CN107301466A - To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship - Google Patents
To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship Download PDFInfo
- Publication number
- CN107301466A CN107301466A CN201610237159.4A CN201610237159A CN107301466A CN 107301466 A CN107301466 A CN 107301466A CN 201610237159 A CN201610237159 A CN 201610237159A CN 107301466 A CN107301466 A CN 107301466A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- time
- incidence relation
- formula
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
Abstract
The invention discloses a kind of to business load and resource distribution and the Forecasting Methodology and system of property relationship, including：Predictive content is determined, and original workload data is collected according to the predictive content；The original workload data is converted into the pattern of the input suitable for forecast model；According to the predictive content and the original workload data and pattern of the input that are collected into, the forecast model for being applicable the predictive content is set up, wherein, the forecast model is nonlinear model；Input variable is set, the input variable and the workload data are inputted into the forecast model and are predicted analysis, forecast analysis conclusion is drawn.
Description
Technical field
The present invention relates to business support field, more particularly to it is a kind of to the pre- of business load and resource distribution and property relationship
Survey method and forecasting system.
Background technology
Intranet (IT, Internet) system is the core component of generation, supply and management business data, IT system
The quality of performance concerns the ups and downs of enterprise, and the effective management of the performance progress to IT system and prediction are to ensure application performance
One vital task.
But in information-based overall situation, although information technology is obtained for unprecedented fast in the application of all trades and professions
Hail exhibition, but the IT of every profession and trade is put into based on system Construction and application and development, effective management to system and to future system
The prediction of system performance does not obtain enough attention but.And service generally have to and business change match it is perspective, otherwise
Easily there is the performance fault of system, and cause to produce harmful effect to business development.
Because business data is more and more huger, frequently task scheduling is related in treatment scale data procedures, it is multiple
The reasons such as miscellaneous Data Stream Processing, the performance of IT system becomes no longer to be easily mastered as before.And on the other hand, due to money
The finiteness in source, service provider always wants to reach highest CSAT using minimum software and hardware resources；So,
Carry out performance prediction to IT system and avoid risk also just to be particularly important.The core that performance to IT system is effectively predicted
The heart is the performance model for setting up IT system, one be capable of accurate description IT system performance model, the use for reducing system
Cost simultaneously ensures service feature, with important effect.
Existing system performance prediction mode, generally all based on simple linear patterns of change, when business load increase,
Central processing unit (CPU, Central Processing Unit) utilization rate, input and output (IO, Input Output) utilization rate
And service response time is all linearly increased or reduced with certain proportion；Think simultaneously, improving cpu performance or I O process energy
During the resource distributions such as power, the response time of system business can become faster；But actual conditions are then not quite similar, above-mentioned relation is not
Simple linear relationship, utilizes traditional Forecasting Methodology, it is impossible to meet prediction requirement of the people to existing system；For example due to business
Function be continuously increased and customer volume growth, how rationally whether existing system disclosure satisfy that business need can not judge,
Ground distributing system resource turns into the problem currently faced；Meanwhile, if to be reconfigured to existing system resource, such as carry
High cpu performance, there is much influences on the performance of operation system, and traditional prediction method can not also be judged, without it is particularly relevant according to
According to.
In summary, Classical forecast system is primarily present following deficiency：
1st, traditional prediction method is only applicable to solve the problems, such as simple linear, and complicated nonlinear change can not be opened up effectively
Show；
2nd, predictive content is simple, and availability is not strong, and forecasting system degree is not high；
3rd, prediction dimension is few, and many condition various dimensions can not be predicted；
4th, corresponding forecast model can not be set up according to actual conditions, prediction specific aim is not strong；
5th, prediction theory is single, and autgmentability is bad；The change produced to resource distribution and software upgrading is unpredictable.
The content of the invention
In order to solve the above technical problems, being closed the embodiments of the invention provide a kind of to business load and resource distribution with performance
The Forecasting Methodology and system of system.
It is provided in an embodiment of the present invention to business load and resource distribution and the Forecasting Methodology of property relationship, including：
Predictive content is determined, and original workload data is collected according to the predictive content；
The original workload data is converted into the pattern of the input suitable for forecast model；
According to the predictive content and the original workload data and pattern of the input that are collected into, set up and be applicable institute
The forecast model of predictive content is stated, wherein, the forecast model is nonlinear model；
Input variable is set, the input variable and the workload data are inputted into the forecast model progress
Forecast analysis, draws forecast analysis conclusion.
In the embodiment of the present invention, the foundation is applicable the forecast model of the predictive content, including：
The first incidence relation set up between utilization rate and following parameter：Service time, arrival rate, transaction processor number
Amount；
The first incidence relation set up between CPU response times and following parameter：Service time, utilization rate, issued transaction
Device quantity；
The first incidence relation set up between response time and following parameter：Service time, queuing time；
The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor quantity, service time,
Arrival rate；
The 5th incidence relation set up between queuing time and following parameter：ErlangC functions, service time, office
Manage device quantity, utilization rate.
In the embodiment of the present invention, the forecast model includes equation below：
U=(St λ)/M
Rt-cpu=St/ (1-UM)
Rt=St+Qt
Qt=EcSt/m (1-U)
Wherein, U represents utilization rate, the busy extent for characterizing server；St represents service time, characterizes a service
Time needed for the single affairs of device processing；λ represents arrival rate, characterizes the quantity of the affairs within a specified time reached；M or m generations
Table transaction processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes an affairs and spends in systems
All times；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；Ec represents ErlangC
Function, for calculating queuing time.
It is described to be predicted analysis in the embodiment of the present invention, including：
Response time is calculated according to first incidence relation, second incidence relation and the 3rd incidence relation；Or
Person,
Closed according to first incidence relation, the 4th incidence relation, the 5th incidence relation and the 3rd association
System calculates the response time.
It is described to be predicted analysis in the embodiment of the present invention, including：
Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；
Bring service time St and known input variable U and M into formula Rt-cpu=St/ (1-UM), calculate response
Time Rt；
Queuing time Qt is calculated according to formula Rt=St+Qt, wherein, Qt=Rt-St.
It is described to be predicted analysis in the embodiment of the present invention, including：
Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；
Bring service time St and known input variable m and λ into formula
In, calculate Ec functional values；
St, Ec, m, U are substituted into formula Qt=EcSt/m (1-U), queuing time Qt is calculated；
Response time Rt is calculated according to formula Rt=St+Qt.
In the embodiment of the present invention, methods described also includes：
The data drawn according to forecast analysis, generate chart；
Forecast analysis conclusion is obtained according to the icon and the forecast analysis conclusion is exported.
It is provided in an embodiment of the present invention to business load and resource distribution and the forecasting system of property relationship, including：
Data acquisition unit, original workload data is collected for determining predictive content, and according to the predictive content；
Data storage, for the original workload data to be converted into the pattern of the input suitable for forecast model,
And workload data is stored；
Prediction meanss are analyzed, for according to the predictive content and the original workload data being collected into and defeated
Entry format, sets up the forecast model for being applicable the predictive content, wherein, the forecast model is nonlinear model；Input is set
Variable, the input variable and the workload data are inputted into the forecast model and are predicted analysis, is drawn pre-
Survey analytical conclusions.
