s=u1*u1+u2*u2；

Step 4: if s >=1 or s=0, continue step 1, if s ＜ 1, enter step 5；

Step 5: take

Step 6: output x1* σ²+ μ and x2* σ²The simulation that+μ consumes as memory source.

Beneficial effect: the present invention compared with prior art has the advantage that

1, the objective characteristics of resource consumption, therefore resource consumption when in the present invention, resource consumption model considers assembly operating Simulation the most accurate.

2, the resource in the present invention disappear analogy method and particular platform unrelated, there is platform independence, therefore, be conducive to move Plant and use universality, convenience；The simulation process of the present invention is simple.

Accompanying drawing explanation

Fig. 1 is the flow chart of the present invention；

Fig. 2 is the cpu resource consumption figure of assembly 1 in embodiment；

Fig. 3 is the memory source consumption figure of assembly 1 in embodiment；

Fig. 4 is the load balancing desired value distribution that the inventive method is applied to accessed by distributed system based on DDS Figure.

Detailed description of the invention

Below technical solution of the present invention is described in detail, but protection scope of the present invention is not limited to described enforcement Example.

As it is shown in figure 1, a kind of component software resource consumption analogy method of the present invention, specifically include following steps:

Step one: mathematical distribution and memory source according to the cpu resource consumption rate of assembly operating parameter determination assembly disappear The mathematical distribution of consumption, sets up cpu resource consumption models and memory source consumption models；

Step 2: respectively according to cpu resource consumption models and memory source consumption models, constructs cpu resource simulated assembly Simulated assembly is consumed with memory source；

Step 3: according to the scale of software architecture to be assessed, disposes and runs simulated assembly, securing component real-time Cpu resource consumption state data and memory source consumption state data；

Step 4: data base is assessed in the write of real time resources consumption state data accordingly.

Above-mentioned steps one determines mathematical distribution and resource consumption that the cpu resource consumption rate of assembly and memory source consume Model is specific as follows:

1) construction method of cpu resource consumption models is: simulate for cpu resource consumption during assembly operating, sets CPU Check figure be core_num, the sampling period is ST, in ST consume CPU time be cost_time, CPU consumption rate cpu_ Usage represents, therefore the computing formula of CPU consumption rate is as follows.

$cpu\_usage=\frac{\cos t\_time}{ST*core\_num}$

Definition stochastic variable T represents the assembly elapsed time in each sampling period to CPU, and assembly is in sampling period ST Consume the time T of CPU₁With its elapsed time T to CPU in the sampling period subsequently₂It it is separate, it means that group Part operation time in sampling period ST and its operation time in the next sampling period are incoherent, and assembly is often The operation time in the individual sampling period meets the feature without memory of exponential, and therefore stochastic variable T meets exponential Characteristic, i.e. T～e (λ).Wherein, T～e (λ) represented in each sampling period, and assembly meets finger to the resource consumption T of CPU Number distribution.Owing to stochastic variable T represents the assembly holding time (cost_ in the most above-mentioned formula in each employing cycle to CPU Time), then CPU consumption rate meets the characteristic of exponential equally.

The CPU usage maximum of note assembly is M, then be M (0 ＜ M ＜ 100) for a CPU consumption rate maximum Assembly, it takies cpu resource substantially without occupancy with M at capacity during running, therefore takes its meansigma methods The factor is α, and the meansigma methods that i.e. it takies CPU within the whole time period is α * M, knows according to expectation formulaFor group Part, according to the cumulative distribution function of exponential, its CPU usage probability less than t is as follows:

$P(x\≤t)=F\left(t\right)=1-e^{-\mathrm{\λ}t}=1-e^{-\frac{1}{\mathrm{\α}*M}*t}$

In actual applications, cpu resource consumption rate can be influenced by many factors, is M for actual maximum occupation value, Maximum CPU consumption rate when it runs is in certain scope, and it is less than M with the biggest probability, i.e.?When taking t=0.9M, p ＞ 0.9, obtain meansigma methods factor-alpha Value is 0.4.Therefore it is M when the CPU quota of assembly, can generate and meetSimulated assembly.

