1 / 49

Cmpe 589

Cmpe 589. Spring 2006. PROCESS ASSESSMENT. Definitions and motivation Software process models SEI capability maturity model Maturity assessment process Experience. MOTIVATION. Risk reduction Quality improvement Productivity increase. Capability Maturity Model (CMM).

joann
Download Presentation

Cmpe 589

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Cmpe 589 Spring 2006

  2. PROCESS ASSESSMENT • Definitions and motivation • Software process models • SEI capability maturity model • Maturity assessment process • Experience

  3. MOTIVATION • Risk reduction • Quality improvement • Productivity increase

  4. Capability Maturity Model (CMM) • Developed by the Software Engineering Institute (SEI) with DoD funding • Designed for large organizations doing routine development • Assessment AND evaluation • What's the difference? • Five levels of maturity • Key processes

  5. THE SEI PROCESS MATURITY MODEL Results Productivity Quality Risk

  6. Levels • Level 1: Initial • Instable; dependent on individuals • Level 2: Repeatable • Policies; use of experience in planning; discipline • Level 3: Defined • Documented process; process group; readiness and completion criteria • Level 4: Managed • Quantitative goals; data collection • Level 5: Optimized • Continuous process improvement

  7. LEVEL 1: Initial Process • Ill-defined inputs; cost and schedule overruns • Undefined process; no repeatability • Simple metrics of size, staff effort • Baseline for later comparison

  8. LEVEL 2: Repeatable Process • Identified process inputs, outputs, and constraints • No knowledge of how outputs are produced • Measures of size: • Lines of code (LOC), function points, object and method counts • Requirements volatility • Extent of personnel experience determines success • Domain / applications, development architecture, tools / methods, overall years of experience, turnover • Key areas • Requirements, management, project planning, project tracking, subcontract management, QA, CM

  9. LEVEL 3: Defined Process • Activities with definitions and entry / exit criteria • Measures of requirements complexity, design modules, code complexity, test paths, pages of documentation • Software Engineering Process Groups (SEPGs) • Quality metrics • Failures discovered, failure density for each activity area • Key areas • Organizational process definition, training program, integrated management, product engineering, intergroup coordination, peer reviews

  10. LEVEL 4: Managed Process • Feedback from early activities is used to set priorities for later stages • Data collected • Process type, extent of reuse (production and consumption), when are defects detected, testing completion criteria, use of configuration management, change control, traceability links, module completion rate • Key areas • Process measurement and analysis, quality management

  11. LEVEL 5: Optimizing Process • Measures of activities are used to change the process • Analogy with Statistical Process Control (SPC) • Key Areas • Defect prevention, technology innovation, process change management

  12. Assessment • Selection of assessment team • Management commitment • Assessment agreement • Preparation • Training, survey questionnaire • Assessment • Questionnaire analysis; discussions with projects and functional area representatives; findings; feedback; presentation • Report • Follow Up • Action plan, reassessment after 18 months

  13. Assessment Details - 1 • Assessment team has 6-8 members, some internal, some external • Either SEI or a vendor • Team members have > 10 years experience; team leader has > 20 years experience • Assessment itself takes 3-5 days • 78 YES / NO questions

  14. Assessment Details - 2 • Four or five projects are examined per organization • Interviews with 8-10 functional area representatives (FARs) from each area • QA, integration testing, coding and unit test, requirements and design • Implementation process takes 12-18 months • Follow-up at the end of this time

  15. Questionnaire - 1 2.3 Data Management and Analysis Data management deals with the gathering and retention of process metrics. Data management requires standardized data definitions, data management facilities, and a staff to ensure that data is promptly obtained, properly checked, accurately entered into the database and effectively managed. Analysis deals with the subsequent manipulation of the process data to answer questions such as, "Is there a relatively high correlation between error densities found in test and those found in use?" Other types of analyses can assist in determining the optimum use of reviews and resources, the tools most needed, testing priorities, and needed education.

