Specific embodiment
Here exemplary embodiment will be illustrated in detail, its example is illustrated in the accompanying drawings.Following
When description is related to accompanying drawing, unless otherwise indicated, the same numbers in different accompanying drawings represent same or analogous
Key element.Implementation method described in following exemplary embodiment does not represent the institute consistent with the application
There is implementation method.Conversely, they are only with described in detail in such as appended claims, the application
The example of the consistent apparatus and method of a little aspects.
The disclosure illustrates how the property to the platform by taking a platform for processing life fee payment service as an example
More accurate and reliable test can be carried out.Such as the example of Fig. 1, a business platform 11 is illustrated, should
Business platform 11 can be interacted with many external agencies, such as mechanism 12 (for example, bank),
Mechanism 13 (for example, water rate enterprise is received in Shanghai), mechanism 14 (for example, combustion gas expense enterprise is received in Beijing) etc.,
For process to should mechanism business, for example, interacted to process transferred account service with mechanism 12, with mechanism
13 interactions pay business to process water rate, are interacted with mechanism 14 and pay business to process combustion gas expense.
For each external agency, business platform can be entered by least one business interface with the mechanism
Row communication, interface 15a, interface 15b, interface 15c as illustrated in figure 1 etc.;For example, business is flat
Platform 11 can be started logical with mechanism 13 when the service request that water rate is paid is received with calling interface 15b
Letter.The disclosure, for the performance test of business platform, can test the platform to bear great industry
How long business amount, or the platform can complete the business of request to the response time of service request;This property
Can also be represented with the performance indications of above-mentioned business interface, e.g., the number of concurrent of business interface, connect
Response time of mouth etc..
Fig. 2 illustrates the principle of the performance test methods of the disclosure, refers to Fig. 2, it is assumed that develop industry
Be engaged in the redaction of platform, be to the test of the redaction platform it is online under carry out, and at the business of Fig. 1
Reason is run on line, and both can be referred to as " system under line " and " inline system ".Under online
In system, including " platform testing system ", the business that the platform testing system is used to test redaction is put down
Platform, then, it is necessary to the parameter of some tests is input into the platform testing system, for test environment
Build so that the business platform of redaction is tested in the simulation test environment built.
Also include " big data modeling " in system under the line of Fig. 2, the system is used to set up to industry
It is engaged in the service feature model that is estimated of business interface performance of platform, will be using arriving during modeling
Business platform service data, and the present embodiment can be by obtaining resource consumption mould in business platform service data
Intend parameter, interface response delay parameter and service test data, by these data inputs to platform test system
System, is built into the test environment for platform testing system.For example, resource consumption analog parameter can be with mould
Intend the consumption of the resources such as CPU, the internal memory of business platform, it is outer that interface response delay parameter can be used for simulation
The response delay time of portion mechanism, and service test data is used for the business number that analog service platform is received
According to such as to the request execution transferred account service of mechanism 12.
The principle of the performance test methods of the above-mentioned disclosure illustrated in Figure 2 is, according to by inline system
The business platform service data for obtaining builds the test environment of system under line, and this mode can be more preferable
Environment on artificial line, closer to business on line, so as to obtain more accurately test result.As follows will be detailed
Thin description how the |input paramete according to needed for business platform service data obtains platform testing system:
Data acquisition and cleaning:
The data acquisition of inline system, the business that can gather inline system by data acquisition module 24 is put down
Platform service data, including to daily record on line and the collection of service feature data.For example, business can be obtained
Platform calls some daily record datas of external agency's interface, such as, in certain time to certain external agency
Perform some type of business, external agency is to response time of the business, and platform CPU at that time
Consumption etc..Also include other kinds of parameter in daily record data, no longer describe in detail.Wherein, platform CPU disappears
Consumption or platform internal memory consumption etc. belong to the data of platform resource consumption, can be gathered by snapshot tool.
For example, the platform resource consumption data for corresponding to multiple sampling time point respectively can be gathered by snapshot tool.
Wherein, the data such as platform CPU consumption or platform internal memory consumption are properly termed as resource consumption analog parameter,
Belong to the service environment data of system level during business platform runs.And it is above-mentioned in certain time to certain
Individual external agency performs the daily record datas such as some type of business, and being properly termed as should during business platform runs
With the service operation data of aspect, business platform service data includes above-mentioned service environment data and business
Service data.
