Tweet
Tweet

Slide 1

Slide 1 text

On Debugging the Performance of Configurable Software Systems Miguel Velez (CMU -> Google) Norbert Siegmund (Leipzig) Pooyan Jamshidi (UofSC & Google) Sven Apel (Saarland) Christian Kästner (CMU) Developer Needs and Tailored Tool Support

Slide 2

Slide 2 text

Most software is con fi gurable

Slide 3

Slide 3 text

Indexing Encryption Compression Logging Core functionality Configurable System Need: Need: Need:

Slide 4

Slide 4 text

Indexing Encryption Compression Logging Core functionality Configurable System

Slide 5

Slide 5 text

No content

Slide 6

Slide 6 text

6 Performance Execution time Energy consumption Operational costs

Slide 7

Slide 7 text

Developer • Design, implement, and maintain efficient software • Debug surprising performance behaviors

Slide 8

Slide 8 text

8 59% - Con fi guration-related performance issues 61% - Average fi x time of 5 weeks 50% - Con fi guration-related patches

Slide 9

Slide 9 text

No content

Slide 10

Slide 10 text

Traditional Off-the-Shelf Profilers

Slide 11

Slide 11 text

Searching Inefficient Coding Patterns

Slide 12

Slide 12 text

Program Debugging Techniques 12

Slide 13

Slide 13 text

13

Slide 14

Slide 14 text

14 Limited empirical evidence of the usefulness to help debug the performance of con fi gurable systems

Slide 15

Slide 15 text

15

Slide 16

Slide 16 text

16

Slide 17

Slide 17 text

Developer • Understand how options affect performance • Design, implement, and maintain efficient software Help developers debug the performance of configurable software systems

Slide 18

Slide 18 text

Help developers debug the performance of configurable software systems

Slide 19

Slide 19 text

Identify Information Needs Influencing Options Option Hotspots Cause-Effect Chain Help developers debug the performance of configurable software systems

Slide 20

Slide 20 text

Identify Information Needs Influencing Options Option Hotspots Cause-Effect Chain Help developers debug the performance of configurable software systems Tailor Ingredients CPU Profiling Program Slicing Performance Modeling Global Local

Slide 21

Slide 21 text

No content

Slide 22

Slide 22 text

22 User Study to Identify Information Needs

Slide 23

Slide 23 text

23 User Study to Identify Information Needs Overview of Ingredients

Slide 24

Slide 24 text

24 User Study to Identify Information Needs Overview of Ingredients Tailor and Evaluate Ingredients

Slide 25

Slide 25 text

25 Overview of Ingredients Tailor and Evaluate Ingredients User Study to Identify Information Needs

Slide 26

Slide 26 text

Information Needs RQ1: What information do developers look for when debugging the performance of con fi gurable systems? RQ2: What is the process that developers follow and the activities that they perform to obtain this information?

Slide 27

Slide 27 text

Information Needs RQ1: What information do developers look for when debugging the performance of con fi gurable systems? RQ2: What is the process that developers follow and the activities that they perform to obtain this information? Exploratory User Study

Slide 28

Slide 28 text

Exploratory User Study

Slide 29

Slide 29 text

Exploratory User Study Influencing Options The option or interaction causing the unexpected performance behavior

Slide 30

Slide 30 text

Exploratory User Study Influencing Options Option Hotspots The methods where options affect the performance of the system

Slide 31

Slide 31 text

Exploratory User Study Influencing Options Option Hotspots Cause-Effect Chain How in fl uencing options are used in the implementation to directly and indirectly affect the performance of option hotspots

Slide 32

Slide 32 text

Exploratory User Study Influencing Options Option Hotspots Cause-Effect Chain Compared problematic con fi guration to a non-problematic con fi guration

Slide 33

Slide 33 text

No content

Slide 34

Slide 34 text

User Study to Identify Information Needs Influencing Options Option Hotspots Cause-Effect Chain Help developers debug the performance of configurable software systems

Slide 35

Slide 35 text

35 Tailor and Evaluate Ingredients User Study to Identify Information Needs Overview of Ingredients

Slide 36

Slide 36 text

The option or interaction causing the unexpected performance behavior Influencing Options

Slide 37

Slide 37 text

The option or interaction causing the unexpected performance behavior Influencing Options Global Performance-In fl uence Models

Slide 38

Slide 38 text

… 23 seconds 28 seconds 20 seconds Con fi gurations Performance T = 25 + 3· - 5· + 9· · Performance-Influence Models

Slide 39

Slide 39 text

… 23 seconds 28 seconds 20 seconds Con fi gurations Performance T = 25 + 3· - 5· + 9· · Performance-Influence Models

Slide 40

Slide 40 text

… 23 seconds 28 seconds 20 seconds Con fi gurations Performance T = 25 + 3· - 5· + 9· · Performance-Influence Models

Slide 41

Slide 41 text

Modeling Global T = 25 + 3· - 5· + 9· · Influencing Options

Slide 42

Slide 42 text

The methods where options affect the performance of the system Option Hotspots

Slide 43

Slide 43 text

The methods where options affect the performance of the system Option Hotspots Local Performance-In fl uence Models

Slide 44

Slide 44 text

Modeling Local Tfoo = 5 + 3· + 9· · Option Hotspots

Slide 45

Slide 45 text

How in fl uencing options are used in the implementation to directly and indirectly affect the performance of option hotspots Cause-Effect Chain

