CN100383820C - A Quantification Method of Computer Performance Index Based on Fuzzy Logic - Google Patents
A Quantification Method of Computer Performance Index Based on Fuzzy LogicDownload PDFInfo
- Publication number
- CN100383820C CN100383820CCNB031375480ACN03137548ACN100383820CCN 100383820 CCN100383820 CCN 100383820CCN B031375480 ACNB031375480 ACN B031375480ACN 03137548 ACN03137548 ACN 03137548ACN 100383820 CCN100383820 CCN 100383820C
- Authority
- CN
- China
- Prior art keywords
- fuzzy
- output value
- load output
- control rule
- degree
- 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)
- Multi Processors (AREA)
Abstract
Translated fromChineseDescription
Translated fromChinese技术领域technical field
本发明属于计算机性能评估技术领域，特别是针对服务器的性能评估的性能指标量化方法。The invention belongs to the technical field of computer performance evaluation, in particular to a performance index quantification method for server performance evaluation.
背景技术Background technique
随着Internet/Intranet技术的迅速发展，许多公司、企业和网站建立了不少基于网络的应用服务系统，如商业网站、文件传输、视频点播等。由于服务具有相当的开放性，服务强度存在一定的波动，可能服务器在大部分时间能够正常的运行，但当遇到访问峰值的时候，容易发生服务器响应变满甚至服务中断，为了避免类似情况的发生，应及时对系统运行状态进行评估，以便在系统出现问题之前采取相应的保护措施，其中反映系统当前运行状态的主要参数包括：CPU利用率、内存利用率、磁盘带宽和网络带宽等。With the rapid development of Internet/Intranet technology, many companies, enterprises and websites have established many application service systems based on the network, such as commercial websites, file transfer, video on demand and so on. Due to the openness of the service, there are certain fluctuations in the service intensity. The server may be able to operate normally most of the time, but when encountering a peak access, it is easy for the server to respond to full or even service interruption. In order to avoid similar situations If it happens, the system operation status should be evaluated in time, so that corresponding protection measures can be taken before the system goes wrong. The main parameters reflecting the current system operation status include: CPU utilization, memory utilization, disk bandwidth and network bandwidth.
系统性能评价在具体系统应用中有很大的现实意义。在单机服务系统中，根据系统各状态参量的变化情况总结出系统负载的变化规律，并分析出系统瓶颈的所在，从而决定当前系统是否能满足当前的服务需求。在分布式系统中，管理员也需要随时了解当前各节点的性能状态，并对可能出现超载的节点提前采取一定的保护措施。尤其在分布式服务应用领域，服务节点的系统性能评价是至关重要的一个环节。只有调度器了解了各节点当前的性能状况，才能有效的进行负载的均衡，性能高的节点分得的任务较多，性能差的节点分得的任务较少。以往，对系统性能的评价很大程度上依赖与对系统配置的了解，缺乏一种即时的动态负载估算方法，而且，在系统配置各不相同时，比较不同节点之间性能更是存在些问题。System performance evaluation has great practical significance in specific system applications. In the stand-alone service system, according to the change of each state parameter of the system, the change law of the system load is summarized, and the bottleneck of the system is analyzed, so as to determine whether the current system can meet the current service demand. In a distributed system, administrators also need to know the current performance status of each node at any time, and take certain protective measures in advance for nodes that may be overloaded. Especially in the field of distributed service applications, the system performance evaluation of service nodes is a crucial link. Only when the scheduler understands the current performance status of each node can it effectively balance the load. Nodes with high performance are assigned more tasks, and nodes with poor performance are assigned fewer tasks. In the past, the evaluation of system performance largely relied on the understanding of system configuration, lacking a real-time dynamic load estimation method, and, when the system configuration is different, there are some problems in comparing the performance of different nodes .