After data on the line for being gathered, line number can be entered to these data according to predefined metadata
According to cleaning, obtain subsequently modeling the format data collection for using.I.e. metadata is the number for describing data attribute
According to, metadata can be pre-defined, further according to data on the line of metadata cleaning collection, finally give full
The data acquisition system of sufficient metadata form.Here data cleansing is selected and used from the data of collection
Target data, not all data are all used to model.Table 1 below illustrates part metadata
Form:
The metadata form of table 1
Field name |
Explanation of field |
Data type |
biz_time |
Business time of origin |
String |
biz_type |
Type of service |
String |
sub_biz_type |
Business subtype |
String |
charge_inst |
Enter an item of expenditure in the accounts mechanism |
String |
chargeoff_inst |
Write off mechanism |
String |
biz_cpu |
Business time of origin piece CPU occupation rates |
Integer |
biz_mem |
Business time of origin piece memory usage |
Integer |
biz_load |
The Load of business time of origin piece |
Float |
biz_tps |
The TPS of business time of origin piece |
Integer |
biz_concurrency |
The interface number of concurrent of business time of origin piece |
Integer |
biz_rspTime |
The interface response time of business time of origin piece |
Integer |
biz_errorTimes |
The interface error number of business time of origin piece |
Integer |
By the data cleansing of this step, the format data for finally giving is the metadata of above-mentioned table 1
Corresponding data set, business model is made up of the part necessity metadata in above-mentioned table 1, such as number of concurrent,
TPS, interface response time etc..
The treatment such as classified calculating, denoising is carried out to the data set after treatment, each business interface training is obtained
Data set, the corresponding data set of different business interface is respectively stored in different data files.For example,
After by classification, denoising, each business interface generates corresponding data file list, and form is such as
“BizType_SubBizType_ChargeInst_ChargeOffInst_Date.csv”.Metadata in individual data file is such as
Shown in table 2 below:
The data file list of table 2
Form:BizType_SubBizType_ChargeInst_ChargeOffInst_Date.csv |
HK_HK_CMB_CMB_20150728.csv |
JF_WATER_BJWATER_BJCEB_20150728.csv |
JF_ELECTRIC_WUHANELECTRIC_BJCEB_20150728.csv |
HK_HK_ABC_ABC_20150728.csv |
HK_HK_ICBC_ICBC_20150728.csv |
In above-mentioned table 2, wherein each data file is the corresponding number of treated each business model
According to, and used as the training dataset of decision Tree algorithms, the corresponding metadata of each data set is as shown in table 3:
Table 3 is used for the metadata of decision Tree algorithms
Field name |
Explanation of field |
Data type |
week |
Business time of origin:What day |
Integer |
hour |
Business time of origin:Hour |
Integer |
slice |
The timeslice of minute level |
Integer |
calledTimes |
Interface interchange number of times |
Integer |
rspPerCalled |
Timeslice inner joint response time average |
Float |
rspTime |
The interface response time in timeslice after removal noise |
Float |
volatility |
Coefficient of variation algorithm example:rspTime/3/rspPerCalled |
Float |
Exemplary, the one of data file in above-mentioned table 2 can be with as shown in table 4 below, can be with
See, each row of data in table 4 represents the statistical information of business in some timeslice, wherein time
Piece can be 15 minutes, or 1 minute, or even second level, depending on specific portfolio.
The data file of table 4
The performance modeling of business interface can be carried out according to performance data training set, the industry of business interface is obtained
Business performance model.For example, the foundation of the service feature model of this step, can be big data modeling
Perform.The service feature model W of Zhou Weidu can be set up by decision Tree algorithms;Or, over time
Change, it is according to all dimensional model W (W1/W2/...) and monthly using the output of cosine similarity algorithm
With reference to all dimensional model W', to cause Performance Evaluation more accurate.
In the embodiment of the present disclosure, can by the performance data that is used in the modeling process of above-mentioned performance model,
The platform testing system for incoming traffic platform is obtained to build some parameters of test environment, is said respectively
It is bright as follows:
Resource consumption analog parameter:Can be according to monitoring system (operation platform that monitoring business runs)
Monitoring Data is that service environment data are obtained, and is mainly used in realizing the playback of the resource consumption of inline system,
And the load simulation of the operating system aspect of business platform.
For example, in above-mentioned business model, platform resource consumption can include:Place when business occurs
Reason device CPU consumption, memory consumption when business occurs etc., and these platform resource consumption datas can be with
Data acquisition system including corresponding to multiple sampling time point respectively, such as, and the business time of origin in table 1
Piece CPU occupation rates, business time of origin piece memory usage etc..Can be surveyed in platform according to these data
The resource consumption of test system simulated testing system, simulates as the resource consumption of the system level of business platform.
Interface responds delay parameter:Can be obtained according to service operation data, be mainly used in realizing to outside
The simulation of the response time delay of mechanism, such as, when business platform receives service request and is handed over external agency
Mutually during treatment, generally can all there be certain treatment response time in external agency.For example, interface response time delay
Parameter, can be the interface response time in the rspTime- timeslices in table 4 after removal noise.Can be with
The interface delay of external system is simulated in platform testing system according to these data.
