+
-
Copyright 2007Olivier L. de Weck and James M. Lyneis
Dynamics of Software Projects
Prof. Olivier de Weck
Guest Lecture at Ecole Polytechnique
October 5, 2007
10/05/07 2
+
-
Copyright 2007de Weck and Lyneis
Topics
Dynamic project problems
The system dynamics viewpoint
Causes of project dynamics
Application of System Dynamics to Modeling Small to Medium Size Software Projects
10/05/07 3
+
-
Copyright 2007de Weck and Lyneis
Topics
Dynamic project problems
The system dynamics viewpoint
Causes of project dynamics
Application of System Dynamics to Modeling Small to Medium Size Software Projects
10/05/07 4
+
-
Copyright 2007de Weck and Lyneis
Project Dynamics
What does “dynamic” mean in the context of a project?What are some key measures typically monitored on a project?Sketch “desired” (or planned) and typical actuals for these measures.
10/05/07 5
+
-
Copyright 2007de Weck and Lyneis
Sketch Plan and Actuals for ...
Time
ProjectStaffing
10/05/07 6
+
-
Copyright 2007de Weck and Lyneis
Sketch Plan and Actuals for ...
FractionComplete
Time
.5
1
10/05/07 7
+
-
Copyright 2007de Weck and Lyneis
Sketch Plan and Actuals for ...
Productivity(Normalised)
Time
1
2
10/05/07 8
+
-
Copyright 2007de Weck and Lyneis
Trouble-free Projects Behave as Planned: Design Labor (Equivalent People)
Simulated Data100.
75.
50.
25.
0.
TIME91 92 93 94 95 96
Disguised resultsfrom an actualsoftware project
Year 1 Year 5Year 4Year 3Year 2 Year 6
10/05/07 9
+
-
Copyright 2007de Weck and Lyneis
Design Progress (Percent Complete)Simulated100.
75.
50.
25.
0.
TIMEYear 1 Year 4Year 3Year 2 Year 5 Year 6
Disguised resultsfrom an actual
software project
10/05/07 10
+
-
Copyright 2007de Weck and Lyneis
Design Labor (Equivalent People)Simulated Data80.
60.
40.
20.
0.
TIME
Disguised results from an actual Ship-building project
Year 1 Year 4Year 3Year 2
10/05/07 11
+
-
Copyright 2007de Weck and Lyneis
Design Progress (Percent Complete)Simulated100.
75.
50.
25.
0.
TIMEYear 1 Year 4Year 3Year 2
Disguised results from an actual Ship-building project
10/05/07 12
+
-
Copyright 2007de Weck and Lyneis
So while some projects are trouble-free, more typically, projects ...
10/05/07 13
+
-
Copyright 2007de Weck and Lyneis
Typical Project “Disasters”
The Channel Tunnel -- original estimate, $3 billion; final cost, $10 billionBoston’s “Big Dig” -- original mid-1980’s estimate, $2.5 billion; latest estimate, $14.5 billion (9/2001)
10/05/07 14
+
-
Copyright 2007de Weck and Lyneis
Microsoft’s next major release of Windows operating system
January 2003: Longhorn will ship in 2004 http://redmondmag.com/news/article.asp?EditorialsID=5674
August 2004: Will Ship late 2006http://arstechnica.com/news.ars/post/20040828-4139.html
Windows Vista (AKA “Longhorn)
10/05/07 15
+
-
Copyright 2007de Weck and Lyneis
FBI’s Virtual Case File
Virtual Case File is would allow agents to share info easilyCongress and others believe VCF is “critical to enabling the FBI and intelligence agencies to ‘connect the dots’ in preventing [terrorist] attacks”InformationWeek Feb 4, 2005
May 2004: Project will be delivered by end of 2004 Feb 2005: Mueller acknowledges he doesn’t know how much it will cost ($170 million already spent) or when it will be deliveredFBI may have to scrap what they’ve done so far and start fresh
InformationWeek Feb 4, 2005 from the Associated Press
10/05/07 16
+
-
Copyright 2007de Weck and Lyneis
Software Projects are Tough
On average software projects overshoot schedule by 50%⅓ of large-scale software projects never complete (cancelled)¾ of all large systems that are completed are operating failures*
Source: Gibbs, W. 1994. “Software’s Chronic Crisis”. Scientific American. September. pp. 86-95
10/05/07 17
+
-
Copyright 2007de Weck and Lyneis
Lamri Survey
http://www.lamri.com/CMMI_Survey_results.pdf
10/05/07 18
+
-
Copyright 2007de Weck and Lyneis
So while some projects are trouble-free, more typically, projects ...