现有的性能评价方法大多都是用经典数学的方法对计算机性能进行评价，如对各性能参数加权求和。但由于计算机性能和各状态参量之间并不是简单的线性关系，所以采用加权求和这种纯线形方法显然不能够准确的反映出计算机的真实性能，所以，这种方法虽然简单，但功能上有较大的局限。Most of the existing performance evaluation methods use classical mathematical methods to evaluate computer performance, such as weighted summation of each performance parameter. However, since there is not a simple linear relationship between the performance of the computer and each state parameter, the pure linear method of weighted summation obviously cannot accurately reflect the real performance of the computer. Therefore, although this method is simple, it is functionally There are larger limitations.
经典数学处理问题的非常重要的前提是讨论的范围必须是清晰的，也就是要求一个对象要么在这个范围内，要么不在这个范围内，不允许出现模棱两可的情况。但现实问题并非都是这样。例如， 描述CPU负载状况时，可以说当前CPU比较空闲，但空闲与繁忙两者是一个相对概念，并没有一个精确的阈值来区分它们。因此利用经典数学的方法对计算机性能进行评价不能满足实际应用的需求。The very important prerequisite for dealing with problems in classical mathematics is that the scope of discussion must be clear, that is, it is required that an object is either within this scope or not within this scope, and ambiguity is not allowed. But the real problem is not always like this. For example, when describing the CPU load status, it can be said that the current CPU is relatively idle, but idle and busy are relative concepts, and there is no precise threshold to distinguish them. Therefore, the use of classical mathematical methods to evaluate computer performance cannot meet the needs of practical applications.
发明内容Contents of the invention
本发明的目的是为了解决对计算机的性能进行评价的问题，提出一种基于模糊逻辑的计算机性能指标量化方法，可将原来模糊的不便比较的概念数字化，根据计算机的多个状态参量给出一个综合的量化指标，从而使得使用者对于计算机性能有个更为直观的了解，以便于分配任务。也便于不同计算机之间的性能进行比较，此方法尤为适用于分布式网络管理系统和集群调度系统。The purpose of the present invention is to solve the problem of evaluating the performance of the computer, and propose a computer performance index quantification method based on fuzzy logic, which can digitize the original fuzzy concept that is inconvenient to compare, and give a Comprehensive quantitative indicators, so that users have a more intuitive understanding of computer performance, so as to assign tasks. It is also convenient to compare the performance of different computers, and this method is especially suitable for distributed network management systems and cluster scheduling systems.
本发明提出的一种基于模糊逻辑的计算机性能指标量化方法，其特征在于，包括：A fuzzy logic-based computer performance index quantification method proposed by the present invention is characterized in that it comprises:
首先将采集到的多个计算机输入状态参量进行模糊化处理，并设置系统的一个负载输出值的量化范围及模糊化参数；First, fuzzify the collected multiple computer input state parameters, and set the quantization range and fuzzy parameters of a load output value of the system;
然后制定所述多个输入状态参量与一个负载输出值对应的模糊关系的模糊控制规则库；Then formulate a fuzzy control rule base of the fuzzy relationship corresponding to the plurality of input state parameters and a load output value;
根据每个采样时刻的输入状态参量具体数值，利用模糊控制规则库的每条规则进行模糊推理，得到该各控制规则对应的负载输出值和匹配程度；According to the specific value of the input state parameter at each sampling moment, use each rule of the fuzzy control rule base to perform fuzzy reasoning, and obtain the load output value and matching degree corresponding to each control rule;
最后再将该对应的负载输出值和匹配程度进行加权合并处理，得到一个最终量化的负载输出值作为性能的评价指标。Finally, the corresponding load output value and matching degree are weighted and merged to obtain a final quantized load output value as a performance evaluation index.
所述对多个计算机状态参量进行模糊化处理的方法，包括：The method for fuzzifying a plurality of computer state parameters includes:
确定参与计算机性能评价的输入状态参量(通常可以用于评价计算机性能的输入状态参量有CPU利用率、内存利用率、运行队列长度、磁盘带宽利用率等)；Determine the input state parameters that participate in computer performance evaluation (usually the input state parameters that can be used to evaluate computer performance include CPU utilization, memory utilization, run queue length, disk bandwidth utilization, etc.);
根据所选定的每个输入状态参量的状态数值大小划分为多个状态等级；例如，可根据CPU利用率这个状态参量的数值大小划分为三个状态等级：高、中、低；According to the state value of each selected input state parameter, it can be divided into multiple state levels; for example, it can be divided into three state levels according to the value of the state parameter of CPU utilization: high, medium, and low;
设置表示每个输入状态参量的具体数值对应该参量各状态等级的隶属程度的隶属度函数；隶属度用0～1之间的数字来表示。Set the membership degree function that indicates the specific value of each input state parameter corresponds to the degree of membership of each state level of the parameter; the degree of membership is represented by a number between 0 and 1.