Service test data:The data can be obtained according to service operation data, be primarily referred to as business platform
Which kind of business received, for example, can include business time of origin, type of service, subservice type,
Enter an item of expenditure in the accounts mechanism, the mechanism that writes off etc..
For example, during the foundation of above-mentioned service feature model, having obtained many according to metadata
Business model, and these business models are classified, can be according to parameters such as mechanism, business, will be logical
Crossing the business model of same business interface has carried out sorting out set.Service test data construction can be foundation
The relatively low business model of performance data selection performance, to optimize the performance of business platform accordingly.Such as, exist
In each business model, there are the performance indications of the business interfaces such as number of concurrent, interface response time, can
Each business model is ranked up with according to these performance indications, is screened, choose sequence digit default
The business model of scope, such as business model of selection Top N, such as the business of CPU consumption Top3,
Or the business of CPU+ interface response times integrated ordered Top3;Business is constructed according to these business models
Test data, forms the service environment of platform test, and the service test data can include:Type of service,
The parameters such as subservice type, the mechanism that enters an item of expenditure in the accounts, the mechanism that writes off.Can for example be surveyed using orthogonal arrage finishing service
Try the construction of data.
Can be by the resource consumption analog parameter of above-mentioned acquisition, interface response delay parameter and operational trials number
According to, the platform testing system for testing business platform is input into, carry out performance test with to business platform.
By said process it can be seen that, the disclosure provide performance test methods such as Fig. 3 treatment:
In step 301, business platform service data is obtained, the business platform service data includes:
Service operation data and service environment data;By obtaining resource consumption simulation ginseng in the service environment data
Number, delay parameter and service test data are responded by the interface that platform service is obtained in service operation data.
In step 302, the resource consumption analog parameter, interface response delay parameter and business are surveyed
Examination data, as the input of the platform testing system for testing business platform performance, enter to business platform
Row performance test.
Wherein, performance test is being carried out to business platform, can be with output performance baseline curve, the performance base
Change of the line number according to the performance indications for weighing the business interface during each version.For example, business
Platform would generally Persisting exploitation, evolution according to demand, and the test mode of the disclosure can be using industry on line
Business data carry out the daily build of each version, iteration as the input data of test platform, lasting monitoring
Each version, the performance change trend in each stage.
Above-mentioned performance test methods, are modeled, and obtain use by the big data according to inline system
In the |input paramete of test system, the running environment on line can preferably be simulated, more accurately
Obtain test result;Also, service test data is also, according to data gained on line, to improve and currently connect
Business model coverage rate in mouth test, also more presses close to business on line;The basis that the method can continue
Data configuration and renewal test environment, obtain a dynamic performance change trend on line.
Fig. 4 provides a kind of performance testing device, and the device is used to realize the performance test side shown in Fig. 3
Method, as shown in figure 4, the device can include:Parameter acquisition module 41 and performance test module 42.
Parameter acquisition module 41, for obtaining business platform service data, the business platform service data
Including:Service operation data and service environment data;By obtaining resource consumption in the service environment data
Analog parameter, delay parameter and operational trials are responded by the interface that platform service is obtained in service operation data
Data;
Performance test module 42, for by the resource consumption analog parameter, interface response delay parameter and
Service test data, as the input of the platform testing system for testing business platform performance, to business
Platform carries out performance test.
Further, referring to Fig. 5, parameter acquisition module 41 can include:Classification submodule 411, row
Sequence submodule 412 and construction submodule 413.
Classification submodule 411, for obtaining the same business interface of correspondence in business platform service data
Multiple business models;
Sorting sub-module 412, for the performance data in the business platform service data, by institute
Multiple business models are stated to be ranked up;
Construction submodule 413, for choosing business model of the sequence digit in preset range, construction is described
Service test data.
Further, performance test module 42, are additionally operable to the performance base of the business interface of outgoing traffic platform
Line number evidence, the performance base-line data is used to weigh the performance indications of the business interface during each version
Change.
For example, the resource consumption analog parameter, including:The processor consumption resource, interior of business platform
Deposit consumption resource.But the resource consumption analog parameter includes but is not limited to above-mentioned parameter.
The performance testing device of the present embodiment, is modeled, and obtain by the big data according to inline system
The |input paramete in test system is taken, the running environment on line can preferably be simulated, it is more accurate
True obtains test result.
The preferred embodiment of the application is the foregoing is only, is not used to limit the application, it is all at this
Within the spirit and principle of application, any modification, equivalent substitution and improvements done etc. should be included in
Within the scope of the application protection.