In the embodiment of the present invention, the analysis prediction meanss are additionally operable to first set up between utilization rate and following parameter
Incidence relation：Service time, arrival rate, transaction processor quantity；First set up between CPU response times and following parameter is closed
Connection relation：Service time, utilization rate, transaction processor quantity；First set up between response time and following parameter associates
System：Service time, queuing time；The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor number
Amount, service time, arrival rate；The 5th incidence relation set up between queuing time and following parameter：ErlangC functions, service
Time, transaction processor quantity, utilization rate.
In the embodiment of the present invention, the forecast model includes equation below：
U=(St λ)/M
Rt-cpu=St/ (1-UM)
Rt=St+Qt
Qt=EcSt/m (1-U)
Wherein, U represents utilization rate, the busy extent for characterizing server；St represents service time, characterizes a service
Time needed for the single affairs of device processing；λ represents arrival rate, characterizes the quantity of the affairs within a specified time reached；M or m generations
Table transaction processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes an affairs and spends in systems
All times；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；Ec represents ErlangC
Function, for calculating queuing time.
In the embodiment of the present invention, the analysis prediction meanss are additionally operable to be closed according to first incidence relation, described second
Connection relation and the 3rd incidence relation calculate the response time；Or, closed according to first incidence relation, the 4th association
System, the 5th incidence relation and the 3rd incidence relation calculate the response time.
In the embodiment of the present invention, the analysis prediction meanss are additionally operable to perform following process：According to formula U=(St λ)/M
Service time St is calculated, wherein, St=UM/ λ；Bring service time St and known input variable U and M into formula Rt-cpu=
St/(1-UM), calculate response time Rt；Queuing time Qt is calculated according to formula Rt=St+Qt, wherein, Qt=Rt-St.
In the embodiment of the present invention, the analysis prediction meanss are additionally operable to perform following process：According to formula U=(St λ)/M
Service time St is calculated, wherein, St=UM/ λ；Bring service time St and known input variable m and λ into formulaIn, calculate Ec functional values；St, Ec, m, U are substituted into
In formula Qt=EcSt/m (1-U), queuing time Qt is calculated；Response time Rt is calculated according to formula Rt=St+Qt.
In the embodiment of the present invention, the system also includes：Output device, it is raw for the data drawn according to forecast analysis
Into chart；Forecast analysis conclusion is obtained according to the icon and the forecast analysis conclusion is exported.
In the technical scheme of the embodiment of the present invention, predictive content is determined, and original work is collected according to the predictive content
Load data；The original workload data is converted into the pattern of the input suitable for forecast model；According in the prediction
The original workload data and pattern of the input held and be collected into, set up the forecast model for being applicable the predictive content；
Input variable is set, the input variable and the workload data are inputted into the forecast model and are predicted point
Analysis, draws forecast analysis conclusion.
By the implementation to technical scheme of the embodiment of the present invention, automation collection various dimensions workload data, is system
Prediction provides reliable data supporting；Traditional linear prediction method is abandoned, scientific and reasonable mathematical forecasting model is established, in advance
Survey ratio of precision traditional mode greatly improve, and can innovation realization IT system performance early warning, be easy to evade performance issue and failure
Risk；It can independently select to be suitable for the predictor formula of the forecast model, make prediction data accurate, be that follow-up system dilatation adjustment is carried
For important evidence, the software and hardware input and output ratio of IT system is set to reach reasonable level；It is suitable according to the selection of specific business demand
Prediction scene, rational input variable is set.
Brief description of the drawings
Fig. 1 is the embodiment of the present invention one to business load and resource distribution and the structure group of the forecasting system of property relationship
Into schematic diagram；
Fig. 2 shows for the flow to business load and resource distribution and the Forecasting Methodology of property relationship of the embodiment of the present invention two
It is intended to；
Fig. 3 shows for the flow to business load and resource distribution and the Forecasting Methodology of property relationship of the embodiment of the present invention two
It is intended to；
Fig. 4 is the calculating logic figure of the Rt response times of the embodiment of the present invention；
Fig. 5 changes the comparison diagram of ratio for the cpu busy percentage of the embodiment of the present invention between front and rear；
Change ratio comparison diagram between before and after the response time that Fig. 6 is the embodiment of the present invention；
Forecasting system generation CPU response times of the Fig. 7 for the embodiment of the present invention and the comparison diagram before arrival rate；
Fig. 8 is the embodiment of the present invention three to business load and resource distribution and the structure group of the forecasting system of property relationship
Into schematic diagram.
Embodiment
The characteristics of in order to more fully hereinafter understand the embodiment of the present invention and technology contents, below in conjunction with the accompanying drawings to this hair
The realization of bright embodiment is described in detail, appended accompanying drawing purposes of discussion only for reference, not for limiting the embodiment of the present invention.
The embodiment of the present invention is directed to prior art shortcoming, it is proposed that system performance prediction model, sets up corresponding analysis pre-
Examining system, prediction when business load increase to a certain degree when, whether systematic function can be affected, can be by which kind of degree
Influence；And which kind of degree forecast analysis load increases to, systematic function can produce obvious nonlinear change；Meanwhile, carrying
During high system resource configuration what has influence on system response time；Finally, the performance prediction result based on conditions present is formed.
The basic thought of the embodiment of the present invention is：Performance prediction is carried out for IT business system, analysis forecast model is set up
And system, according to business load change and system resource configuration change situation, performance prediction is carried out to the system business response time.
System needs acquisition system main frame and business index of correlation data as the input variable of forecasting system, by forecast model computing,
Which kind of change when the forecasting system response time produce, and is that business load increase and resource distribution are changed to systematicness
The influence of energy is provided and predicted the outcome.
With Classical forecast systematic comparison, the forecasting system of the embodiment of the present invention is different from traditional linear relationship and calculates mould
Type, can predict the non-linear behaviour problem that conventional linear prediction mode can not be shown, also be greatly improved in terms of accuracy.
Fig. 1 is the embodiment of the present invention one to business load and resource distribution and the structure group of the forecasting system of property relationship
Into schematic diagram, as shown in figure 1, the forecasting system includes：Data acquisition unit, data storage and analysis prediction meanss composition.
First, data acquisition unit is disposed in the operation system of prediction to be analyzed, and is controlled by analysis prediction meanss,
It is responsible for the time point of control collection, and the indication information and operational state of mainframe gathered etc., number is stored data into after collection
According in memory；When data meet forecast analysis requirement, analysis prediction meanss read related data from data storage, together
When, the corresponding prediction input variable of curriculum offering predicted as needed；Finally, according to forecast model in analysis prediction meanss
Measuring and calculating analysis, draws corresponding forecast analysis data, chart and prediction conclusion, above-mentioned conclusion can for the follow-up business growth of load and
Influence of the resource distribution change to performance provides related foundation and suggestion.
The analysis prediction meanss of foundation the content predicted is needed between load and resource distribution and performance to exist certain
Relation, and performance here is often referred to the ability that CPU unit interval interior energy handles affairs, so rational predictor formula need to be set up
And system model.