所述设置系统负载输出值的量化范围可根据具体应用情况进行具体设定，例如可定义在0～100之间，0表示空载状态，100表示满载状态；The quantitative range of the set system load output value can be specifically set according to the specific application situation, for example, it can be defined between 0 and 100, 0 means no-load state, and 100 means full-load state;
所述模糊化参数可包括：将所述负载输出值的大小分为若干个级别，定义相应的状态的隶属度函数。The fuzzy parameters may include: dividing the magnitude of the load output value into several levels, and defining a membership function of a corresponding state.
所述模糊控制规则库是根据专家的实际经验知识归纳得到，其中，每条模糊控制规则为由所述多个输入状态参量的不同状态等级通过一定的与/或逻辑关系构成的条件，对应于所述一个负载输出值的相应状态等级，该控制规则可用″如果…则…″的条件语句表示。The fuzzy control rule library is obtained according to the actual experience knowledge of experts, wherein each fuzzy control rule is a condition formed by a certain and/or logical relationship between different state levels of the multiple input state parameters, corresponding to For the corresponding state level of the one load output value, the control rule can be expressed by a conditional statement of "if...then...".
所述模糊推理方法，可采用Mamdani模糊推理算法，包括：首先对一条模糊控制规则中的每个输入状态参量的隶属度函数得到该参量的隶属度，然后以隶属度最小的值为准，结合该模糊控制规则中所对应的输出参量的状态等级隶属度函数，求出该模糊控制规则的负载输出值及该值的匹配程度。Described fuzzy reasoning method, can adopt Mamdani fuzzy reasoning algorithm, comprise: first obtain the degree of membership of this parameter to the degree of membership function of each input state parameter in a fuzzy control rule, then take the value of degree of membership minimum as the criterion, combine The state grade membership function of the corresponding output parameter in the fuzzy control rule is used to obtain the load output value of the fuzzy control rule and the matching degree of the value.
所述加权合并处理方法为：将通过每条控制规则得到的匹配程度对其相应的负载输出值进行加权运算，得到一个最终量化的负载输出值，计算公式如下：
本发明的原理：Principle of the present invention:
本发明的参量模糊化的处理过程就是将原来精确的输入参量根据事先定义好的模糊集和隶属度函数进行模糊化处理，用状态和隶属关系来描述一个参量的状态，比如，将CPU利用率划分为繁忙和空闲两个状态，当前CPU利用率为30％，根据隶属度函数经过模糊化处理后，就可以描述为当前CPU利用率对繁忙状态的隶属程度为0.3，对空闲状态的隶属程度为0.8。The parameter fuzzification process of the present invention is to fuzzify the original accurate input parameters according to the pre-defined fuzzy set and membership function, and describe the state of a parameter with the state and the membership relationship, for example, the CPU utilization rate It is divided into two states: busy and idle. The current CPU utilization rate is 30%. According to the membership degree function, after fuzzy processing, it can be described as 0.3 for the current CPU utilization rate for the busy state and 0.3 for the idle state. is 0.8.
模糊控制规则是基于操作人员长期积累的控制经验和领域专家的有关知识，它是对被控对象进行控制的一个知识数据库，这个数据库建立的是否准确，即是否准确总结了成功的操作经验和领域专家的知识，将直接决定了模糊系统性能的好坏。Fuzzy control rules are based on the long-term accumulated control experience of operators and the relevant knowledge of domain experts. It is a knowledge database for controlling the controlled object. Whether the database is established accurately, that is, whether it accurately summarizes successful operating experience and domain Expert knowledge will directly determine the performance of the fuzzy system.
