CN101605144B - Web software system throughput optimization method - Google Patents
Web software system throughput optimization method Download PDFInfo
- Publication number
- CN101605144B CN101605144B CN 200910054349 CN200910054349A CN101605144B CN 101605144 B CN101605144 B CN 101605144B CN 200910054349 CN200910054349 CN 200910054349 CN 200910054349 A CN200910054349 A CN 200910054349A CN 101605144 B CN101605144 B CN 101605144B
- Authority
- CN
- China
- Prior art keywords
- throughput
- controller
- web software
- internal layer
- online number
- 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.)
- Expired - Fee Related
Links
Images
Landscapes
- Feedback Control In General (AREA)
Abstract
The invention discloses a Web software system throughput optimization method, applicable to optimization control of Web software system and belonging to the software development technical field. The invention solves the problems that server resource can not be fully utilized and throughput is not enough caused by improper online population control and service performance is suddenly dropped and even corrupted caused by too high load of the server when the Web software system is under high load condition. The invention collects operating data such as request quantity, average response time, service throughput and the like of the Web software system in the recent time at regular time by an optimization controller of the system, and a feedback control mechanism is utilized to dynamically adjust online population limit according to operating feedback information, so that the throughput of the server is further improved or is stabilized at a higher level. By adopting the invention, the service quality, stability and reliability of the Web software system are guaranteed, and the system service throughput is improved.
Description
Technical field
The present invention is applicable to the optimal control of Web software systems, belongs to the software development technique field, the dynamic optimization method of throughput when being specifically related to a kind of Web software systems operation.
Background technology
The Web software systems are as the system that software service is provided by network (local area network (LAN) or Internet); in running, often can run into the high load condition of a large number of users Concurrency Access; under high load condition; Web server bears very large access pressure; if can not effectively control this moment; the consumption of the server resources such as CPU, internal memory, database, network may sharply increase so, causes server performance to descend, serve the violent decline of throughput even server failing.The service throughput current with system the time online number and average response time the two is relevant, online number quantity increases can improve throughput, but will cause the response time sharply to descend after surpassing certain level, cause on the contrary rapid reduction even the server failing of throughput.Therefore, the high capacity Concurrency Access that may occur during for operation, the Web service system generally all can guarantee by online number control (i.e. temporary transient refusal login when simultaneously online number is too high) service quality of server performance and logged-in user.Yet arranging of online number restriction must be within zone of reasonableness, arranges too highly may cause that server can't bear the heavy load, handling property sharply descends, and crosses lowly may cause server resource to be fully used.The running environment of the dynamic change such as the dynamic availability of resource and access load is closely related when reasonably online number restriction and server performance, operation, is difficult to arrange simply a fixed value.Yet traditional online number control all is to set in advance maximum number of persons logging, perhaps by the system manager according to the system operation situation manual adjustments, thereby cause system under high load condition, often to be in " low saturated " or " supersaturation " state.Therefore, corresponding Optimal Control Problem can be summed up as: by the adaptive optimization of the online number restriction of server is regulated, make system bring into play its service ability under high load condition as far as possible, realize the maximization of service throughput.
Cybernetics is study of various system regulation and control rule science, in the cybernetics, has or not impact to be divided into closed-loop control (being FEEDBACK CONTROL) and open loop control to control method according to the output variable of system on the control action of system.The FEEDBACK CONTROL opinion is dynamically controlled system according to the output feedback of system, and the output valve of controlled device is tried one's best close to the desired value of system.FEEDBACK CONTROL can significantly reduce systematic error, improve control precision, reduce disturbance to the impact of system's output, the basic framework of FEEDBACK CONTROL is as shown in Figure 1: controller is controlled controlled device according to the difference of desired value and real output value, and purpose is to make the real output value of controlled device as far as possible close to desired value.In the last few years, cybernetics gradually studied person was incorporated in the software development, its objective is the optimal control that realizes software systems or software development process.
Summary of the invention
The optimal control method that the purpose of this invention is to provide a kind of online number of Web software systems based on when operation feedback, thereby the feedback information during according to operation is implemented in the self adaptation of informant's number restriction and regulates, and realizes the maximization of throughput under the Web software systems high load condition.