Fig. 2 shows for the flow to business load and resource distribution and the Forecasting Methodology of property relationship of the embodiment of the present invention two
It is intended to, as shown in Fig. 2 the flow includes：
1st, predictive content is determined.Here, each single item prediction work, no matter it is complicated or simple, it is all based on one substantially
This problem is answered in the problem of type or needs, prediction.
2nd, collection work load data.Here, according to forecast demand, gathered data is determined, and collect, store, use this
A little data.
The 3rd, workload data is described.Here, the original workload data of collection needs, and is converted into being applied to prediction mould
The pattern of the input of type.
4th, suitable forecast model is formulated.Here, according to predictive content and obtainable workload data and description
Option formulates the forecast model for being applicable the predictive content.
5th, it is predicted analysis.Here, it should be understood that IT prediction service risks before predicting, and related countermeasure is formulated,
To mitigate the harm that risk and reduction service level are brought.
6th, prediction conclusion is drawn.Here, the data drawn according to forecast analysis, generate correlation graph, and it is accurate to finally obtain
Forecast analysis conclusion, provide foundation for the system decision-making.
Fig. 3 shows for the flow to business load and resource distribution and the Forecasting Methodology of property relationship of the embodiment of the present invention two
It is intended to, as shown in figure 3, the Forecasting Methodology to business load and resource distribution and property relationship comprises the following steps：
Step 301：Predictive content is determined, and original workload data is collected according to the predictive content.
In the embodiment of the present invention, data acquisition unit determines predictive content, and collects original work according to the predictive content
Load data.
Step 302：The original workload data is converted into the pattern of the input suitable for forecast model.
In the embodiment of the present invention, the original workload data is converted into suitable for forecast model by data storage
Pattern of the input, and workload data is stored.
Step 303：According to the predictive content and the original workload data and pattern of the input that are collected into, build
The vertical forecast model for being applicable the predictive content.
Here, the forecast model is nonlinear model.
The foundation is applicable the forecast model of the predictive content, including：
The first incidence relation set up between utilization rate and following parameter：Service time, arrival rate, transaction processor number
Amount；
The first incidence relation set up between CPU response times and following parameter：Service time, utilization rate, issued transaction
Device quantity；
The first incidence relation set up between response time and following parameter：Service time, queuing time；
The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor quantity, service time,
Arrival rate；
The 5th incidence relation set up between queuing time and following parameter：ErlangC functions, service time, office
Manage device quantity, utilization rate.
In the embodiment of the present invention, the forecast model that analysis prediction meanss are set up is characterized with reference to IT system key and actually made
Formed with summary of experience.The particular content of forecast model is as follows：
U=(St λ)/M (1)
Rt-cpu=St/ (1-UM) (2)
Rt=St+Qt (3)
Qt=EcSt/m (1-U) (5)
Wherein, formula (1) is represented：Utilization rate=service time × arrival rate/number of transactions.
Formula (2) is represented：CPU response times=service time/(the M powers of 1- utilization rates).
Formula (3) is represented：Response time=service time+queuing time.
Formula (4) is represented：ErlangC functions, for calculating queuing time.
Formula (5) is summarized on ErlangC functional foundations to be formed, and calculates queuing time.
In above-mentioned formula, formula (2) is the algorithm that the first calculates the response time, and formula (3) is second of calculating response
The algorithm of time.
The implication of above-mentioned variable is as follows：U represents utilization rate, the busy extent for characterizing server；When St represents service
Between, characterize the time needed for the single affairs of server process；λ represents arrival rate, characterizes the thing within a specified time reached
The quantity of business；M or m represent transaction processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes a thing
It is engaged in all times spent in systems；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；
Ec represents ErlangC functions, for calculating queuing time.
Step 304：Input variable is set, the input variable and the workload data are inputted to the prediction mould
Analysis is predicted in type, forecast analysis conclusion is drawn.
In the embodiment of the present invention, forecasting system is reasonably using above formula to the workload data collected and setting
Input variable calculated, draw correlation analysis data, chart and conclusion, wherein response time Rt calculating logics schematic diagram is joined
According to Fig. 4.It is described to be predicted analysis, including：According to first incidence relation, second incidence relation and the 3rd association
Relation calculates the response time；Or, according to first incidence relation, the 4th incidence relation, the 5th incidence relation
And the 3rd incidence relation calculate the response time.
Specifically, the built-in above-mentioned algorithm of forecasting system, according to different prediction scenes, user can reasonably Selection utilization
Algorithm above content is calculated the data collected and the input variable of setting, draws correlation analysis data, chart and knot
By.Logic chart shown in Fig. 4 defines the response time Rt of system two kinds of computational methods, and Rt is that an affairs are spent in systems
All times taken, this " time in system " is commonly known as the response time.Two kinds of computational methods are specific as follows：
First method：Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；By service time
St and known input variable U and M bring formula R intot-cpu=St/ (1-UM), calculate response time Rt；According to formula Rt=
St+Qt calculates queuing time Qt, wherein, Qt=Rt-St.
Specifically, using fundamental forecasting formula Rt-cpu=St/ (1-UM) response time Rt is predicted.1st, service is calculated
Time St；Firstly the need of service time St is calculated, by logic chart it can be seen that service time St=UM/ λ, U are utilization rate,
Refer to the busy extent of server；M (or m) is transaction processor quantity；λ refers to arrival rate, i.e., the affairs reached within preset time
Quantity, also can be regarded as system load situation, thus formula can calculate the service time St of system；When the 2nd, calculating response
Between Rt；By service time St and known input variable U, M brings fundamental forecasting formula R into togethert-cpu=St/ (1-UM) counted
Calculate, calculate response time Rt value；3rd, queuing time Qt is calculated；Queuing time Qt is derived from using formula Rt=St+Qt
=Rt-St.
Second method：Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；By service time
St and known input variable m and λ bring formula intoIn, meter
Calculation draws Ec functional values；St, Ec, m, U are substituted into formula Qt=EcSt/m (1-U), queuing time Qt is calculated；According to public affairs
Formula Rt=St+Qt calculates response time Rt.
Specifically, using ErlangC functions, the prediction to the response time is improved, new queuing time calculation formula is utilized
Qt is calculated, Rt calculation formula Rt=St+Qt is reused and is predicted；It is made up of two parts, service time St and queuing
Time Qt, computational methods adding up and comprising the following steps that for both：1st, service time St is calculated；Service time St calculating side
Method in method one using formula S t=UM/ λ with unanimously, being calculated；2nd, Ec functional values are calculated；Ec functions calculate three needed
Input variable is m, St and λ；St has been calculated in the first step, and m and λ are the known input variable in forecasting system, generation
Enter to calculate Ec values in function；3rd, queuing time Qt is calculated；Because the calculation formula Qt=EcSt/m (1-U) of Qt in logic chart can
Know, it is necessary to tetra- variate-values of Ec, St, m and U are calculated, St and Ec values are calculated in being walked above two, by itself and
Two variate-values of m, U are substituted into formula together calculates queuing time Qt predicted values；4th, response time Rt is calculated；By by St
Read group total is carried out with Qt values, response time Rt predicted value is finally drawn.