模糊推理方法是结合控制规则，按照预先选取的模糊推理算法进行推理。每条控制规则都会对输入参量相作用，然后给出一个输出值Oi，以及该规则和实际情况的匹配程度αi。本发明的模糊推理方法就是Mamdani模糊推理算法的一种具体应用。The fuzzy reasoning method is combined with the control rules, and the reasoning is carried out according to the pre-selected fuzzy reasoning algorithm. Each control rule will act on the input parameters, and then give an output value Oi , and the degree of matching between the rule and the actual situation αi . The fuzzy reasoning method of the present invention is a specific application of the Mamdani fuzzy reasoning algorithm.
关于Mamdani模糊推理算法：About the Mamdani fuzzy inference algorithm:
Mamdani模糊推理算法由Mamdani教授在1974年发明的，在很多文献中都有详细的介绍，该算法的实现过程主要包括：首先对一条模糊控制规则中的每个输入状态参量的隶属度函数得到该参量的隶属度，然后以隶属度最小的值为准，结合该模糊控制规则中所对应的输出参量的状态等级隶属度函数，求出该模糊控制规则的负载输出值及该值的匹配程度。The Mamdani fuzzy inference algorithm was invented by Professor Mamdani in 1974, and it has been introduced in detail in many documents. The implementation process of the algorithm mainly includes: first, obtain the membership function of each input state parameter in a fuzzy control rule. According to the membership degree of the parameter, the minimum value of the membership degree shall prevail, combined with the state level membership function of the corresponding output parameter in the fuzzy control rule, to obtain the load output value of the fuzzy control rule and the matching degree of the value.
现以双输入单输出的系统为例作一下说明，输入参量为X、Y，输出参量为R，假设输入参量实际值分别为x和y，待匹配的规则为(Xi，Yj)→Rk，以图形方式表示如图1所示。Let’s take a dual-input and single-output system as an example. The input parameters are X and Y, and the output parameter is R. Assume that the actual values of the input parameters are x and y respectively, and the matching rule is (Xi , Yj )→ Rk , graphically represented in Figure 1.
图中(a)、(b)分别为输入参量X在状态i时的隶属度函数曲线Xi，Y在状态j时的隶属度函数曲线Yj，横坐标为输入参量的值X、Y，纵坐标为该参量的隶属度μX、μY；(c)为输出参量R在状态k时的隶属度函数曲线，横坐标为输出参量的值R，纵坐标为输出参量的隶属度μR。根据隶属度函数曲线，可以分别求出x对Xi的隶属度，y对Yj的隶属度，以最小隶属度(即Yj的隶属度)为准向输出参量的隶属度函数图像作平行线，在输出参量的隶属度函数曲线内截得一个梯形(图2(c)中的灰色区域)，求出梯形的重心(图中的黑点)，重心的横坐标为根据该规则计算出的输出值Ok，重心的纵坐标为该规则的匹配程度αk。对每条规则都如此处理。(a) and (b) in the figure are respectively the membership function curve Xi of the input parameter X in state i, and the membership function curve Yj of Y in state j, and the abscissa is the value X and Y of the input parameter, The ordinate is the degree of membership μX , μY of the parameter; (c) is the function curve of the degree of membership of the output parameter R in state k, the abscissa is the value R of the output parameter, and the ordinate is the degree of membership of the output parameter μR . According to the membership function curve, the membership degree of x to Xi and the membership degree of y to Yj can be obtained respectively, and the minimum membership degree (that is, the membership degree of Yj ) is taken as the quasi-parallel to the membership function image of the output parameter Line, cut a trapezoid (the gray area in Figure 2(c)) in the membership function curve of the output parameter, and find the center of gravity of the trapezoid (the black point in the figure), and the abscissa of the center of gravity is calculated according to this rule The output value Ok of the center of gravity is the matching degree αk of the rule. Do this for each rule.