In order to achieve the above object, technical scheme of the present invention is as follows:
The present invention is by the timing monitoring to Web software systems operation conditions, then the acquisition system carries out dynamic adjustments according to these feedback informations to online number restriction by controller in these service datas such as the request quantity within (such as 1 minute), average response time, service throughput for the previous period.The adaptive control process that this monitoring, FEEDBACK CONTROL and optimization are regulated continues to carry out at Web software systems run duration.Core concept of the present invention is based on three characteristic features of Adaptable System, decision-making and dynamic the adjustment are implemented when monitoring when namely moving, operation, realize the optimal control of Web software system throughput, use simultaneously feedback realization adaptive control decision-making wherein.Because FEEDBACK CONTROL is take default desired value as the optimal control target, and optimal throughput, online number and the average response time in Web software systems when operation are among the dynamic change, therefore the present invention adopts two-layer feedback controller: outer controller is dynamically controlled the desired value of internal layer controller according to the goodput situation of change of system, and the internal layer controller then carries out dynamic adjustments according to the desired value that sets to online number restriction.
The inventive method specifically comprises the steps:
A. be that the Web software systems set an initial online number restriction, and so that this system under corresponding online number controlling mechanism, moving;
B. regularly collect the operation information of Web software systems within recently a period of time;
C. the feedback control mechanism in the optimal controller carries out dynamic adjustments according to described operation information to online number restriction, so that the throughput of server is further enhanced or is stabilized on the higher level;
D. repeated execution of steps b-d behind the certain hour of interval.
Wherein, described Web software system throughput optimization method step c adopts two-layer feedback controller to be implemented in the optimal control of informant's number restriction; Wherein, internal layer adopts the higher PID controller of precision, and is outer then adopt the lower simple controller of precision (Simple Controller); The internal layer feedback controller is regulated online number restriction as desired value with the average response time that sets, and outer feedback controller then carries out dynamic adjustments according to the goodput feedback of system to the desired value of internal layer controller.
Described internal layer feedback controller adopts the PID controller, and corresponding control algolithm as shown in Equation 1.Theoretical by numerical computations, can come the evaluation integration with Rectangular Method, replace differential with backward difference, thereby formula (1) is converted to formula (2).But the calculating of formula (2) is relevant with all states in system's past, therefore can further be converted to the increment type control function formula shown in formula (3), and this formula is only relevant with nearest 3 next states, need not add up, and is difficult for causing the accumulation of error.K in the formula
p, K
i, K
dThese 3 parameters (PID tuning) are set in advance by artificial, and their value can affect the level and smooth degree of Mean Time of Systemic Response curve, therefore can determine by the level and smooth degree of observing curve in actual applications the value of these parameters.
u(t)=Kp*e(t)+Ki*∑e(t)+Kd*(e(t)-e(t-1))..........(2)
u(t)=u(t-1)+Kp*(e(t)-e(t-1))+Ki*e(t)+Kd*(e(t)-2*e(t-1)+e(t-2))..........(3)
PID controller for internal layer, in the corresponding formula (3), the actual average response time that the expectation average response time that e (t) expression t calculates constantly and monitoring obtain poor, u (t) expression t control variables constantly, the pass of the correction value Δ (t) of the constantly online number restriction of this control variables and t is:
A(t)=limit(t-1)*u(t)/set_time(t)
Wherein set_time (t) represents current expectation average response time, and limit (t-1) then represented the online number restriction in a upper moment.The Δ that calculates accordingly (t) will be used to the adjustment (increase or reduce) of online number restriction, and purpose is to make the average response time of server be stabilized in desired value.
Described outer feedback controller adopts simple controller, according to the goodput feedback of system the desired value of internal layer controller is carried out dynamic adjustments, and purpose is to make system be in " saturated " state (high-throughput) of optimization under the high load capacity environment as far as possible.This simple controller adopts didactic strategy: if throughput rises, then increase the desired value of internal layer PID controller; If throughput descends, then reduce the desired value of internal layer PID controller.Each increase is different with the quantity that reduces, and purpose is the generation in order to prevent from shaking.
The present invention is owing to adopted above-mentioned technical scheme, make it compared with prior art, have following advantage and good effect: the optimization of the online number restriction of Web software systems of feedback was regulated when the inventive method can realize based on operation, thereby adapt to the dynamic change of the running environment such as Web server computational resource, Internet resources and access load, ensure service quality, stability and the reliability of Web software systems, improve the system service throughput.
