67
The Impact of Switching to a Rapid Release cycle on Integration Delay of Addressed Issues Daniel A. Costa Shane McIntosh Uirá Kulesza Ahmed E. Hassan UFRN McGill UFRN Queen’s

The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

  • Upload
    sailqu

  • View
    85

  • Download
    0

Embed Size (px)

Citation preview

Page 1: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

The Impact of Switching to a Rapid Release cycle on Integration Delay of

Addressed Issues

Daniel A. Costa Shane McIntosh Uirá Kulesza Ahmed E. Hassan

UFRN McGill UFRN Queen’s

Page 2: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Issues can represent bug fixes, enhancements, and new features

Bug fix

New feature

Enhancement

1

Page 3: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Software projects must attract and retain users to survive

2

Page 4: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Slow delivery of fixed issues can drive users away

So, when does this fix will get added? Firefox release 22? 23?

(Issue report #883554)

3

Page 5: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Proponents of rapid release cycles argue that fixed issues can be delivered faster

4

Page 6: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We study the rapid and traditional releases of Mozilla Firefox

Traditional Rapid

VS

5

Page 7: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Traditional Rapid111 releases 73 releases

We study the rapid and traditional releases of Mozilla Firefox

VS

5

Page 8: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Traditional Rapid111 releases 73 releases

34,673 issues 37,443 issues

We study the rapid and traditional releases of Mozilla Firefox

VS

5

Page 9: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Traditional Rapid111 releases 73 releases

34,673 issues 37,443 issues

We study the rapid and traditional releases of Mozilla Firefox

VS

1.0 to 4.0 10 to 275

Page 10: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

6

Page 11: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1)Comparing traditional

and rapid release strategies

VS

6

Page 12: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2)Comparing traditional

and rapid release strategies

Studying minor and major releases

3.6 3.6.1 3.6.2VS

6

Page 13: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

6

Page 14: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

6

Page 15: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We measure the delivery delay

3.01.0 2.0

7

Page 16: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We measure the delivery delay

3.01.0 2.0

7

Page 17: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We measure the delivery delay

3.0130 days1.0 2.0

7

Page 18: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Issues are addressed more quickly in rapid releases but are delivered more slowly

500

50

5

1

Day

s to

deliv

er fi

xed

issu

es

TraditionalRapid

8

Page 19: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Issues are addressed more quickly in rapid releases but are delivered more slowly

500

50

5

1

Day

s to

deliv

er fi

xed

issu

es

TraditionalRapid

Data density

8

Page 20: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Issues are addressed more quickly in rapid releases but are delivered more slowly

500

50

5

1

Day

s to

deliv

er fi

xed

issu

es

TraditionalRapid

MedianMedian

Data density

8

Page 21: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Issues are addressed more quickly in rapid releases but are delivered more slowly

500

50

5

1

Day

s to

deliv

er fi

xed

issu

es

TraditionalRapid

8

Page 22: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Fixed issues are delivered more consistently in rapid releases

TraditionalRapid

I can be faster!

I have consistent

gains!

9

Page 23: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

10

Page 24: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional releases can deliver issues

more quickly

10

Page 25: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional releases can deliver issues

more quickly

10

Page 26: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Minor releases are how traditional releases deliver issues more quickly

500

50

5

1

Major traditional

Minor traditional

Day

s to

deliv

er fi

xed

issu

es

11

Page 27: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Minor releases are how traditional releases deliver issues more quickly

500

50

5

1

Major traditional

Minor traditional

Day

s to

deliv

er fi

xed

issu

es Major traditional

Rapid

11

Page 28: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional releases can deliver issues

more quickly

12

Page 29: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional minor releases lead to faster

delivery of issues

Traditional releases can deliver issues

more quickly

12

Page 30: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditional

and rapid release strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional minor releases lead to faster

delivery of issues

Traditional releases can deliver issues

more quickly

12

Page 31: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Factors that can cause delivery delay

Number of Impacted Files

13

Page 32: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Factors that can cause delivery delay

Number of Impacted Files

I Think this bug is due to…

Number of comments

13

Page 33: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Factors that can cause delivery delay

Number of Impacted Files

Number of comments Experience of the resolvers

VSI Think this bug is due to…

13

Page 34: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Factors that can cause delivery delay

Number of Impacted Files

Number of comments Experience of the resolvers

VS

And more

I Think this bug is due to…

13

Page 35: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

A fixed issue is delayed if it is not shipped in the next upcoming release

3.01.0 2.0

14

Page 36: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

A fixed issue is delayed if it is not shipped in the next upcoming release

3.01.0 2.0Not delayed

14

Page 37: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

A fixed issue is delayed if it is not shipped in the next upcoming release

3.01.0 2.0 Delayed

14

Page 38: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We build logistic regression models to study factors associated with delay

Trad.