对上述每条规则推理得到的结果采用的加权方法进行合并处理，具体方法为：将通过每条控制规则得到的匹配程度对其相应的负载输出值进行加权运算，得到一个最终量化的负载输出值，计算公式如下：
本发明的优点：Advantages of the present invention:
1.参数选择灵活，算法具有很好的扩展性。用户可以自己选择需要观测的状态参量，甚至可以根据自己的测量自定义一些参数参与系统性能的评价。1. The parameter selection is flexible, and the algorithm has good scalability. Users can choose the state parameters that need to be observed, and even customize some parameters according to their own measurements to participate in the evaluation of system performance.
2.输出结果直观，用户无需了解计算机各参数的实际意义，只要根据最后的输出结果就可以判断系统当前的负载情况。2. The output result is intuitive, the user does not need to understand the actual meaning of each parameter of the computer, as long as the final output result can judge the current load situation of the system.
3.根据输出的量化后的负载指标，可以很容易的比较不同配置的计算机之间的性能差异，这在分布式调度系统中有较为积极的意义。3. According to the output quantified load index, the performance difference between computers with different configurations can be easily compared, which has more positive significance in the distributed scheduling system.
附图说明Description of drawings
图1为Mamdani模糊推理算法示意图。Figure 1 is a schematic diagram of Mamdani fuzzy reasoning algorithm.
图2为本发明方法的总体流程框图。Fig. 2 is an overall flowchart of the method of the present invention.
图3为本发明实施例的CPU利用率的隶属度函数曲线。FIG. 3 is a membership function curve of CPU utilization according to an embodiment of the present invention.
图4为本发明实施例的内存利用率的隶属度函数曲线。FIG. 4 is a membership function curve of memory utilization according to an embodiment of the present invention.
图5本发明实施例的负载指标的隶属度函数曲线。Fig. 5 is the membership function curve of the load index in the embodiment of the present invention.
图6本发明实施例的模糊推理过程示意图。Fig. 6 is a schematic diagram of the fuzzy reasoning process of the embodiment of the present invention.
具体实施方式Detailed ways
本发明提出的一种基于模糊逻辑的计算机性能指标量化方法结合附图及实施例详细说明如下：A kind of fuzzy logic-based computer performance index quantification method that the present invention proposes is described in detail as follows in conjunction with accompanying drawing and embodiment:
本发明方法如图2所示，包括：As shown in Figure 2, the inventive method comprises:
首先将采集到的多个计算机状态参量X1、X2…进行模糊化处理，Firstly, the collected computer state parameters X1, X2... are fuzzy processed,
然后制定所述多个输入状态参量与一个负载输出值对应的模糊关系的模糊控制规则库；Then formulate a fuzzy control rule base of the fuzzy relationship corresponding to the plurality of input state parameters and a load output value;
根据每个采样时刻的输入状态参量具体数值，利用模糊控制规则库的每条规则进行模糊推理，得到该各控制规则对应的负载输出值和匹配程度；According to the specific value of the input state parameter at each sampling moment, use each rule of the fuzzy control rule base to perform fuzzy reasoning, and obtain the load output value and matching degree corresponding to each control rule;
最后再将该对应的负载输出值和匹配程度进行加权合并处理，得到一个最终量化的负载输出值作为性能的评价指标Y。Finally, the corresponding load output value and matching degree are weighted and merged to obtain a final quantized load output value as the performance evaluation index Y.