Description of drawings
To embodiments of the invention and in conjunction with the description of its accompanying drawing, can further understand purpose of the present invention, specific structural features and advantage by following.Wherein, accompanying drawing is:
Fig. 1 is the basic framework schematic diagram of FEEDBACK CONTROL;
Fig. 2 is basic process schematic diagram of the present invention;
Fig. 3 is two-layer feedback controller structural representation of the present invention.
Embodiment
This section has provided a specific implementation based on JSP/Java and Tomcat.
As shown in Figure 2, the main use procedure based on this execution mode is:
(1) revise Tomcat configuration file server.xml, make its with the information recording/of all requests in the running log file of appointment.The form of every Web request record is " %h%t%r%s%D ", wherein %h represents the IP address of accessing, %t represents the access time, %r represents the mode (post or get) of accessing, the resource of access and the http protocol version of use, %s represents to access the http state that returns, and %D represents to process the used time of request;
(2) the Web software systems bring into operation as current online number restriction with a predefined value, and carry out login according to the number restriction and control (being that refusal was logined after online number surpassed restriction);
(3) optimal controller is analyzed the running log file, obtain the total number of request, total processing time (the %D addition of every record and get) within nearest 1 minute and the service number (the various operands that the user carries out after logining successfully) of finishing, and calculate the indexs such as average response time (total processing time/total number of request) and throughput (the service number of finishing);
(4) outer controller is regulated the expectation average response time of internal layer controller according to the situation of change (rise or descend) of goodput, and the internal layer controller is then regulated online number restriction according to desired value and actual average response time;
(5) the Web software systems are logined control according to the online number restriction after upgrading, if restriction improves then allows more logging in system by user, otherwise reduce the online user's quantity that allows;
(6) after 1 minute repeated execution of steps (3)-(6) at the interval.
As shown in Figure 3, step (4) adopts two-layer feedback controller to be implemented in the dynamic adjustments of informant's number restriction; Wherein, internal layer adopts the PID controller, the simple controller of outer employing; Described internal layer feedback controller is regulated online number restriction as desired value with the average response time that sets, and described outer feedback controller then carries out dynamic adjustments according to the goodput feedback of system to the desired value of internal layer controller; This simple controller adopts didactic strategy: if throughput rises, then increase the desired value of internal layer PID controller; If throughput descends, then reduce the desired value of internal layer PID controller.Each increase is different with the quantity that reduces, and purpose is the generation in order to prevent from shaking.
Claims (6)
1. a Web software system throughput optimization method is characterized in that comprising the steps:
A. be that the Web software systems set an initial online number restriction, and so that this system under corresponding online number controlling mechanism, moving;
B. regularly collect the Web software systems at the nearest operation information of T in the time;
C. the feedback control mechanism in the optimal controller carries out dynamic adjustments according to described operation information to online number restriction, so that the throughput of server is further enhanced or is stabilized on the higher level;
D. interval T repeated execution of steps b-d after the time.
2. method according to claim 1 is characterized in that, described operation information comprises average response time and service throughput.
3. method according to claim 1 and 2 is characterized in that, described step c adopts two-layer feedback controller to be implemented in the dynamic adjustments of informant's number restriction; Wherein, internal layer adopts the PID controller, the simple controller of outer employing; Described internal layer feedback controller is regulated online number restriction as desired value with the average response time that sets, and described outer feedback controller then carries out dynamic adjustments according to the goodput feedback of system to the desired value of internal layer controller.
4. method according to claim 3 is characterized in that, in the two-layer feedback controller that described step c adopts, the PID controller of internal layer adopts following FEEDBACK CONTROL formula:
u(t)=u(t-1)+K
p*(e(t)-e(t-1))+K
i*e(t)+K
d*(e(t)-2*e(t-1)+e(t-2))
Wherein, the actual average response time that the expectation average response time that e (t) expression t calculates constantly and monitoring obtain poor, u (t) expression t control variables constantly, wherein K
pThe coefficient of proportional component (P), K
iThe coefficient of integral element (I), K
dIt is the coefficient of differentiation element (D);
Described control variables u (t) with the pass of the correction value Δ (t) of the constantly online number restriction of t is:
Δ(t)=limit(t-1)*u(t)/set_time(t)
Wherein, set
_The expectation average response time that time (t) expression is current, limit (t-1) then represented the online number restriction in a upper moment, and the Δ that calculates accordingly (t) will be used to the adjustment of online number restriction.