  16. Questionnaire - 2 2.3.1 Has a managed and controlled process database been established for process metrics data across all projects? 2.3.2 Are the review data gathered during design reviews analyzed? 2.3.3 Is the error data from code reviews and tests analyzed to determine the likely distribution and characteristics of the errors remaining in the product? 2.3.4 Are analyses of errors conducted to determine their process related causes? 2.3.5 Is a mechanism used for error cause analysis? 2.3.6 Are the error causes reviewed to determine the process changes required to prevent them.? 2.3.7 Is a mechanism used for initiating error prevention actions? 2.3.8 Is review efficiency analyzed for each project? 2.3.9 Is software productivity analyzed for major process steps?

  17. Case Study: Hughes Aircraft - 1 • 1987: Level 2; 1990: Level 3 • $45K assessment cost; $400K improvement cost; $2M savings; 2% increased overhead; 18 months implementation (78 staff months); 5x improvement in expenditure estimation • 500 people in part of one division

  18. Case Study: Hughes Aircraft - 2 • Major 1987 recommendations • Central data repository, process group, more involvement in requirements process, technology transition organization • Major 1990 recommendations • More division-wide data analysis; opportunities for automation

  19. CMM Benefits • Level 2 leads to superior product quality • CMM encapsulates industry best practices • DOD sponsorship has enforced process improvement throughout Defense community • Quality movement has led to CMM being quite widely used in other sectors • Enhanced understanding of the development process • Increased control and risk reduction • Migration path to a more mature process • More accurate cost estimation and scheduling • Objective evaluations of changes in tools and techniques • Standardized training • Marketing

  20. CMM Criticisms • Lots of room for interpretation of assessment rules • Purpose and potential misuse of model • Originally for self-assessment and organizational learning • Increasingly used by DoD for contractor evaluation and qualification • Tends to ignore different needs of different development environments • Emphasis on DoD contractual development • Emphasis on big, mission-critical projects • Deemphasis of design risk • Deemphasis on satisfaction of customer requirements

  21. SURVEY • SEI • 1990: 113 Projects • 85% At Level 1 • 14% At Level 2 • 1% At Level 3

  22. Process Improvement

  23. Process Mistakes(McConnell) • Overly optimistic schedules • Insufficient risk management • Contractor failure • Insufficient planning • Abandonment of planning under pressure • Wasted time during fuzzy front end • Short-changed upstream activities

  24. More Process Mistakes • Inadequate design • Short changes quality assurance • Insufficient management controls • Premature or overly frequent convergence • Omitting necessary tasks from estimates • Planning to catch up later • "Code-like-hell" programming

  25. McConnell Fixes • Identify and fix classical mistakes • Miniature milestones • Milestone-based schedule • Track schedule meticulously • Record reasons for missed milestones • Recalibrate early and often • Don't commit to a new schedule until you can do so meaningfully Manage risks painstakingly

  26. Process Improvement(Shewart Cycle) • Plan • What and why? • Do • How, when, and how much? • Check • How will you know it worked? • Act • How do you plan to fully adopt? Act Plan Do Check

  27. The Dilbert Cycle Blame someone else for catastrophe Make wild guess at what is wrong ACT Adopt unproven process or technology

  28. Plan • Identify problem or opportunity • Vision • Where do you want to be? • Describe the current process • Is the process in control? Is it repeatable? • Identify possible weaknesses • Strategic impact • Cost / benefit analysis, how effort supports core-competence, risk of not doing

  29. Do • Determine what changes might help • Develop or purchase an evolution plan • Implementation tactics • Major deliverables, organization maturity considerations, key role of infrastructure, coordination with other groups, timing to match user projects, hand-off criteria • Select and implement a change on a pilot project • People and resources • Sponsor and champion roles, user partners, people and expertise needed, hardware, software, space • Schedule risks, contingency plans

  30. Check • Measure effect of change • Baseline measurement, prechange environmental characteristics, expected effect • Postproject review, degree of adoption • Percentage of engineers using improvement, range of uses, maturity of usage

  31. Act • Adopt across organization • Support strategy • Infrastructure changes, documentation, training, consulting, packaging, maintenance, feedback • Continue cycle

  32. Metrics and Measurement Get the data first, then distort it with your judgment. --Mark Twain You can’t control what you can’t measure. --Tom Demarco If you don’t know where you are going any map will do, If you don’t know where you are, a map won’t help. --Watts Humphr