上述方法的实施例详细说明如下：The embodiment of above-mentioned method is described in detail as follows:
本实施例是根据CPU利用率和物理内存利用率两个输入状态参量得到一个最终负载输出值作为系统性能量化评价指标。具体步骤包括：In this embodiment, a final load output value is obtained as a quantitative evaluation index of system performance according to two input state parameters of CPU utilization rate and physical memory utilization rate. Specific steps include:
一、对两个输入状态参量进行模糊化处理：本实施例将CPU利用率(用CPU表示该参量)和内存利用率(用MEM表示该参量)按照大小分为三个等级，高、中、低，即定义模糊集为{高，中，低}，并定义其隶属度函数，分别如图3、4所示。图中曲线L、M、H分别表示高中低三个状态的隶属度函数，横坐标为输入参量的实际值，纵坐标为隶属度，隶属度用0～1之间的数字来表示。输入参量每取一个具体的值就可以确定其对应高中低三个状态的隶属度。例如在本实施例中：当采集到CPU利用率具体值为35％，则其属于低等级的隶属度为0.83，属于中等级的隶属度为0，属于高等级的隶属度为0；内存利用率具体值为85％，则其属于低等级的隶属度为0，属于中等级的隶属度为0.17，属于高等级的隶属度为0.6。One, carry out fuzzy processing to two input state parameters: CPU utilization rate (represent this parameter with CPU) and memory utilization rate (represent this parameter with MEM) are divided into three grades according to size in this embodiment, high, medium, Low, that is, define the fuzzy set as {high, medium, low}, and define its membership function, as shown in Figure 3 and Figure 4 respectively. The curves L, M, and H in the figure represent the membership degree functions of the three states of high, middle and low respectively. The abscissa is the actual value of the input parameter, and the ordinate is the degree of membership. The degree of membership is represented by a number between 0 and 1. Every time the input parameter takes a specific value, its membership degree corresponding to the three states of high, middle and low can be determined. For example, in this embodiment: when the specific value of the CPU utilization rate is 35%, the membership degree belonging to the low level is 0.83, the membership degree belonging to the middle level is 0, and the membership degree belonging to the high level is 0; If the specific rate is 85%, the membership degree of the low grade is 0, the membership degree of the middle grade is 0.17, and the membership degree of the high grade is 0.6.
二、设置系统的一个负载输出值量化范围和模糊化参数：本实施例将该负载输出值的量化范围定义在0～100之间，0表示空载状态，100表示满载状态，并将系统负载输出值分为五个级别：低，较低，中，较高，高，并定义其隶属度函数如图5所示。曲线L、ML、M、MH、H分别表示低、较低、中、较高、高状态的隶属度函数。横坐标为输出负载值，纵坐标为隶属度。从图中可以确定，每个负载值对应的五个状态的隶属度。2. Set a quantization range and fuzzy parameter of a load output value of the system: In this embodiment, the quantization range of the load output value is defined between 0 and 100, where 0 represents the no-load state, 100 represents the full-load state, and the system load The output values are classified into five levels: low, low, medium, high, high, and define their membership functions as shown in Figure 5. Curves L, ML, M, MH, and H represent membership functions of low, low, medium, high, and high states, respectively. The abscissa is the output load value, and the ordinate is the degree of membership. It can be determined from the figure that each load value corresponds to the degree of membership of the five states.
三、制定模糊控制规则库：3. Formulate the fuzzy control rule base:
本实施例由九种输入和输出的模糊对应关系(每种对应关系为一条规则)组成的控制规则库，如表1所示。In this embodiment, the control rule base is composed of nine types of fuzzy correspondences between inputs and outputs (each correspondence is a rule), as shown in Table 1.
表1模糊控制规则库Table 1 Fuzzy control rule base
本实施例规则中的输入状态参量之间采用“与”逻辑关系；根据以上控制规则，系统当前CPU和内存的状态参量的不同组合情况对应系统负载输出的状态等级。比如表1中的一条规则表示为：“如果”CPU利用率为低级别L且MEM利用率为高级别H，“则”系统负载为中级别M。The input state parameters in the rules of this embodiment adopt an "AND" logical relationship; according to the above control rules, the different combinations of the current CPU and memory state parameters of the system correspond to the state levels of the system load output. For example, a rule in Table 1 is expressed as: "if" the CPU utilization is a low level L and the MEM utilization is a high level H, then "then" the system load is a medium level M.
