One side is a consulting company preaching in project management training: "CMM2-level enterprises are not suitable for implementing 6Sigma, and after the CMM4 level, the metrics is relatively perfect." One side is at the World Software Engineering Conference in 2004, and national experts reached a consensus. : "CMM / CMMI and 6Sigma can be combined, promoted to each other." what should we do? I used to advocate: the dispute is temporarily putting on one side, and the time should be improved, otherwise the result is probably: "We are progressing, but we have more backwards with the competitors." Some colleagues accepted my opinion, so there is one Q: "Do you have a successful case in the software to implement 6Sigma?" I haven't yet, but now I have several typical cases, and they have a characteristic, let us share this one.
1. Can 6Sigma help solve the software technical problem?
The first project was seen in the end of last year, and participated in the 6Sigma Excellent Project Release of a Division. The project name is "XX network management system to improve the alarm throughput", the problem is that the alarm processing throughput rate of the UNIX server is only 8 / sec on a large number of alarms, and the CPU is 90%, causing the operation of other modules. get on. Users are very dissatisfied, requiring our company to solve this problem as soon as possible, increase the throughput to at least 48 / sec, and the system cost cannot have a large increase. how to solve this problem? A solution is to improve hardware configuration, thereby increasing processing performance, but this will greatly increase procurement cost, and performance does not have great improvements, in fact, reducing the user's leader, such input income ratio Not compuent, this program is rejected. The project team spent a lot of time and effort, still looking for 6Sigma after being able to find a suitable solution. Everyone knows that the choice of the 6Sigma project is the "difficulty, influential" problem, so members of this project group will be completed, and 6Sigma can bring dawn in the dark.
In addition to the definition and measurement phase, the analysis idea of this project is this: first is the brainstorming fish bone chart, and all the possible reasons for everyone can think of; then these reasons are organized into FMEA according to the alarm logic process process, and RPN analysis, A few factors having a RPN value greater than 100 are screened as a potential key factor; then test these potential factors one by one, confirmed. The breakthrough of the entire project appears in the first factor test, and the test data is shown in FIG. 1, the abscissa indicates the input alarm traffic, and the ordinate represents the alarm processing delay. The curve in the figure shows a periodic inflection point, and after the inflection point, the alarm traffic increases, and the server's processing delay is more reduced. This phenomenon does not have the trial for this reason, without which data is unable to see. Analysis of this phenomenon, it is difficult to get our software engineer, soon concluded: TCP protocol parameters are not set. After modifying this parameter, re-do the same trial, the data is shown in Figure 2, which can be seen that its alarm throughput increases to increase linear relationship with input traffic, and the bottleneck has been eliminated. This is not only confirmed that this factor is a key factor, but also verifies the effectiveness of improvement measures. Several factors are also similar, by testing for each suspicious factor, or confirming this factor as a critical factor, or filtering factors that have little affected; then for each key factor finding technological solutions, it is not I'm going down. The success of this project created 1.660,000 yields per year.
Reviewing this project, it should be an old saying: "Solving the problem is often 99% efforts to find key reasons, and modify only 1% effort." 6Sigma does not provide technical solutions, but its ideas guides us Make in the right direction, and the data is to protect our direction. Although the project is a software project, the problem itself is clearly measured, which is also a reason why it adapts to 6Sigma features to succeed. Figure 1 Alarm processing flow test data diagram for a key factor one
Figure 2 A project modified alarm processing throughput after protocol parameters
Second, is the result of subjective judgment convincing?
This case is that the black band "reduces the abnormal code fault rate", which is one of the main fault types from CQ: The abnormal code fault is high, which reflects the responsible person to learn and progress from mistakes. Also give a lot of colleagues who still find the right project: CQ library is a very convenient project treasure house.
This project is analyzed for the measurement system of fault classification, which is a typical data analysis of discrete data. In the process of research and development, we often encounter the situation of "only can not be said", everyone is subjective judgment "shooting head", this analysis is convincing? Subjective judgment is not equal to the head, this project can be used as a reference, and the feeling of things can achieve a substantially consistent and accurate standard for everyone's subjective judgment by developing certain guidelines. Summary of the person in charge of this project:
After determining the classification rules of the fault, the classification of the fault may have different results for the classification of the same fault. This problem may be two reasons: (1) The standard for fault classification is not clear enough, and R & D personnel who participate in the fault classification are different from fault understanding. (2) R & D personnel who participate in the fault classification have no ability to classify the fault by classification standard. The solution is to measure system analysis before the fault classification, confirming whether the fault classification criteria has become clear, whether the R & D personnel who participate in the classification failure have fault classification capabilities. Analysis of discrete measurement system can be analyzed for measurement system analysis for fault classification. The basic steps for performing discrete measurement system analysis are:
1. 30 faults are randomly extracted in the fault you need to analyze. 2. A number of development managers jointly determine the classification results of each fault by fault classification criteria. (As a "true value" - this author) 3, the R & D personnel who participate in the fault classification are classified according to the fault classification standard. (As a "measured value" - this article author) 4, after a week, the R & D personnel who participated in the fault classification will be re-press the fault classification. 5. Analysis of discrete measurement system, determine the accuracy, repetition and reproducibility of fault classification.