Finally, the data drawn according to forecast analysis, generate chart；Forecast analysis conclusion is obtained and defeated according to the icon
Go out the forecast analysis conclusion.
With reference to concrete application scene to the embodiment of the present invention to business load and resource distribution and property relationship
Forecasting Methodology is described in further detail.
Predict case
At present, mobile Internet turns into the new trend of IT industry developments, and wechat is used as the client work that people all know
Tool, increasing people uses, and how to be provided the user using it with good service and facility, already as numerous enterprises not
The problem of disconnected thinking.
For telecom operators, wechat is also playing the value of service user.Sichuan movement is directed to visitor always
Family service lifting, is the pioneer of mobile Internet marketing service.Early in September, 2013, this most popular exchange in wechat
" Sichuan movement wechat business hall " is just opened in platform, Sichuan movement, causes the concern of vast wechat user.Yet with business
Function be continuously increased and customer volume growth, how rationally whether existing system disclosure satisfy that business need can not judge,
Ground distributing system resource turns into the problem currently faced；Thus, arise at the historic moment, we invent set analysis prediction meanss, to working as
During business load increase, come forecasting system performance will occur how to change；Meanwhile, if to be weighed to existing system resource
During new configuration, cpu performance is such as improved, there are much influences to the performance of operation system.
The data provided according to business department, they predict that within a following season " Sichuan movement wechat business hall " is used
Amount will increase by 500, and this means that portfolio i.e. system load may increase by 500；In this case, " move in Sichuan
Whether the performance of wechat business hall " system can be affectedWhich kind of it can be influenceed by degreeFor such case, me is utilized
Existing analysis prediction meanss analysis is predicted to " Sichuan move wechat business hall " system, analyze which kind of journey load increases to
Degree, systematic function can produce significant change, and influence during raising resource distribution to system response time, and related content is specific
It is as follows：
Collection work load data
First, using data acquisition unit in prediction meanss is analyzed, " Sichuan movement wechat business hall " system work is collected negative
Carry data and describe it, inputted as the part of system prediction, be that next step prediction is prepared.
Workload (i.e. arrival rate λ) is represented using main frame " user calls " (user call)
Cpu busy percentage (U) sampled data comes from main frame NMON data
Sample collection time range (2015/06/14---2015/06/18)
Specimen sample number totally 26 groups, time point：9,10,11,14,15,16 corresponding hours points.
| Sample | Sample time | Arrival rate (λ) | Busy degree (U) | Service time ms (St) | CPU response times ms (Rt) | Queuing time ms |
| 1 | 2015/06/14 09 | 13.58 | 0.2139 | 2.016108 | 2.016108 | 0 |
| 2 | 2015/06/14 10 | 14.65 | 0.2216 | 1.936326 | 1.936326 | 0 |
| 3 | 2015/06/14 11 | 13.76 | 0.2138 | 1.988130 | 1.988130 | 0 |
| 4 | 2015/06/14 14 | 14.77 | 0.2151 | 1.864075 | 1.864075 | 0 |
| 5 | 2015/06/14 15 | 15.49 | 0.2410 | 1.991673 | 1.991673 | 0 |
| 6 | 2015/06/14 16 | 13.73 | 0.2172 | 2.024347 | 2.024347 | 0 |
| 7 | 2015/06/15 09 | 14.33 | 0.2340 | 2.090869 | 2.090869 | 0 |
| 8 | 2015/06/15 10 | 14.81 | 0.2072 | 1.790892 | 1.790892 | 0 |
| 9 | 2015/06/15 11 | 13.98 | 0.1960 | 1.794892 | 1.794892 | 0 |
| 10 | 2015/06/15 14 | 14.67 | 0.2058 | 1.795109 | 1.795109 | 0 |
| 11 | 2015/06/15 15 | 14.94 | 0.2243 | 1.921745 | 1.921745 | 0 |
| 12 | 2015/06/15 16 | 14.10 | 0.2050 | 1.860454 | 1.860454 | 0 |
| 13 | 2015/06/16 09 | 12.76 | 0.2098 | 2.104471 | 2.104471 | 0 |
| 14 | 2015/06/16 10 | 16.85 | 0.2297 | 1.744854 | 1.744854 | 0 |
| 15 | 2015/06/16 11 | 13.46 | 0.1950 | 1.854230 | 1.854230 | 0 |
| 16 | 2015/06/16 14 | 14.34 | 0.2079 | 1.855840 | 1.855840 | 0 |
| 17 | 2015/06/16 15 | 14.49 | 0.2329 | 2.058026 | 2.058026 | 0 |
| 18 | 2015/06/16 16 | 13.37 | 0.2123 | 2.033164 | 2.033164 | 0 |
| 19 | 2015/06/17 09 | 13.37 | 0.1916 | 1.834469 | 1.834469 | 0 |
| 20 | 2015/06/17 10 | 14.15 | 0.2048 | 1.852645 | 1.852645 | 0 |
| 21 | 2015/06/17 11 | 13.00 | 0.1908 | 1.878693 | 1.878693 | 0 |
| 22 | 2015/06/17 14 | 13.72 | 0.2132 | 1.989700 | 1.989700 | 0 |
| 23 | 2015/06/17 15 | 14.93 | 0.2370 | 2.032090 | 2.032090 | 0 |
| 24 | 2015/06/17 16 | 13..0 | 0.2173 | 2.060731 | 2.060731 | 0 |
| 25 | 2015/06/18 09 | 12.23 | 0.1993 | 2.077436 | 2.077436 | 0 |
| 26 | 2015/06/18 10 | 14.43 | 0.2342 | 2.077258 | 2.077258 | 0 |
Table 1
Table 1 is workload chart, and workload chart institute display data is " Sichuan movement wechat business hall " system at present
The known input variable of system, suitably using these input variables, reliable basis for forecasting is provided for next step system prediction.
System prediction
1st, application load keeps constant, lifts cpu performance, cpu busy percentage and response time change.
2nd, application load is increased to which kind of degree, the rate of change of CPU response times produces significant change.
Predict scene one
Application load keeps constant, in the case of lifting cpu performance 10%, prediction cpu busy percentage and response time.
Prediction logic figure in Fig. 4, using providing fundamental forecasting method and correlation predictive formula in first method
It is predicted calculating, it can be deduced that such as the data of table 2 below:
Table 2
According to the prediction data generated in forecasting system, after cpu performance improves 10%, the anaplasia before and after cpu busy percentage
The comparison diagram of change ratio is as shown in Figure 5.Similarly, according to prediction data, after cpu performance raising 10% is improved, before the response time
Change ratio comparison diagram between afterwards as shown in Figure 6.
The result calculated according to forecasting system, front and rear comparative analysis can draw following prediction conclusion：
1st, application load keeps constant, in the case of lifting cpu performance 10%, and the average down ratio 5% of cpu busy percentage is left
It is right.
2nd, application load keeps constant, and in the case of lifting cpu performance 10%, it is left that the response time averagely reduces ratio 5%
It is right.