2) construction method of memory source consumption models is: simulate for memory source consumption during assembly operating, due to The internal memory that assembly is consumed is dynamically change, and memory consumption rate can fluctuate in the range of one, and in the case of only a few Reach peak value, and because partial code needs memory-resident, so the memory consumption of assembly all can be more than certain value, assembly Memory consumption meets the characteristic of normal distribution.If the maximum memory consumption of assembly is M', the real-time consumption of assembly is random Variable X, X～N (μ, σ²), wherein N (μ, σ²) be normal distribution, the real-time of assembly consume X meet average be μ variance be σ² Normal distribution.Wherein, parameter μ determines meansigma methods, and parameter σ determines changing slope, and σ is the biggest, and then image is the most slow, it is meant that P The probability of (μ-σ ＜ x ＜ μ+σ) is the biggest, and sampled value, further off mean μ, illustrates that the constant interval of internal memory is bigger；σ is the least, anticipates It is less that taste internal memory constant interval, therefore according to practical situation, sets larger by the value of σ, and such as about 5, the scope of σ probably exists 0-6.Internal memory can be floated in a big scope.The minima of EMS memory occupation is set as μ-3 σ, and its maximum is μ+3 σ. Therefore the maximum memory consumption of assignment component is M' and according to μ+3 σ=M'(wherein μ=M'/2), meet parameter bar by generation The simulated assembly of the normal distribution of part, completes assembly and is in operation the estimation to internal memory resource consumption.

The method building corresponding simulated assembly in above-mentioned steps two is specific as follows:

For cpu resource consumption rate, obey exponential, sets U as being uniformly distributed in obedience (0,1), then build The execution logic of simulated assembly is: x=-λ * log (1-U).

For memory source consumption, Normal Distribution N (μ, σ²), following step gives generation and meets normal distribution N (μ,σ²) algorithm of sample data.If u1 and u2 is the variable obeying U (0,1), the structure of simulated assembly needs to firstly generate Two random number u1 and u2, its quadratic sum s have to be larger than 0 and is less than 1, and such s also obeys U (0,1), and separate with u1/u2； So, the X1 built with this, X2 are the separate variablees obeying standard normal distribution, wherein meet ρ distribution.Wherein X1, X2 are the variablees obeying standard normal distribution built, then with this, structure average is u, and variance is σ²Normal distribution Sample data, build formula be: x1* σ²+μ,x2*σ²+ μ, thus carry out emulated memory resource consumption.Because memory source consumption It is exactly Normal Distribution N (μ, σ²).Secondly, utilizing X1 and X2 to build and meet average for u, variance is σ²The sample of normal distribution Notebook data.Specifically determine that step is as follows:

Step 1: take u1=random number rand ()/RAND_MAX；U2=random number rand ()/RAND_MAX；Rand () uses In producing a system random number, RAND_MAX is then the maximum of random number；

Step 2: take v1=2*u1-1, v2=2*u2-1；

Step 3: take s=u1*u1+u2*u2；

Step 4: if s >=1 or s=0, continue step 1, if s ＜ 1, enter step 5；

Step 5: take

Step 6: output x1* σ²+ μ and x2* σ²The simulation that+μ consumes as memory source.

Cpu resource consumption calculations in above-mentioned steps three, for the assembly that cpu resource consumption is M (0 < M < 100) of assembly, Generate and CPU core number core_num thread as much, thread will perform corresponding code, in main loop runs in code Deposit distribution instruction, thread SLEEP instruction and internal memory to release order, carried out the cpu resource consumption of simulated assembly by such operation. If the cpu resource occupancy of certain simulated assembly that the method in above-mentioned steps generates is tm (non-percentage ratio), then to certain mould Intending assembly, the step making thread sleep operation is as follows:

Step (1): the cpu resource elapsed time ktm of this assembly, computational methods: ktm=tm/100* sampling period ST；

Step (2): calculate the sleeptime=ST-ktm length of one's sleep that this assembly needs；

Step (3): call the interface that in testing machines, operating system provides and obtain the current sample period time started of CPU lasttime；

Step (4): call the interface that in testing machines, operating system provides and obtain CPU current time currenttime；

Step (5): if currenttime-lasttime < ktm, represent that present day analog component run-time is not enough, need Continue to run with, enter (4)；Work as currenttime-lasttime=ktm, be then downwardly into 6)；

Step (6): renewal lasttime is currentime；

Step (7): utilize system to call, makes thread sleep a period of time, length value during this period of time with above-mentioned in Sleeptime value is equal；

To (7), each simulated assembly is repeatedly performed above-mentioned steps according to above-mentioned steps (1), until completing the most to be evaluated

Estimate the simulated assembly specified by architecture and dispose number.