If the repetitiveness of the fault classification is discovered by the measurement system, the same R & D personnel are not the same as the result of the same fault, and it is generally the problem of the quality of the research and development personnel. The measures taken are needed to strengthen the classification standard. . If there is a problem with the reproducibility between different R & D staff, it is generally due to the unclear classification standard, and it is necessary to further clarify the classification standard. If repetitive, reproducibility meets the requirements, basically guarantees the standard definition of fault classification is clear, and the skills of R & D personnel who participate in fault class have met the requirements. In fact, this method of measurement system analysis is not the first use. Last year, a green belt project in our institute is extremely similar, and the principle is shown in Figure 3.
Figure 3 Analysis of the analysis of the measurement system of a green belt project
This is a case of typical discrete data MSA. Many developers are very surprised when showing: "The original MSA can do so", or "It is possible to quantify data." Many people always say that data is quantified in the process of research and development, so it is not suitable for 6Sigma projects. In fact, it is not 6Sigma that cannot be used in R & D field, but many times we didn't find the right way. So think more, do more, this kind of teacher teaches us to work, it works. Third, how to improve the implementation force? I remember that a leader said: "Our company has never lacking planning, we are lacking is good execution." Some of the problems in the software is the problem, such as the implementation of the specification is not strict, the process is unreasonable, etc. Some people will ask: "If the solution is the implementation problem, you can choose rationalization suggestions according to its influence, or the team project is solved, and it is not suitable for the 6Sigma project." In fact, one 6sigma project does not know the key at the beginning. What is the factor, two execution is not simple, know and do two yards, just as a black belt: "Hear you will forget, see you will remember, you will understand." The case is another black belt "Improve Functional Tester's Software R & D Efficiency", the definition of research and development efficiency is software development and maintenance of human costs. The characteristics of this project are first very detailed. Each step can be used in accordance with 6Sigma's ideas, the method is complete, clearly described, which can be used as a model guide for beginners. However, for me, it is more important to its powerful execution. In the analysis phase, three key factors have been confirmed: 1. Modularity is not high; 2. Interface document is not standardized; 3. The system department, software department, hardware communication mechanism is uncomfortable; in order to solve the first problem, this project proposes the establishment of 10 modules, and it is currently only 3 modules; in order to solve the second problem, it is necessary to establish an interface document. Template, but more important is to get the user's recognition and operation; and the third problem is a need for the highest elder officials of three departments - ministers to communicate and coordinate. These projects are done above! How did you do it? Look at his team members, some research institutes, a general manager of the product, the minister of the three related departments, and related sectors, development managers, and some developers, such team architecture, and worry Does the plan do not implement it? The case is such a simple, and each project person in charge requires careful choice of your team members to make everyone do their best. The team needs various people: first is the representatives of all parties, and then be good at analysts, good at operator, good at coordinators, and processes or organizational person in charge. I remember that there was a project before, and members were basically the project manager. Such a team communication is smooth, but it doesn't have time to do it. How do you develop? Finally only cancels. Think about how many projects fall before reaching the end, there are few solutions to find a solution, but the solution that is not allowed to sell to the usage, so that it is common. why? Most is because there is no representative of the user in the team, and there is no one who doesn't like anything.
Some people may say that in fact, it is necessary to pull the leadership, getting the leadership, let the leaders go forward to implement and track, and everything is OK. This is wrong, because we can't get the leadership, will only be the leader to get us: Choosing the project must grasp the leaders' concerns; even if it is not the most concerned, if you are in his list, you are in front of the question . If you want to do, the leader is very trying to solve the problem, so invite him to join the project team, ask him to do some tracking work naturally, if the selected topics are far from the forefront of the leadership, let him disperse energy, At the same time, it consumes his resources to do things, he naturally refuses. This is the reason why we currently produce projects from the top. At present, our company is a certain degree of human society, recognizing this reality, and actively adjusting our approach to adapting the status quo, is a wise move for doing things. The above is my recent collection of more typical and successful software 6Sigma projects. However, I have to admit that the software project remains very small in the thousands of projects already completed. If it is because there is no successful case, affecting everyone's confidence, or don't know how to do it, I hope this article can provide some reference for everyone. CMM / CMMI and 6Sigma combination and mutual promotion, in the field of both parties, there is a large expansion space. At present, our company also has a lot of black tape EPG, and the selected project is this new topic. One year later, let's review, I hope to have a bigger breakthrough, let us also leave a glorious stroke in the history of 6Sigma.