Predict scene two
Increase application load to which kind of degree, the rate of change of CPU response times produces significant change.
The data collected according to workload table in table 1, system-computed show that weighted average arrival rate is 14.068trx/
Ms (only represents unit, User call represent " affairs " number here, referred in fact " arrival rate ") in this trx.
Weighted formula
Forecasting system logic chart according to Fig. 4, using the computational methods of utilization rate U in second method, to single
Load is calculated, and obtained CPU average utilizations are 21.42%.
Thus baseline service time St is：
Although forecasting system is preferable in prediction utilization rate using fundamental forecasting formula, when they often underestimate response
Between.More precisely, they underestimate queuing time.It means that it will cause user to believe that system may be handled than reality more
Many work.By using the computational methods of second step ErlangC functions in second method in forecasting system logic chart, we
The prediction to the response time can be improved.By three variable m-cpu quantity of ErlangC functions, St service times, λ q queues are reached
Rate is updated in forecasting system input variable, is calculated：
Ec(m, St, λq)=Ec(128,1.9498,14.098)=0.00000
On ErlangC functional foundations, calculating Qt using new queuing time function is：
The baseline response time, Rt was：
Rt=St+Qt=1.9498ms/trx
According to three input variable arrival rate (λ), cpu quantity (128), service time (St), using in forecasting system
ErlangC functions calculate Erlang C values, Qt are calculated according to the 3rd step in method in logic chart two, finally using pre-
When response time Rt calculation formula Rt=St+Qt calculates CPU responses in the 4th step in method two in examining system logic chart
Between, while setting input variable arrival rate to be increased with 10% in forecasting system, draw following prediction chart：
| Arrival rate increase | Arrival rate (λ) | Cpu busy percentage | Service time | Response time | Response time changes | erlangc | Qt (queuing time) |
| 0% | 14.068 | 0.2142 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 10% | 15.475 | 0.2356 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 20% | 16.882 | 0.2570 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 30% | 18.288 | 0.2785 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 40% | 19.695 | 0.2999 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 50% | 21.102 | 0.3213 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 60% | 22.509 | 0.3427 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 70% | 23.916 | 0.3641 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 80% | 25.322 | 0.3856 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 90% | 26.729 | 0.4070 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 100% | 28.136 | 0.4284 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 110% | 29.543 | 0.4498 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 120% | 30.950 | 0.4712 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 130% | 32.356 | 0.4927 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 140% | 33.763 | 0.5141 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000000 | 0.0000000000 |
| 150% | 35.170 | 0.5355 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000001 | 0.0000000000 |
| 160% | 36.577 | 0.5569 | 1.9489 | 1.94890 | 0.0000000000 | 0.0000000010 | 0.0000000000 |
| 170% | 37.984 | 0.5783 | 1.9489 | 1.94890 | 0.0000000001 | 0.0000000083 | 0.0000000001 |
| 180% | 39.390 | 0.5998 | 1.9489 | 1.94890 | 0.0000000004 | 0.0000000595 | 0.0000000004 |
| 190% | 40.797 | 0.6212 | 1.9489 | 1.94890 | 0.0000000021 | 0.0000003618 | 0.0000000021 |
| 200% | 42.204 | 0.6426 | 1.9489 | 1.94890 | 0.0000000103 | 0.0000018953 | 0.0000000103 |
| 210% | 43.611 | 0.6640 | 1.9489 | 1.94890 | 0.0000000442 | 0.0000086432 | 0.0000000442 |
| 220% | 45.018 | 0.6854 | 1.9489 | 1.94890 | 0.0000001659 | 0.0000346374 | 0.0000001659 |
| 230% | 46.424 | 0.7069 | 1.9489 | 1.94890 | 0.0000005488 | 0.0001229514 | 0.0000005488 |
| 240% | 47.831 | 0.7283 | 1.9489 | 1.94890 | 0.0000016153 | 0.0003904496 | 0.0000016153 |
| 250% | 49.238 | 0.7497 | 1.9489 | 1.94890 | 0.0000042560 | 0.0011167709 | 0.0000042560 |
| 260% | 50.645 | 0.7711 | 1.9489 | 1.94891 | 0.0000100991 | 0.0028979870 | 0.0000100991 |
| 270% | 52.052 | 0.7925 | 1.9489 | 1.94892 | 0.0000217028 | 0.0068707262 | 0.0000217028 |
| 280% | 53.458 | 0.8140 | 1.9489 | 1.94894 | 0.0000424219 | 0.0149762771 | 0.0000424219 |
| 290% | 54.865 | 0.8354 | 1.9489 | 1.94898 | 0.0000758096 | 0.0302455792 | 0.0000758096 |
| 300% | 56.272 | 0.8568 | 1.9489 | 1.94902 | 0.0001241281 | 0.0569308098 | 0.0001241281 |
| 310% | 57.679 | 0.8782 | 1.9489 | 1.94909 | 0.0001863698 | 0.1005124669 | 0.0001863698 |
| 320% | 59.086 | 0.8996 | 1.9489 | 1.94916 | 0.0002559154 | 0.1674773887 | 0.0002559154 |
| 330% | 60.492 | 0.9211 | 1.9489 | 1.94922 | 0.0003183205 | 0.2648425974 | 0.0003183205 |
| 340% | 61.899 | 0.9425 | 1.9489 | 1.94925 | 0.0003502839 | 0.3999646909 | 0.0003502839 |
| 350% | 63.306 | 0.9639 | 1.9489 | 1.94922 | 0.0003186049 | 0.5796498372 | 0.0003186049 |
| 370% | 66.120 | 1.0067 | 1.9489 | 1.94879 | -0.0001124881 | 1.0961426540 | -0.0001124881 |
| 380% | 67.526 | 1.0282 | 1.9489 | 1.94828 | -0.0006181612 | 1.4417484926 | -0.0006181612 |
| 390% | 68.933 | 1.0496 | 1.9489 | 1.94750 | -0.0013966671 | 1.8501493857 | -0.0013966671 |
| 400% | 70.340 | 1.0710 | 1.9489 | 1.94639 | -0.0025111266 | 2.3229008325 | -0.0025111266 |
Table 3
Input variable and the prediction data drawn according to known in table 3, forecasting system generation CPU response times and arrival rate
Comparison diagram before, as shown in fig. 7, the CPU response times generated according to forecasting system show with load arrival rate tendency chart, is somebody's turn to do
Figure abscissa is that the variable quantity of load is ladder-type change with 10%, and the longitudinal axis is the response time of each affairs, it can be seen that one
Denier load exceedes to (when i.e. current regular traffic loads 2.9 times), the rate of change of response time is turned when 290% or so
Point, response time value changes are clearly.
Predict the outcome：
The prediction data drawn according to analysis prediction meanss and the CPU response times of generation and loading trends figure, when " four
When river movement wechat business hall " system load is exceeded to 290% or so, there is flex point in the change rate curve of response time,
Change clearly, will cause customer experience to decline, service quality is greatly reduced；If there is 5 times of numbers of users in following one month
Increase, existing system will can not meet business need completely；Meanwhile, it need to such as keep existing " Sichuan movement wechat business hall " performance
Level, using improving cpu performance or increasing the expansion method of CPU quantity, to meet the increase of future services load, it is necessary to increase
The substantially CPU of (5-2.9) × (10/5)=4.2 times quantity or performance, to ensure that " Sichuan movement wechat business hall " system is stable
Effec-tive Function.