  33. Getting the data • How do we decide what data to collect? • Goal, Question Metric (GQM) (Basili) • Goal, e.g. reduce defects to 6/KLOC in code • Question, e.g. what is the defect rate now? • Metric, e.g. defect count per KLOC

  34. Key Metrics • Lines of code • Staff months • Calendar months • Defects

  35. Defect Analysis • Origin (where) • Requirements / specification, design, code, environment support, documentation • Type (what) • Mode (why) • Missing, unclear, wrong, changed, better way

  36. Defect Types • Requirement / analysis: functionality • Specification / design: HW / SW / User interfaces, functional description • Design: Communication, data definition, module design, logic description, error checking, standards • Code: Logic, computation, data handling, module interface implementation, standards • Environment: Test SW, Test HW, development tools, integration SW

  37. Root Cause Analysis • Determination of the underlying process deficiency that causes a class of product defects • Collect defect data • Determine causes • Organizational buy-in (tie to business goals) • Assign responsibility • Variations • One-shot, postproject, continuous

  38. One-Shot Root Cause Analysis(Requires prior defect data collection but no prior causal analysis) • Train on defect analysis; indicate goals • Select 50-75 defects randomly • Classify; pie chart • Pick top two defect types; fishbone diagram • Improvement recommendations • Present results; assign responsibility; implement change

  39. Postproject Root Cause Analysis(Prior understanding of defect patterns) • Premeeting • Identify primary business goal • Data analysis (champion and facilitator) • Select two defect types for brainstorming • Invitations to engineers • Bring examples of their faults (in selected categories) • Guess at root cause • Suggested prevention mechanism

  40. Meeting • State meeting goal and rules • Select issues • Review defects/causes/solutions brought to meetings Analyze defects (fishbone diagram) • Brainstorm solutions • Assure organizational commitment • Plan for change (responsibilities and dates)

  41. Postmeeting • Champion and facilitator review meeting • Prepare meeting summary • Champion captures process baseline data • Before and after process description and data

  42. Fishbone Diagram(Cause and Effect Diagram) • Write suggested causes of a class of defects on note cards • Organize into affinity groups • Use lines to indicate "is a cause of" or "is a prerequisite of" • Spine is major defect class • Major branches are affinity groups

  43. Example

  44. Third Party Audits / Assessments • CMM: maturity model • ISO9001: international standard • SQPA: business practices; functional model • QMS: business model • Baldridge: Department of commerce; weighted criteria

  45. CMM V Continuously Improving • Quantitative strategic goals • Processes improved IV • Improvement based on data Quantitatively Controlled • Definition of quantitative goals III • Process metrics captured • Managing processes by data Well Defined • Development of org. std. process II • Projects use org. std. process • Sharing organizational learning Planned & Tracked I • Work is planned & managed • Projects using defined process Performed Informally • Controlling local chaos • Individual heroics 0 • Essential elements performed Legend: • Doing systems engineering Not Performed Level Title • Characterized by • Achieved when • Primary Concept • SE process area not being done • N/A • Organizational starting point SA- 11

  46. ISO 9000 Series • Actually a series of specifications that allow companies to be “certified” • Doesn’t focus on improvement—just current state • Addresses minimum criteria for a quality system • Strongly paper-trail oriented • Checklist assessment • Generic—not software specific • Principle: Every important process should be documented

  47. Guidelines for Process Improvement • Incremental rather than big-bang implementation • Contrast with “Paradigm Shift", "BPR” • Use data • Treat root causes not symptoms • Workers know best how to do a process • Avoid consultants

  48. More Guidelines • Today’s problems come from yesterday’s solutions • The harder you push, the harder the system pushes back • The easy way out usually leads back in • The cure can be worse than the disease

  49. Still More Guidelines • Faster is slower • Cause and Effect are not closely related in time and space • Small changes can produce big results, but the areas of highest leverage are the least obvious • You can have your cake and eat it too, but not at the same time • Dividing an elephant in half does not produce two small elephants • "Fix the problem, not the blame"

More Related