... Surprise us with late requests for additional time or resources... Have schedule and budget overruns... Seem to get stuck at “90%” complete... Result in new products with flaws discovered after release... Hinder learning because of difficulty in comparing the performance of different, “unique” projects
10/05/07 19
+
-
Copyright 2007de Weck and Lyneis
On typical development projects ...Design Labor (Equivalent People)
Simulated Actual
Year
800.
600.
400.
200.
0 0 3 6 9 12 15 18 21
ApproximateOriginal Plan
Disguised results from actual aerospace project
… Staffing experiences an extended tail
10/05/07 20
+
-
Copyright 2007de Weck and Lyneis
Program Staff, Simulated vs. Data (Equivalent Staff)
400.
300.
200.
100.
TIME
0 Year 1 Year 2 Year 3 Year 4 Year 5 Year 6
Simulated Original Plan Actual
Disguised results from actual vehicle project
… Or a second staffing peak
10/05/07 21
+
-
Copyright 2007de Weck and Lyneis
Two software development projects for the same aerospace company experienced dramatically different results
[see next 3 pages; figures show disguised results shifted so that the 2 projects started in the same year]
10/05/07 22
+
-
Copyright 2007de Weck and Lyneis
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
150
100
50
0
Program staff (Program A ...)
10/05/07 23
+
-
Copyright 2007de Weck and Lyneis
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
150
100
50
0
Program staff (Program B)
10/05/07 24
+
-
Copyright 2007de Weck and Lyneis
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
150
100
50
0
Program staff (Program B vs. Program A)
10/05/07 25
+
-
Copyright 2007de Weck and Lyneis
What caused the differences between Program A and Program B?
Differences in work scope?
External Conditions?
Management policies and processes?
And how can a company learn from these differences, and therefore
Bid better?
Plan better?
Manage better?
10/05/07 26
+
-
Copyright 2007de Weck and Lyneis
Progress has been made: Many firms have cut development times in half ...
Source: Smith, Preston G. and Donald G Reinersten, Developing Products in Half the Time (2nd edition), Wiley 1998.
10/05/07 27
+
-
Copyright 2007de Weck and Lyneis
PERT and CPM (Critical Path Method)Waterfall, Spiral, …Emphasis on “soft,” people factorsMicrosoft Project
.. and Learning is not happeningWhy???
Project problems persist in spite of numerous advances in the last 30 years
10/05/07 28
+
-
Copyright 2007de Weck and Lyneis
Topics
Dynamic project problems
The system dynamics viewpoint
Causes of project dynamics
Application of System Dynamics to Modeling Small to Medium Size Software Projects
10/05/07 29
+
-
Copyright 2007de Weck and Lyneis
Events “Change Occurred,Project behind schedule”
Three ways of looking at a problem --
10/05/07 30
+
-
Copyright 2007de Weck and Lyneis
Events
Patterns of
BehaviorEst. CompletionDate
Add.Resources
“Project behind schedule”
Three ways of looking at a problem --
10/05/07 31
+
-
Copyright 2007de Weck and Lyneis
System Structure
EstimatedCompletion
Date
AddResources
BehindSchedule
RemainingWork
Experience Dilution,Quality Problems
Three ways of looking at a problem --
10/05/07 32
+
-
Copyright 2007de Weck and Lyneis
Events
Patterns of Behavior
System Structure
Est. CompletionDate
Add.Resources
EstimatedCompletion
AddResources
BehindSchedule
RemainingWork
Experience Dilution,Quality Problems
“Project behind schedule”
Three ways of looking at a problem --
10/05/07 33
+
-
Copyright 2007de Weck and Lyneis
The System Dynamics Viewpoint
Events
Patterns of Behavior
System Structure
External Factors
10/05/07 34
+
-
Copyright 2007de Weck and Lyneis
The System Dynamics Viewpoint
Events
Patterns of Behavior
System Structure
Low
Medium
High
Ability toInfluence
External Factors
10/05/07 35
+
-
Copyright 2007de Weck and Lyneis
Why??