By the forecasting system to operation system carry out analytical performance prediction, can be very good prediction business load change and
The influence to operation system performance is changed in resource distribution, and prediction data is clear, and prediction chart is very clear, and prediction conclusion is accurate,
The service operation risk being likely to occur for future provides reference frame, and related counter-measure is made in advance, and safeguards system is good for
Health, stably, efficient operation.
Fig. 8 is the embodiment of the present invention three to business load and resource distribution and the structure group of the forecasting system of property relationship
Into schematic diagram, as shown in figure 8, the forecasting system to business load and resource distribution and property relationship includes：
Data acquisition unit 81, original workload data is collected for determining predictive content, and according to the predictive content；
Data storage 82, for the original workload data to be converted into the input lattice suitable for forecast model
Formula, and workload data is stored；
Analyze prediction meanss 83, for according to the predictive content and the original workload data being collected into and
Pattern of the input, sets up the forecast model for being applicable the predictive content, wherein, the forecast model is nonlinear model；Set defeated
Enter variable, the input variable and the workload data are inputted into the forecast model and are predicted analysis, is drawn
Forecast analysis conclusion.
The analysis prediction meanss 83, are additionally operable to set up the first incidence relation between utilization rate and following parameter：Service
Time, arrival rate, transaction processor quantity；The first incidence relation set up between CPU response times and following parameter：During service
Between, utilization rate, transaction processor quantity；The first incidence relation set up between response time and following parameter：Service time, row
Team's time；The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor quantity, service time, arrive
Up to rate；The 5th incidence relation set up between queuing time and following parameter：ErlangC functions, service time, transaction processor
Quantity, utilization rate.
The forecast model includes equation below：
U=(St λ)/M
Rt-cpu=St/ (1-UM)
Rt=St+Qt
Qt=EcSt/m (1-U)
Wherein, U represents utilization rate, the busy extent for characterizing server；St represents service time, characterizes a service
Time needed for the single affairs of device processing；λ represents arrival rate, characterizes the quantity of the affairs within a specified time reached；M or m generations
Table transaction processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes an affairs and spends in systems
All times；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；Ec represents ErlangC
Function, for calculating queuing time.
The analysis prediction meanss 83, are additionally operable to according to first incidence relation, second incidence relation and the
Three incidence relations calculate the response time；Or, closed according to first incidence relation, the 4th incidence relation, the described 5th
Connection relation and the 3rd incidence relation calculate the response time.
The analysis prediction meanss 83, are additionally operable to perform following process：Service time is calculated according to formula U=(St λ)/M
St, wherein, St=UM/ λ；Bring service time St and known input variable U and M into formula Rt-cpu=St/ (1-UM), calculate
Draw response time Rt；Queuing time Qt is calculated according to formula Rt=St+Qt, wherein, Qt=Rt-St.
The analysis prediction meanss 83, are additionally operable to perform following process：Service time St is calculated according to formula U=(St λ)/M,
Wherein, St=UM/ λ；Bring service time St and known input variable m and λ into formula
In, calculate Ec functional values；St, Ec, m, U are substituted into formula Qt=EcSt/m (1-U), queuing time Qt is calculated；
Response time Rt is calculated according to formula Rt=St+Qt.
The system also includes：Output device 84, for the data drawn according to forecast analysis, generates chart；According to institute
Icon is stated to obtain forecast analysis conclusion and export the forecast analysis conclusion.
It will be appreciated by those skilled in the art that shown in Fig. 8 to business load and resource distribution and the prediction of property relationship
Each device in system realizes that function can refer to the foregoing Forecasting Methodology to business load and resource distribution and property relationship
Associated description and understand.
, can be in any combination in the case where not conflicting between technical scheme described in the embodiment of the present invention.
In several embodiments provided by the present invention, it should be understood that disclosed method and smart machine, Ke Yitong
Other modes are crossed to realize.Apparatus embodiments described above are only schematical, for example, the division of the unit, only
Only a kind of division of logic function, can have other dividing mode, such as when actually realizing：Multiple units or component can be tied
Close, or be desirably integrated into another system, or some features can be ignored, or do not perform.In addition, shown or discussed each group
Into part coupling each other or direct-coupling or communication connection can be by some interfaces, equipment or unit it is indirect
Coupling is communicated to connect, and can be electrical, machinery or other forms.
The above-mentioned unit illustrated as separating component can be or may not be it is physically separate, it is aobvious as unit
The part shown can be or may not be physical location, you can positioned at a place, can also be distributed to multiple network lists
In member；Part or all of unit therein can be selected to realize the purpose of this embodiment scheme according to the actual needs.
In addition, each functional unit in various embodiments of the present invention can be fully integrated into a second processing unit,
Can also be each unit individually as a unit, can also two or more units it is integrated in a unit；
Above-mentioned integrated unit can both be realized in the form of hardware, it would however also be possible to employ hardware adds the form of SFU software functional unit real
It is existing.
The foregoing is only a specific embodiment of the invention, but protection scope of the present invention is not limited thereto, any
Those familiar with the art the invention discloses technical scope in, change or replacement can be readily occurred in, should all be contained
Cover within protection scope of the present invention.
Claims (14)
1. it is a kind of to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that methods described includes：
Predictive content is determined, and original workload data is collected according to the predictive content；
The original workload data is converted into the pattern of the input suitable for forecast model；
According to the predictive content and the original workload data and pattern of the input that are collected into, set up applicable described pre-
The forecast model of content is surveyed, wherein, the forecast model is nonlinear model；
Input variable is set, the input variable and the workload data are inputted into the forecast model and are predicted
Analysis, draws forecast analysis conclusion.
2. it is according to claim 1 to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that
The foundation is applicable the forecast model of the predictive content, including：
The first incidence relation set up between utilization rate and following parameter：Service time, arrival rate, transaction processor quantity；
The first incidence relation set up between CPU response times and following parameter：Service time, utilization rate, transaction processor number
Amount；
The first incidence relation set up between response time and following parameter：Service time, queuing time；
The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor quantity, service time, arrival
Rate；
The 5th incidence relation set up between queuing time and following parameter：ErlangC functions, service time, transaction processor
Quantity, utilization rate.
3. it is according to claim 2 to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that
The forecast model includes equation below：
U=(St λ)/M
Rt-cpu=St/ (1-UM)
Rt=St+Qt
<mrow>
<msub>
<mi>E</mi>
<mi>c</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>m</mi>
<mo>,</mo>
<msub>
<mi>S</mi>
<mi>t</mi>
</msub>
<mo>,</mo>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>m</mi>
</msup>
<mrow>
<mi>m</mi>
<mo>!</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<mo>)</mo>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>k</mi>
</msup>
<mrow>
<mi>k</mi>
<mo>!</mo>
</mrow>
</mfrac>
<mo>+</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>k</mi>
</msup>
<mrow>
<mi>m</mi>
<mo>!</mo>
</mrow>
</mfrac>
</mrow>
</mfrac>
</mrow>
Qt=EcSt/m (1-U)
Wherein, U represents utilization rate, the busy extent for characterizing server；St represents service time, characterizes at a server
Manage the time needed for single affairs；λ represents arrival rate, characterizes the quantity of the affairs within a specified time reached；M or m represent thing
Business processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes the institute that an affairs are spent in systems
There is the time；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；Ec represents ErlangC functions,
For calculating queuing time.