Rapid

15

Page 39: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We build logistic regression models to study factors associated with delay

Regression models

Trad.Learn from

RapidLearn from

15

Page 40: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We build logistic regression models to study factors associated with delay

Regression models

not delayedTrad.Learn from Answer

delayed

not delayedRapidLearn from Answer

delayed

15

Page 41: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

We build logistic regression models to study factors associated with delay

Regression models

not delayedTrad.Learn from Answer

delayed

not delayedRapidLearn from Answer

delayed

VS

15

Page 42: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0

Traditional

Prioritization is different along strategies

16

Page 43: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0

Traditional

Prioritization is different along strategies

Backlog of issues

16

Page 44: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0

Traditional

Prioritization is different along strategies

16

Page 45: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

Prioritization is different along strategies

16

Page 46: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ay16

Page 47: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ay16

Page 48: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ay

Release cycle backlog

16

Page 49: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ay16

Page 50: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ay

Position 2 in the cycle backlog

16

Page 51: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ayPr

obab

ility

of d

elay

0.95

0.900.850.80

Position in the cycle backlog0 1,000 3,000

Position 2 in the cycle backlog

16

Page 52: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Traditional releases behave as a queue while rapid releases behave as a stack

VS

17

Page 53: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

1.0 2.0 3.0VS

Traditional releases behave as a queue while rapid releases behave as a stack

17

Page 54: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Released Completed

1.0 2.0 3.0VS

Traditional releases behave as a queue while rapid releases behave as a stack

17

Page 55: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Released Completed

1.0 2.0 3.0

3.0

2.0

1.0

VS

Traditional releases behave as a queue while rapid releases behave as a stack

17

Page 56: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Released Completed

1.0 2.0 3.0

3.0

2.0

1.0

VS

Traditional releases behave as a queue while rapid releases behave as a stack

CompletedReleased

17

Page 57: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Case study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditionaland rapid releasing

strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional minor releases lead to faster

delivery of issues

Traditional releases can deliver issues

more quickly

1

Page 58: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Case study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditionaland rapid releasing

strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional minor releases lead to faster

delivery of issues

Traditional releases can deliver issues

more quickly

Traditional and rapid releases prioritize issues differently

1

Page 59: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

In Summary

Page 60: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Are addressed issues delivered more quickly in rapid releases?

Empirical study structure

(RQ.1) (RQ.2) (RQ.3)Comparing traditionaland rapid releasing

strategies

Studying minor and major releases

Studying delay factors of each release

strategy

3.6 3.6.1 3.6.2VS

Traditional minor releases lead to faster

delivery of issues

Traditional releases can deliver issues

more quickly

Traditional and rapid releases prioritize issues differently

Page 61: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues
Page 62: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Fixed issues are delivered more consistently in rapid releases

TraditionalRapid

I can be faster!

I have consistent

gains!

Page 63: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues
Page 64: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

Minor releases are how traditional releases deliver issues more quickly

500

50

5

1

Major traditional

Minor traditional

Day

s to

deliv

er fi

xed

issu

es Major traditional

Rapid

Page 65: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues
Page 66: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues

3.02.0Position 4 in the backlog

Traditional

2.0

Rapid

3.0

Prioritization is different along strategies

0.950.900.850.80

0 5,000 10,000Position in the backlog

Prob

abili

ty o

f del

ayPr

obab

ility

of d

elay

0.95

0.900.850.80

Position in the cycle backlog0 1,000 3,000

Position 2 in the cycle backlog

Page 67: The Impact of Switching to a Rapid Release Cycle on Integration Delay of Addressed Issues