Project performance problems are fundamentally dynamic problems,
but Managers mental models and typical tools (computer models):
Attribute all problems to external factorsView a project statically (no iteration, no feedback)Treat projects as if they were unique
10/05/07 36
+
-
Copyright 2007de Weck and Lyneis
System Dynamics Concepts
Focus on dynamic behaviorSystem structure as cause of that behaviorEmphasis on internal dynamics and how we can change structureA model is an integral part of organizational learning
10/05/07 37
+
-
Copyright 2007de Weck and Lyneis
Examples of Behavior Modes
"Plan"
Growth
ProjectStaffing
Stability
10/05/07 38
+
-
Copyright 2007de Weck and Lyneis
What Causes Dynamics?
All dynamics are driven by –Feedback processesAccumulation processes
10/05/07 39
+
-
Copyright 2007de Weck and Lyneis
The basic element of system structure is the feedback loop...
Chickens
TIME
...This one a “positive,”or reinforcing feedback loop.
+
+
EGGS CHICKENS
Adapted from Professor John Sterman, MIT
10/05/07 40
+
-
Copyright 2007de Weck and Lyneis
Loops of effects interact with one another, making traditional diagnosis quite difficult...
...Such as with this “negative,” or balancing, feedback loop, which may
slow or reverse growth
Chickens
TIME
+
+
EGGS CHICKENS ROADCROSSINGSX
FARMSIZE
-
+
10/05/07 41
+
-
Copyright 2007de Weck and Lyneis
Effects interacting makes traditional diagnosis quite difficult in a business setting as well
PRICE
UNITCOSTS
USAGE
SERVICECAPACITY
SERVICEQUALITY
10/05/07 42
+
-
Copyright 2007de Weck and Lyneis
Accumulation processes involve ...
Stocks or “levels” -- define the state of the system
Flows or “rates” -- define the rate of change in system states
Bank Account Balance
Deposits
(iThink Rate Symbol)
Deposits
(Vensim/DYNAMO Symbol)
10/05/07 43
+
-
Copyright 2007de Weck and Lyneis
Connecting stocks and flows ...
Bank Account Balance
Deposits Withdrawals
“Clouds”represent stocks outside the system boundary
StaffHiring Firing
10/05/07 44
+
-
Copyright 2007de Weck and Lyneis
Interacting positive and negative feedback loops of cause-effect relationships, with stocks, flows, delays, and non-linearities, are capable of generating all observed modes of behavior.
10/05/07 45
+
-
Copyright 2007de Weck and Lyneis
Topics
Dynamic project problems
The system dynamics viewpoint
Causes of project dynamics
Application of System Dynamics to Modeling Small to Medium Size Software Projects
10/05/07 46
+
-
Copyright 2007de Weck and Lyneis
System dynamics modeling provides a means of ...
... understanding the structure of projects, and how that structure creates behavior;... designing robust projects... learning across projects
10/05/07 47
+
-
Copyright 2007de Weck and Lyneis
But every project is unique ...
Many similarities, some uniquenessWe will discuss the characteristics of different kinds
of projects later in the term.
SD is a framework for assessing dynamic similaritiesBut, there is no “one way”, “right answer”We will discuss heuristics and rules of thumb during
the course.
10/05/07 48
+
-
Copyright 2007de Weck and Lyneis
Examples of Behavior Modes on a Project
ProjectStaffing
Time
TypicalPlan
10/05/07 49
+
-
Copyright 2007de Weck and Lyneis
Examples (continued)
FractionComplete
Time
.5
1
TypicalPlan
10/05/07 50
+
-
Copyright 2007de Weck and Lyneis
Examples (continued)
Productivity(Normalised)
Time
1
2Typical
Plan
10/05/07 51
+
-
Copyright 2007de Weck and Lyneis
The Drivers of Project Dynamics --
The “rework cycle”Feedback effects on productivity and work quality (often “vicious circles”)Knock-on effects between work phasesKnock-on effects between projects
10/05/07 52
+
-
Copyright 2007de Weck and Lyneis
What are typical stocks & flows on a project?