4. it is according to claim 2 to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that
It is described to be predicted analysis, including：
Response time is calculated according to first incidence relation, second incidence relation and the 3rd incidence relation；Or,
According to first incidence relation, the 4th incidence relation, the 5th incidence relation and the 3rd incidence relation meter
Calculate the response time.
5. it is according to claim 3 to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that
It is described to be predicted analysis, including：
Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；
Bring service time St and known input variable U and M into formula Rt-cpu=St/ (1-UM), calculate the response time
Rt；
Queuing time Qt is calculated according to formula Rt=St+Qt, wherein, Qt=Rt-St.
6. it is according to claim 3 to business load and resource distribution and the Forecasting Methodology of property relationship, it is characterised in that
It is described to be predicted analysis, including：
Service time St is calculated according to formula U=(St λ)/M, wherein, St=UM/ λ；
Bring service time St and known input variable m and λ into formula
In, calculate Ec functional values；
St, Ec, m, U are substituted into formula Qt=EcSt/m (1-U), queuing time Qt is calculated；
Response time Rt is calculated according to formula Rt=St+Qt.
7. according to any one of claim 1 to 6 to business load and resource distribution and the Forecasting Methodology of property relationship, its
It is characterised by, methods described also includes：
The data drawn according to forecast analysis, generate chart；
Forecast analysis conclusion is obtained according to the icon and the forecast analysis conclusion is exported.
8. it is a kind of to business load and resource distribution and the forecasting system of property relationship, it is characterised in that the system includes：
Data acquisition unit, original workload data is collected for determining predictive content, and according to the predictive content；
Data storage, for the original workload data to be converted into the pattern of the input suitable for forecast model, and it is right
Workload data is stored；
Prediction meanss are analyzed, for according to the predictive content and the original workload data being collected into and input lattice
Formula, sets up the forecast model for being applicable the predictive content, wherein, the forecast model is nonlinear model；Input variable is set,
The input variable and the workload data are inputted into the forecast model and are predicted analysis, forecast analysis is drawn
Conclusion.
9. it is according to claim 8 to business load and resource distribution and the forecasting system of property relationship, it is characterised in that
The analysis prediction meanss, are additionally operable to set up the first incidence relation between utilization rate and following parameter：Service time, arrival
Rate, transaction processor quantity；The first incidence relation set up between CPU response times and following parameter：Service time, utilization
Rate, transaction processor quantity；The first incidence relation set up between response time and following parameter：Service time, queuing time；
The 4th incidence relation set up between ErlangC functions and following parameter：Transaction processor quantity, service time, arrival rate；Build
Vertical the 5th incidence relation between queuing time and following parameter：ErlangC functions, service time, transaction processor quantity, profit
With rate.
10. according to claim 9 to business load and resource distribution and the forecasting system of property relationship, its feature exists
In the forecast model includes equation below：
U=(St λ)/M
Rt-cpu=St/ (1-UM)
Rt=St+Qt
<mrow>
<msub>
<mi>E</mi>
<mi>c</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>m</mi>
<mo>,</mo>
<msub>
<mi>S</mi>
<mi>t</mi>
</msub>
<mo>,</mo>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>m</mi>
</msup>
<mrow>
<mi>m</mi>
<mo>!</mo>
</mrow>
</mfrac>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>-</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<mo>)</mo>
<msubsup>
<mo>&Sigma;</mo>
<mrow>
<mi>k</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>m</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msubsup>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>k</mi>
</msup>
<mrow>
<mi>k</mi>
<mo>!</mo>
</mrow>
</mfrac>
<mo>+</mo>
<mfrac>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>mS</mi>
<mi>t</mi>
</msub>
<msub>
<mi>&lambda;</mi>
<mi>q</mi>
</msub>
<mo>)</mo>
</mrow>
<mi>k</mi>
</msup>
<mrow>
<mi>m</mi>
<mo>!</mo>
</mrow>
</mfrac>
</mrow>
</mfrac>
</mrow>
Qt=EcSt/m (1-U)
Wherein, U represents utilization rate, the busy extent for characterizing server；St represents service time, characterizes at a server
Manage the time needed for single affairs；λ represents arrival rate, characterizes the quantity of the affairs within a specified time reached；M or m represent thing
Business processor quantity；Rt-cpuRepresent the CPU response times；Rt represents the response time, characterizes the institute that an affairs are spent in systems
There is the time；Qt represents queuing time, characterizes the stand-by period in queue before affairs start to process；Ec represents ErlangC functions,
For calculating queuing time.
11. according to claim 9 to business load and resource distribution and the forecasting system of property relationship, its feature exists
In the analysis prediction meanss are additionally operable to be closed according to first incidence relation, second incidence relation and the 3rd association
System calculates the response time；Or, according to first incidence relation, the 4th incidence relation, the 5th incidence relation with
And the 3rd incidence relation calculate the response time.
12. according to claim 10 to business load and resource distribution and the forecasting system of property relationship, its feature exists
In the analysis prediction meanss are additionally operable to perform following process：Service time St is calculated according to formula U=(St λ)/M, wherein,
St=UM/ λ；Bring service time St and known input variable U and M into formula Rt-cpu=St/ (1-UM), calculate response
Time Rt；Queuing time Qt is calculated according to formula Rt=St+Qt, wherein, Qt=Rt-St.
13. according to claim 10 to business load and resource distribution and the forecasting system of property relationship, its feature exists
In the analysis prediction meanss are additionally operable to perform following process：Service time St is calculated according to formula U=(St λ)/M, wherein,
St=UM/ λ；Bring service time St and known input variable m and λ into formula
In, calculate Ec functional values；St, Ec, m, U are substituted into formula Qt=EcSt/m (1-U), queuing time Qt is calculated；
Response time Rt is calculated according to formula Rt=St+Qt.