四、模糊推理：本实施例中，CPU利用率为35％，内存利用率为85％，根据上述九条组成的规则库得到各规则对应的负载输出值；推理过程如图6所示，首先根据隶属度函数确定参量隶属度，图中第一列的竖线的横坐标为输入参量CPU利用率的具体值，即35％，它和各隶属度函数曲线的交点就是CPU利用率35％时对不同状态的隶属度，图中第二列的竖线的横坐标为输入参量内存利用率的具体制，即85％，它和各隶属度函数曲线的交点就是内存利用率85％时对不同状态的隶属度。图中第三列为输出参量在特定状态下的隶属度函数曲线。按照Mamdani模糊推理规则算法，图中每行的三个图为一组构成针对某一条规则对应的一个Mamdani推理过程，表1中的规则库有9条规则，这样每给出一组输入值就需要进行9次推理，每次推理都可以得到一个输出值Oi和匹配程度αi。在本实施例中，当CPU利用率为35％，内存利用率为85％时，图中只有第4行和第7行的推理图的梯形面积不为零，计算这两个梯形的重心，第4行推理图的梯形重心为(40，0.083)，即通过这条规则推理得到的输出值为40，匹配程度为0.083；第7行推理图的梯形重心为(60，0.26)，即通过这条推理规则得到的负载值为60，匹配程度为0.26，由其他规则推理得到的梯形面积为零，Four, fuzzy reasoning: in the present embodiment, CPU utilization rate is 35%, memory utilization rate is 85%, obtains the load output value corresponding to each rule according to the rule base that above-mentioned nine forms; Reasoning process as shown in Figure 6, at first according to The membership function determines the membership degree of the parameter. The abscissa of the vertical line in the first column in the figure is the specific value of the input parameter CPU utilization rate, that is, 35%. The degree of membership of different states, the abscissa of the vertical line in the second column in the figure is the specific limit of the memory utilization rate of the input parameter, that is, 85%. degree of membership. The third column in the figure is the membership function curve of the output parameter in a specific state. According to the algorithm of Mamdani fuzzy inference rules, the three graphs in each row in the figure constitute a Mamdani inference process corresponding to a certain rule. The rule base in Table 1 has 9 rules, so each set of input values is given Nine inferences are required, each inference can get an output value Oi and matching degree αi . In this embodiment, when the CPU utilization rate is 35% and the memory utilization rate is 85%, only the trapezoidal area of the reasoning diagram in the 4th line and the 7th line in the figure is not zero, and the center of gravity of these two trapezoids is calculated, The trapezoidal center of gravity of the inference diagram in line 4 is (40, 0.083), that is, the output value obtained by reasoning through this rule is 40, and the matching degree is 0.083; the trapezoidal center of gravity of the inference diagram in line 7 is (60, 0.26), that is, by The load value obtained by this inference rule is 60, and the matching degree is 0.26. The area of the trapezoid obtained by other rules is zero,
四、加权合并处理：根据匹配程度对输出值进行加权运算，计算过程如下：(0.083*40+0.26*60)/(0.083+0.26)≈55，即得到一个最终量化的负载输出值Load＝55作为性能的评价指标。4. Weighted merging processing: weighting the output value according to the matching degree, the calculation process is as follows: (0.083*40+0.26*60)/(0.083+0.26)≈55, that is, a final quantized load output value Load=55 is obtained as an indicator of performance.
Claims (6)
Translated fromChinesePriority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031375480ACN100383820C (en) | 2003-06-18 | 2003-06-18 | A Quantification Method of Computer Performance Index Based on Fuzzy Logic |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031375480ACN100383820C (en) | 2003-06-18 | 2003-06-18 | A Quantification Method of Computer Performance Index Based on Fuzzy Logic |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1460966A CN1460966A (en) | 2003-12-10 |
| CN100383820Ctrue CN100383820C (en) | 2008-04-23 |
Family
ID=29591330
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB031375480AExpired - Fee RelatedCN100383820C (en) | 2003-06-18 | 2003-06-18 | A Quantification Method of Computer Performance Index Based on Fuzzy Logic |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100383820C (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101187704B (en)* | 2007-12-17 | 2011-01-26 | 奇瑞汽车股份有限公司 | Reversing radar fuzzy controller |
| JP4897118B1 (en)* | 2011-01-25 | 2012-03-14 | 三菱電機株式会社 | Brake pressure calculation device, brake control system, and program |
| WO2013070226A1 (en)* | 2011-11-10 | 2013-05-16 | Intel Corporation | Computer system performance management with control variables, performance metrics and/or desirability functions |
| US9465374B2 (en) | 2011-11-10 | 2016-10-11 | Intel Corporation | Computer system performance management with control variables, performance metrics and/or desirability functions |