Stocks Flows
10/05/07 53
+
-
Copyright 2007de Weck and Lyneis
The Traditional View of a Program – Task Accomplishment
WORKBEING DONE
PEOPLE PRODUCTIVITY
WORKTO BEDONE
WORKDONE
WORK TO DO STAFF % DONE
TIMETIMETIME
Task Accomplishment
10/05/07 54
+
-
Copyright 2007de Weck and Lyneis
The traditional view, expanded to multiple tasks with logical links ...
... becomes a Critical Path Network
Task A
Task B
Task C
Task D
Task E
Task F Task Z
10/05/07 55
+
-
Copyright 2007de Weck and Lyneis
But traditional approaches fail to consider iteration and rework
WORKBEING DONE
PEOPLE PRODUCTIVITY QUALITY
WORKTO BEDONE
WORKREALLYDONE
REWORK
WORK TO DO STAFF % DONE
RWK
TIME TIME TIME
10/05/07 56
+
-
Copyright 2007de Weck and Lyneis
Or Undiscovered Rework
WORKBEING DONE
PEOPLE PRODUCTIVITY QUALITY
KNOWNREWORK
UNDISCOVEREDREWORK
REWORK DISCOVERY
WORKTO BEDONE
WORKREALLYDONE
WORK TO DO STAFF % DONE
RWK
UR REALLY
TIME TIME TIME
10/05/07 57
+
-
Copyright 2007de Weck and Lyneis
On typical development projects ...Simulated Actual
Year
800.
600.
400.
200.
0 0 3 6 9 12 15 18 21
ApproximateOriginal Plan
Disguised results from actual aerospace project
… Staffing experiences an extended tail
10/05/07 58
+
-
Copyright 2007de Weck and Lyneis
Caused by the need to accomplish rework
0 3 6 9 12 15 18 21
800.
600.
400.
200.
0
Work Assignments of Staff to...
Original Work
Rework
Total
10/05/07 59
+
-
Copyright 2007de Weck and Lyneis
Program Staff, Simulated vs. Data (Equivalent Staff)
400.
300.
200.
100.
TIME
0 Year 1 Year 2 Year 3 Year 4 Year 5 Year 6
Simulated Original Plan Actual
Disguised results from actual vehicle project
… A second staffing peak
On typical development projects ...
10/05/07 60
+
-
Copyright 2007de Weck and Lyneis
Work Assignments of Staff to...
Again caused by the need to execute rework
400.
300.
200.
100.
TIME
0 Year 1 Year 2 Year 3 Year 4 Year 5 Year 6
Original Work Rework
Disguised results from actual vehicle project
Rework
Original Work
10/05/07 61
+
-
Copyright 2007de Weck and Lyneis
The Rework CycleKey to Project Dynamics
WORK
BEING DONE
People Productivity Quality
Rework Discovery
WORKTO BEDONE
KNOWNREWORK
UNDISCOVEREDREWORK
WORKREALLYDONE
10/05/07 62
+
-
Copyright 2007de Weck and Lyneis
Why? The Drivers of Project Dynamics --
The “rework cycle”Feedback effects on productivity and work quality (often “vicious circles”)Knock-on effects between work phasesKnock-on effects between projects
10/05/07 63
+
-
Copyright 2007de Weck and Lyneis
What drives productivity & quality?
UndiscoveredRework
KnownRework
WorkReally Done
WorkTo Be Done
Progress
Rework Discovery
People Productivity Quality
CustomerChanges
Dynamics are initiated by (1) an infeasible initial plan, or (2) changes adding scope and obsoletingcompleted work ...
Scope Growth
10/05/07 64
+
-
Copyright 2007de Weck and Lyneis
What drives productivity & quality?
UndiscoveredRework
KnownRework
WorkReally Done
WorkTo Be Done
Progress
Rework Discovery
People Productivity Quality
ApparentProgress
CustomerChanges Schedule
Acceleration
Overtime
Hiring
Scope Growth
… that initiate corrective actions to get the project back
on track...