14. according to any one of claim 8 to 13 to business load and resource distribution and the forecasting system of property relationship,
Characterized in that, the system also includes：Output device, for the data drawn according to forecast analysis, generates chart；According to institute
Icon is stated to obtain forecast analysis conclusion and export the forecast analysis conclusion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610237159.4A CN107301466A (en) | 2016-04-15 | 2016-04-15 | To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610237159.4A CN107301466A (en) | 2016-04-15 | 2016-04-15 | To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107301466A true CN107301466A (en) | 2017-10-27 |
Family
ID=60136883
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610237159.4A Pending CN107301466A (en) | 2016-04-15 | 2016-04-15 | To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107301466A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726090A (en) * | 2017-10-31 | 2019-05-07 | 慧与发展有限责任合伙企业 | Performance influences the identification of defect in computing system |
| CN109976900A (en) * | 2017-12-28 | 2019-07-05 | 北京京东尚科信息技术有限公司 | The method and apparatus for calling service |
| CN111079991A (en) * | 2019-11-29 | 2020-04-28 | 珠海随变科技有限公司 | Service index prediction method, device, equipment and storage medium |
| CN111275268A (en) * | 2020-02-27 | 2020-06-12 | 中国联合网络通信集团有限公司 | Pricing process efficiency prediction method, device, equipment and storage medium |
| CN112055380A (en) * | 2019-06-06 | 2020-12-08 | 华为技术有限公司 | Method and apparatus for predicting traffic volume |
| CN112988703A (en) * | 2019-12-18 | 2021-06-18 | 中国移动通信集团四川有限公司 | Read-write request balancing method and device |
| CN113626285A (en) * | 2021-07-30 | 2021-11-09 | 平安普惠企业管理有限公司 | Model-based job monitoring method and device, computer equipment and storage medium |
| TWI828917B (en) * | 2019-06-10 | 2024-01-11 | 南韓商三星電子股份有限公司 | Systems and methods for managing input output queue pairs |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101330544A (en) * | 2008-06-17 | 2008-12-24 | 中国移动通信集团江苏有限公司 | Method for implementing multi-center virtual integration of passenger service system |
| CN101374076A (en) * | 2008-10-06 | 2009-02-25 | 中国移动通信集团设计院有限公司 | Method for forecasting and planning server handling ability of telecom business support system |
| CN103475672A (en) * | 2013-09-30 | 2013-12-25 | 南京大学 | Method for setting firewall with lowest cost of cloud computing platform |
| CN103903070A (en) * | 2014-04-15 | 2014-07-02 | 广东电网公司信息中心 | Resource demand measuring and calculating system for application system |
| US20150169240A1 (en) * | 2013-12-18 | 2015-06-18 | Fujitsu Limited | Storage controller, control method, and computer product |
-
2016
- 2016-04-15 CN CN201610237159.4A patent/CN107301466A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101330544A (en) * | 2008-06-17 | 2008-12-24 | 中国移动通信集团江苏有限公司 | Method for implementing multi-center virtual integration of passenger service system |
| CN101374076A (en) * | 2008-10-06 | 2009-02-25 | 中国移动通信集团设计院有限公司 | Method for forecasting and planning server handling ability of telecom business support system |
| CN103475672A (en) * | 2013-09-30 | 2013-12-25 | 南京大学 | Method for setting firewall with lowest cost of cloud computing platform |
| US20150169240A1 (en) * | 2013-12-18 | 2015-06-18 | Fujitsu Limited | Storage controller, control method, and computer product |
| CN103903070A (en) * | 2014-04-15 | 2014-07-02 | 广东电网公司信息中心 | Resource demand measuring and calculating system for application system |
Non-Patent Citations (2)
| Title |
|---|
| 朱世华等: "《现代数字交换与通信网》", 31 August 2009, 西安交通大学出版社 * |
| 殷文浩: "数据库性能预测的研究与应用", 《计算机时代》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109726090A (en) * | 2017-10-31 | 2019-05-07 | 慧与发展有限责任合伙企业 | Performance influences the identification of defect in computing system |
| CN109976900A (en) * | 2017-12-28 | 2019-07-05 | 北京京东尚科信息技术有限公司 | The method and apparatus for calling service |
| CN109976900B (en) * | 2017-12-28 | 2021-07-30 | 北京京东尚科信息技术有限公司 | Method and device for calling service |
| CN112055380A (en) * | 2019-06-06 | 2020-12-08 | 华为技术有限公司 | Method and apparatus for predicting traffic volume |
| CN112055380B (en) * | 2019-06-06 | 2022-04-29 | 华为技术有限公司 | Method and apparatus for predicting traffic volume |
| TWI828917B (en) * | 2019-06-10 | 2024-01-11 | 南韓商三星電子股份有限公司 | Systems and methods for managing input output queue pairs |
| CN111079991A (en) * | 2019-11-29 | 2020-04-28 | 珠海随变科技有限公司 | Service index prediction method, device, equipment and storage medium |
| CN112988703A (en) * | 2019-12-18 | 2021-06-18 | 中国移动通信集团四川有限公司 | Read-write request balancing method and device |
| CN112988703B (en) * | 2019-12-18 | 2022-09-16 | 中国移动通信集团四川有限公司 | Read-write request balancing method and device |
| CN111275268A (en) * | 2020-02-27 | 2020-06-12 | 中国联合网络通信集团有限公司 | Pricing process efficiency prediction method, device, equipment and storage medium |
| CN113626285A (en) * | 2021-07-30 | 2021-11-09 | 平安普惠企业管理有限公司 | Model-based job monitoring method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107301466A (en) | To business load and resource distribution and the Forecasting Methodology and forecasting system of property relationship | |
| Cheng et al. | Estimate at completion for construction projects using evolutionary support vector machine inference model | |
| Lam et al. | Understanding the effect of the learning–forgetting phenomenon to duration of projects construction | |
| Liu et al. | Parallel chance-constrained dynamic programming for cascade hydropower system operation | |
| CN106899660A (en) | Cloud data center energy-saving distribution implementation method based on trundle gray forecast model | |
| Ibrahim et al. | Inter-dependent, heterogeneous, and time-varying service-time distributions in call centers | |
| Simutis et al. | Optimization of cash management for ATM network | |
| CN105243609A (en) | Orderly power utilization decision method of user on the basis of fuzzy analytic hierarchy process | |
| CN114037311B (en) | Information system engineering supervision project risk assessment method | |
| Zhou et al. | Forecast load impact from demand response resources | |
| Miao et al. | Two uncertain chance-constrained programming models to setting target levels of design attributes in quality function deployment | |
| Wang et al. | Comparison of representative heuristic algorithms for multi-objective reservoir optimal operation | |
| Kumar et al. | Cost optimization of an unreliable server queue with two stage service process under hybrid vacation policy | |
| Banihashemi et al. | Evaluating efficiency in construction projects with the TOPSIS model and NDEA method considering environmental effects and undesirable data | |
| JP5953207B2 (en) | Demand plan management system | |
| Laughery | Computer simulation as a tool for studying human-centered systems | |
| CN105023100A (en) | Database and middleware non-index quantitative management platform for platform software | |
| Fernández-Montes et al. | Evaluating decision-making performance in a grid-computing environment using DEA | |
| Antoniol et al. | A queue theory-based approach to staff software maintenance centers | |
| Khalili et al. | Optimal assignment of human resources for maintenance departments using fuzzy queuing systems | |
| Fajar et al. | Optimization of time and cost of multi organization business processes in a port container terminal | |
| De Filippo et al. | Non-linear optimization of business models in the electricity market | |
| Ibrahim et al. | Managing large service systems with self-scheduling agents | |
| Gao et al. | Analysis of an enterprise human resource management performance evaluation model based on the DEA method | |
| CN101021869A (en) | Data value quantitative estimating method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171027 |