5. method according to claim 3, it is characterized in that, in the two-layer feedback controller that described step c adopts, described outer feedback controller adopts the heuristic control strategy: if throughput of system rises, then increase the desired value of described internal layer feedback controller; If throughput descends, then reduce the desired value of internal layer feedback controller.
6. method according to claim 5 is characterized in that, the each quantity that increases or reduce of described desired value is different.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 200910054349 CN101605144B (en) | 2009-07-03 | 2009-07-03 | Web software system throughput optimization method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 200910054349 CN101605144B (en) | 2009-07-03 | 2009-07-03 | Web software system throughput optimization method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101605144A CN101605144A (en) | 2009-12-16 |
CN101605144B true CN101605144B (en) | 2013-04-17 |
Family
ID=41470698
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 200910054349 Expired - Fee Related CN101605144B (en) | 2009-07-03 | 2009-07-03 | Web software system throughput optimization method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101605144B (en) |
Families Citing this family (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101930371A (en) * | 2010-09-16 | 2010-12-29 | 复旦大学 | Software quality run-time optimizing method based on control theory and goal inference |
CN109783108A (en) * | 2019-01-24 | 2019-05-21 | 中国银行股份有限公司 | A kind of optimization method for software and device |
US20200341953A1 (en) * | 2019-04-29 | 2020-10-29 | EMC IP Holding Company LLC | Multi-node deduplication using hash assignment |
CN111447113B (en) * | 2020-03-25 | 2021-08-27 | 中国建设银行股份有限公司 | System monitoring method and device |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101387977A (en) * | 2008-10-30 | 2009-03-18 | 西安交通大学 | Server software regeneration method for maximizing task throughput |
-
2009
- 2009-07-03 CN CN 200910054349 patent/CN101605144B/en not_active Expired - Fee Related
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101387977A (en) * | 2008-10-30 | 2009-03-18 | 西安交通大学 | Server software regeneration method for maximizing task throughput |
Also Published As
Publication number | Publication date |
---|---|
CN101605144A (en) | 2009-12-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11184797B2 (en) | System and method for efficient use of network bandwidth | |
CN100555297C (en) | The method and apparatus of coordinating service performance and application placement management | |
US7644162B1 (en) | Resource entitlement control system | |
Stankovic et al. | Feedback control scheduling in distributed real-time systems | |
US8359290B1 (en) | Smart sync—refreshing application state based on user migration patterns | |
CN101605144B (en) | Web software system throughput optimization method | |
Strulik | Hyperbolic discounting and the time‐consistent solution of three canonical environmental problems | |
Sun et al. | Learning-based Adaptation Framework for Elastic Software Systems. | |
Ghezzi et al. | Performance‐driven dynamic service selection | |
Zhao et al. | Analysis of adaptive incentive protocols for P2P networks | |
Huang | System intelligence: Model, bounds and algorithms | |
Kang et al. | I/O-aware deadline miss ratio management in real-time embedded databases | |
CN112365099B (en) | Non-deterministic separation web server cluster telescoping method | |
Diao et al. | A business-oriented approach to the design of feedback loops for performance management | |
Tu et al. | Control-based quality adaptation in data stream management systems | |
Patikirikorala et al. | An evaluation of multi-model self-managing control schemes for adaptive performance management of software systems | |
CN104283934A (en) | WEB service pushing method and device based on reliability prediction and server | |
Kontarinis et al. | Cloud resource allocation from the user perspective: A bare-bones reinforcement learning approach | |
Bartolini et al. | Self-∗ through self-learning: Overload control for distributed web systems | |
US20230177117A1 (en) | Enhanced Reinforcement Learning Algorithms Using Future State Prediction | |
Lightstone et al. | Control theory: a foundational technique for self managing databases | |
da Silveira et al. | Wage inequality as a source of endogenous macroeconomic fluctuations | |
Chen et al. | Dynamic intelligent feedback scheduling in networked control systems | |
CN115087992B (en) | Deep causal learning for data storage and processing capacity management | |
Wu et al. | Optimal policies based on QoS for adaptive communication system with Markov Decision Process |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20130417 Termination date: 20170703 |