In above-mentioned steps, memory consumption calculates, it is only necessary to creates a thread, performs corresponding code in thread, and code leads to Cross simulated assembly and called operations such as making thread sleep by system, it is achieved memory consumption simulation process during assembly operating, If the above-mentioned simulated assembly meeting normal distribution got is s, proceed as follows:

Step A: utilize the internal memory that function malloc application memory size is s'；

Step B: utilize system to call sleep and make thread sleep 500 milliseconds；

Step C: utilize function free releasing memory.

According to above-mentioned steps A to C, each simulated assembly is performed above-mentioned steps repeatedly, until completing current system to be assessed Simulated assembly specified by structure disposes number.

Embodiment: come concrete as host system using distributed system based on DDS (Data distributing) in the present invention The effectiveness of resource consumption analogy method during assembly operating proposed by the invention is described.The operation got obtained is relevant In assembly CPU consumption rate and memory consumption data, as the parameter of host system architecture test；Carry out host software system The test job of architecture, carries out Performance Evaluation to whole host system architecture.

In a distributed system based on DDS, load balance degree is that measurement system designs the most rational one Important indicator.Load balance degree is the highest, and the load consumption of the most each position is average.Need when calculating system load balancing degree Get the factors such as each bit CPU resource consumption rate and memory consumption rate.Before system does not complete, by side of the present invention Method provides the parameters such as cpu resource consumption rate required in test and memory consumption.

Distributed system based on DDS in the present invention includes: underwater sound observatory control system sys1 and decentralised control simulation System sys2, underwater sound observatory control system sys1 is made up of 200 assemblies, is deployed in respectively on 5 platform positions, and system comprises altogether Five tasks；Decentralised control analog systems sys2 is made up of 3 assemblies, is deployed in respectively on 3 platform positions, and system comprises three altogether Individual task.In test process, resource consumption situation when being run by the inventive method simulated assembly, and send the data to Index evaluation program calculates.

Experiment first passes through the assembly Component1 of underwater sound observatory control system sys1 and carrys out Demo Asset consumption calculations algorithm Flow process, finally the load balancing degrees of two cover systems is tested.

For the assembly Component1 of underwater sound observatory control system sys1, set its cpu resource maximum consumption rate as 10%, memory source maximum consumption is 100MB, carries out dry run, simultaneously by laboratory observation, obtain on windows platform During treating excess syndrome, consumption data is drawn a diagram, observe cpu resource consumption rate and memory consumption situation meet the most respectively exponential and Normal distribution.

Experimental procedure:

Step a: determine the parameter of distribution, simulates for cpu resource consumption, and the parameter lambda of exponential is 0.25, and internal memory provides The mathematical distribution parameter that source takies is:

μ=M/2=50000000byte, σ=50000000/3byte；

Step b: create thread and perform resource consumption simulation.

Cpu resource consumption figure as in figure 2 it is shown, wherein vertical coordinate be frequency, i.e. different consumption rates occur in whole experiment Number of times；Abscissa is CPU consumption rate, and such as 10.000 i.e. represent that CPU usage is 10%.Observation figure understands cpu resource and accounts for By the characteristic meeting exponential；

Memory source takies figure as it is shown on figure 3, wherein vertical coordinate is frequency, and the i.e. different sizes values that consume are in whole experiment The number of times occurred；Abscissa is that memory source consumes size, and such as 40.000 i.e. represent that memory consumption size is 40M.Observe figure Understand memory source consumption and meet the characteristic of normal distribution.

Step c: receive analog data, carry out the calculating of load balancing degrees.

System gathers data when the sampling period starts, including all component issuing subject on each position and topic of subscription Sums etc., read the simulation CPU utilizing the inventive method to obtain and memory consumption data simultaneously, then carry out load balancing degrees and refer to The calculating of scale value.

The load balancing desired value of system as shown in Figure 4, it can clearly be seen that two cover systems load balancing desired value become Changing the most more stable, fluctuating margin is less；But it is owing to the assembly of underwater sound observatory control system sys1 is more, the most more complicated, The load amplitude of variation of each position is relatively big, by contrast, and bearing of assembly simply decentralised control analog systems sys2 less, mutual Carry balanced intensity and be better than underwater sound observatory control system sys1.Utilize result and practical situation that resource occupation simulation algorithm evaluates It is consistent.