| CN105279560B (en)* | 2014-07-09 | 2017-11-10 | 先智云端数据股份有限公司 | Adaptive fuzzy rule control system for software defined storage system |
| CN110019110B (en)* | 2017-07-28 | 2022-11-18 | 腾讯科技（深圳）有限公司 | Capacity management method, device and equipment of service system and service system |
| CN110333992B (en)* | 2019-05-31 | 2022-07-15 | 平安科技（深圳）有限公司 | Service performance analysis method and device, computer equipment and storage medium |
| CN115099569A (en)* | 2022-05-20 | 2022-09-23 | 华南理工大学 | Passenger and taxi matching method based on fuzzy logic controller |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5475844A (en)* | 1992-11-27 | 1995-12-12 | Nec Corporation | Heavily loaded resource evaluation system |
| US5787409A (en)* | 1996-05-17 | 1998-07-28 | International Business Machines Corporation | Dynamic monitoring architecture |
| JPH10320245A (en)* | 1997-05-20 | 1998-12-04 | Nec Corp | Device for evaluating performance of computer |
- 2003
- 2003-06-18CNCNB031375480Apatent/CN100383820C/ennot_activeExpired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5475844A (en)* | 1992-11-27 | 1995-12-12 | Nec Corporation | Heavily loaded resource evaluation system |
| US5787409A (en)* | 1996-05-17 | 1998-07-28 | International Business Machines Corporation | Dynamic monitoring architecture |
| JPH10320245A (en)* | 1997-05-20 | 1998-12-04 | Nec Corp | Device for evaluating performance of computer |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1460966A (en) | 2003-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yager | Some aspects of intuitionistic fuzzy sets | |
| CN119862029B (en) | Edge computing resource allocation optimization method and system based on reinforcement learning | |
| Liu | An improved fuzzy time series forecasting method using trapezoidal fuzzy numbers | |
| CN106126328B (en) | A kind of traffic metadata management method and system based on event category | |
| CN100383820C (en) | A Quantification Method of Computer Performance Index Based on Fuzzy Logic | |
| CN119402440A (en) | A data analysis method, system, device and medium based on distributed communication | |
| CN105743985A (en) | Virtual service migration method based on fuzzy logic | |
| CN104834975A (en) | Power network load factor prediction method based on intelligent algorithm optimization combination | |
| CN110968952A (en) | Passenger traffic forecast method based on intelligent function combination optimization GM(1,1) model | |
| CN118798646A (en) | Regional production safety risk assessment method and system based on convolutional neural network | |
| Ouyang | Structural rule-based modeling with granular computing | |
| Mandal et al. | Multi-granulation fuzzy probabilistic rough sets and their corresponding three-way decisions over two universes | |
| CN113128880A (en) | Stability determination method and device for ecological system of drainage basin | |
| CN117093360A (en) | Computing load balancing method based on self-adaptive fuzzy control | |
| CN120434159A (en) | Method for scheduling computing power network resources | |
| CN119356824A (en) | A computing power scheduling strategy optimization system based on reinforcement learning | |
| CN106445680B (en) | Computing unit migration method based on comprehensive utility | |
| Xiong et al. | 2-tuple linguistic fuzzy ISM and its application | |
| Chang et al. | Cardinality-based fuzzy time series for forecasting enrollments | |
| CN114118543A (en) | Method and device for predicting oxygen content of flue gas based on joint learning | |
| Qu et al. | An improved method of fuzzy time series model | |
| CN115576641B (en) | A multi-objective optimization method and system for virtual machine service quality | |
| Chen et al. | Research on information transmission capability evaluation of information processing system based on fuzzy mathematics comprehensive evaluation method | |
| Hašková | An alternative approach for estimating GDP growth rate: Fuzzy prediction model | |
| CN119473538B (en) | A computing network task scheduling method based on improved three-branch decision-making and related devices |
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 | Granted publication date:20080423 Termination date:20140618 | |
| EXPY | Termination of patent right or utility model |