10/05/07 65
+
-
Copyright 2007de Weck and Lyneis
What drives productivity & quality?
UndiscoveredRework
KnownRework
WorkReally Done
WorkTo Be Done
Progress
Rework Discovery
People Productivity Quality
ApparentProgress
CustomerChanges Schedule
Acceleration
Out-of-SequenceWork, Worksite
Congestion, CoordinationProblems,
Morale Problems
AverageEmployee Skill
and Quality
Fatigue,Burnout
Overtime
Hiring
Scope Growth
… that create vicious circles that undercut intended effects
10/05/07 66
+
-
Copyright 2007de Weck and Lyneis
A system dynamics model usually represents several phases of work, but is more aggregate than a CPM model
SystemEngineering
Software Codeand Test
HardwareDesign
Hardware Buildand Test
Integrateand Test
SoftwareDesign
10/05/07 67
+
-
Copyright 2007de Weck and Lyneis
Summary: The Drivers of Project Dynamics --
The “rework cycle”Feedback effects on productivity and work quality (often “vicious circles”)Knock-on effects between work phasesKnock-on effects between projects
10/05/07 68
+
-
Copyright 2007de Weck and Lyneis
Topics
Dynamic project problems
The system dynamics viewpoint
Causes of project dynamics
Application of System Dynamics to Modeling Small to Medium Size Software Projects
10/05/07 69
+
-
Copyright 2007de Weck and Lyneis
SM Thesis
A Comparative Study of Iterative Prototyping vs. Waterfall Process Applied To Small and Medium Sized Software Projects
Eduardo Malaga ChocanoMIT System Design and Management ProgramJune 2004
10/05/07 70
+
-
Copyright 2007de Weck and Lyneis
Objectives
1. Identify the main strengths and weaknesses of iterative cycle based models vs. sequential-based models applied to small and medium sized software projects.
2. Measure the impact these features have in the management of projects
3. Understand under which conditions each of these approaches is recommended.
4. Study the new trends in this field and propose recommendations regarding their use.
10/05/07 71
+
-
Copyright 2007de Weck and Lyneis
Waterfall Projects
10/05/07 72
+
-
Copyright 2007de Weck and Lyneis
Spiral Projects
10/05/07 73
+
-
Copyright 2007de Weck and Lyneis
Interviews of Peruvian SW Project Leaders
10/05/07 74
+
-
Copyright 2007de Weck and Lyneis
Characteristics of Successful Projects
10/05/07 75
+
-
Copyright 2007de Weck and Lyneis
Source of Problems
10/05/07 76
+
-
Copyright 2007de Weck and Lyneis
Software Project Model (overview)
10/05/07 77
+
-
Copyright 2007de Weck and Lyneis
Software Project Model (detailed)
Model Parameters
-Number of Tasks-Development Rate-Verification Rate-Flag
-sequential-iterative
-Verification Period-Phase Size Table-Insight per Phase-Insight Development-Verification Accuracy-% dedicated to verify
10/05/07 78
+
-
Copyright 2007de Weck and Lyneis
Results (I) – Speed vs Quality
Speed vs QualityRun with nominal parameters
10/05/07 79
+
-
Copyright 2007de Weck and Lyneis
Results (2) – Task Distribution
10/05/07 80
+
-
Copyright 2007de Weck and Lyneis
Results (II) – Amount of Rework
Iterative Approach causes more relative rework
10/05/07 81
+
-
Copyright 2007de Weck and Lyneis
Conclusioniterative
iterative
10/05/07 82
+
-
Copyright 2007de Weck and Lyneis
SummarySoftware Development Methodology have a significant impact on project success
Iterative Approach:more effective use of developer’s timeshort iterations, minimize idle time waiting for tasks to be reworked
Sequential Approach:programmers wait idle while testing is underwayless rework necessary if insight can be developed fastless variance of outcome
Hybrid Approaches existFinal Note – large difference between software for business applications and safety-critical real-time embedded software
business software – very suitable to spiral approachreal time embedded software – rather sequential
Top Related