The preferred embodiment of the present invention described in detail above, but, the present invention is not limited in above-mentioned embodiment Detail, in the technology concept of the present invention, technical scheme can be carried out multiple equivalents, this A little equivalents belong to protection scope of the present invention.

Claims (8)

1. a component software resource consumption analogy method, it is characterised in that comprise the steps:

Step one: consume according to the mathematical distribution of the cpu resource consumption rate of assembly operating parameter determination assembly and memory source Mathematical distribution, sets up cpu resource consumption models and memory source consumption models；

Step 2: respectively according to cpu resource consumption models and memory source consumption models, structure cpu resource simulated assembly and interior Deposit resource consumption simulated assembly；

Step 3: according to the scale of software architecture to be assessed, disposes and runs simulated assembly, the real-time CPU of securing component Resource consumption status data and memory source consumption state data；

Step 4: data base is assessed in the write of real time resources consumption state data accordingly.

A kind of component software resource consumption analogy method the most according to claim 1, it is characterised in that in described step one The computing formula of cpu resource consumption rate is,

$cpu\_usage=\frac{\cos t\_time}{ST*core\_num}$

Wherein core_num is the check figure of CPU, and ST is the sampling period, and cost_time is the CPU time consumed in ST, cpu_ Usage is CPU consumption rate.

A kind of component software resource consumption analogy method the most according to claim 2, it is characterised in that in described step one The mathematical distribution of cpu resource consumption rate specifically determines that method is: according to resource consumption rate with assembly each sampling period ST pair The relation of the elapsed time cost_time of the CPU i.e. computing formula of cpu resource consumption rate obtains the mathematics of cpu resource consumption rate It is distributed as exponential.

4. according to a kind of component software resource consumption analogy method one of claims 1 to 3 Suo Shu, it is characterised in that described In step one, the mathematical distribution of memory source consumption specifically determines that method is: consume dynamically change and the internal memory of internal memory according to assembly It is normal distribution that the variation relation of consumption rate obtains the mathematical distribution of memory source consumption.

A kind of component software resource consumption analogy method the most according to claim 4, it is characterised in that in described step one Set up cpu resource consumption models method particularly includes: cpu resource consumption rate and assembly according to assembly are in each sampling period pair The relation of the holding time of CPU, determines using exponential e (λ) as cpu resource consumption models, makees with 1/ (0.4*M) meanwhile For the value of λ, M is that cpu resource consumes maximum；Wherein λ is a parameter, obeys exponential e (λ) that parameter is λ, and index divides The mathematic expectaion of cloth e (λ) is 1/ λ.

A kind of component software resource consumption analogy method the most according to claim 4, it is characterised in that in described step one Set up memory source consumption models method particularly includes: with normal distribution N (μ, σ²) as memory source consumption models, meanwhile, Using M'/2 as the value of average value mu, using M'/6 as the value of standard deviation sigma, wherein, M' is the maximum that memory source consumes.

A kind of component software resource consumption analogy method the most according to claim 5, it is characterised in that in described step 2 The method building cpu resource simulated assembly is: for cpu resource consumption rate, obeys exponential, sets U as obeying on (0,1) Be uniformly distributed, then the execution logic of the simulated assembly built is: x=-λ * log (1-U).

A kind of component software resource consumption analogy method the most according to claim 1, it is characterised in that in described step 2 The method building memory source simulated assembly is: Normal Distribution, and the structure of simulated assembly needs to firstly generate two at random Number, the quadratic sum of random number have to be larger than 0 less than 1；Then the quadratic sum utilizing the two random number and random number is just determined σ in state distribution²Coefficient；Specifically determine that step is as follows:

Step 1: take u1=random number rand ()/RAND_MAX；U2=random number rand ()/RAND_MAX；Rand () is used for producing A raw system random number, RAND_MAX is then the maximum of random number；

Step 2: take v1=2*u1-1, v2=2*u2-1；

Step 3: take s=u1*u1+u2*u2；

Step 4: if s >=1 or s=0, continue step 1, if s ＜ 1, enter step 5；

Step 5: take

Step 6: output x1* σ²+ μ and x2* σ²The simulation that+μ consumes as memory source.