Slide 46

Slide 46 text

How in fl uencing options are used in the implementation to directly and indirectly affect the performance of option hotspots Cause-Effect Chain CPU Pro fi ling & Program Slicing

Slide 47

Slide 47 text

CPU Profiling Cause-Effect Chain

Slide 48

Slide 48 text

Program Slicing Cause-Effect Chain

Slide 49

Slide 49 text

Overview of Ingredients Influencing Options Option Hotspots Cause-Effect Chain Help developers debug the performance of configurable software systems CPU Profiling Program Slicing Modeling Global Local

Slide 50

Slide 50 text

50 User Study to Identify Information Needs Overview of Ingredients Tailor and Evaluate Ingredients

Slide 51

Slide 51 text

Influencing Options Option Hotspots Cause-Effect Chain CPU Profiling Program Slicing Performance Modeling Global Local

Slide 52

Slide 52 text

Influencing Options Option Hotspots Cause-Effect Chain CPU Profiling Program Slicing Performance Modeling Global Local Information Providers

Slide 53

Slide 53 text

Influencing Options Performance Modeling Global

Slide 54

Slide 54 text

Influencing Options Performance Modeling Global

Slide 55

Slide 55 text

Influencing Options Performance Modeling Global

Slide 56

Slide 56 text

Influencing Options Performance Modeling Global

Slide 57

Slide 57 text

Influencing Options Performance Modeling Global

Slide 58

Slide 58 text

Option Hotspots Performance Modeling Local

Slide 59

Slide 59 text

Option Hotspots Performance Modeling Local

Slide 60

Slide 60 text

Option Hotspots Performance Modeling Local

Slide 61

Slide 61 text

Option Hotspots Performance Modeling Local

Slide 62

Slide 62 text

Cause-Effect Chain CPU Profiling Program Slicing

Slide 63

Slide 63 text

Cause-Effect Chain CPU Profiling

Slide 64

Slide 64 text

Cause-Effect Chain CPU Profiling

Slide 65

Slide 65 text

Cause-Effect Chain CPU Profiling

Slide 66

Slide 66 text

Cause-Effect Chain CPU Profiling

Slide 67

Slide 67 text

Cause-Effect Chain CPU Profiling

Slide 68

Slide 68 text

Cause-Effect Chain Program Slicing

Slide 69

Slide 69 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 70

Slide 70 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 71

Slide 71 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 72

Slide 72 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 73

Slide 73 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 74

Slide 74 text

Main.main(…) Cursor.put(…) Main.init(…) Database.init(…) def init(Database db) Database.checkForNullParam(db.name, "dbName"); Log.msg(Level.INFO, "db " + db.name + " open"); ... if(db.replicated) configReplicated(...); ... this.cacheMode = db.cacheMode; this.commit = db.trans ? true : false; this.sync = db.dups ? true : false; if(db.evict) Evictor.init(db.evictorThreads); ... Cause-Effect Chain Program Slicing

Slide 75

Slide 75 text

Influencing Options Option Hotspots Cause-Effect Chain Performance Modeling Global

Slide 76

Slide 76 text

Influencing Options Option Hotspots Cause-Effect Chain Performance Modeling Global Local Performance Modeling

Slide 77

Slide 77 text

Influencing Options Option Hotspots Cause-Effect Chain CPU Profiling Program Slicing Performance Modeling Global Local Performance Modeling

Slide 78

Slide 78 text

Influencing Options Option Hotspots Cause-Effect Chain CPU Profiling Program Slicing Performance Modeling Global Local Performance Modeling GLIMPS

Slide 79

Slide 79 text

Evaluation

Slide 80

Slide 80 text

Evaluation RQ1: To what extent do the designed information providers help developers debug the performance of con fi gurable software systems?

Slide 81

Slide 81 text

Evaluation RQ1: To what extent do the designed information providers help developers debug the performance of con fi gurable software systems? Validation User Study Confirmatory User Study

Slide 82

Slide 82 text

“Within-subject” study Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS System: Density Converter

Slide 83

Slide 83 text

Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS “Within-subject” study System: Density Converter

Slide 84

Slide 84 text

Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS “Within-subject” study System: Density Converter

Slide 85

Slide 85 text

Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS “Within-subject” study System: Density Converter

Slide 86

Slide 86 text

Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS SCALE +86.3 AConverter.compress(…) +67.4 “Within-subject” study System: Density Converter

Slide 87

Slide 87 text

Validation User Study 30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Control - Did not use GLIMPS Treatment - Used GLIMPS SCALE +86.3 AConverter.compress(…) +67.4 “Within-subject” study System: Density Converter Information providers support information needs

Slide 88

Slide 88 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study

Slide 89

Slide 89 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study

Slide 90

Slide 90 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study

Slide 91

Slide 91 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study

Slide 92

Slide 92 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study

Slide 93

Slide 93 text

30’ 35’ 45’ 05’ 10’ 20’ 25’ 15’ 40’ 50’ Influencing Options Option Hotspots Cause-Effect Chain Confirmatory User Study 55’ 60’ Treatment - Used GLIMPS Control - Did not use GLIMPS System: Berkeley DB Between-subject study Information providers help developers debug the performance of complex con fi gurable systems

Slide 94

Slide 94 text

Influencing Options Option Hotspots Cause-Effect Chain Help developers debug the performance of configurable software systems CPU Profiling Program Slicing Performance Modeling Global Local Tailor and Evaluate Ingredients

Slide 95

Slide 95 text

On Debugging the Performance of Configurable Software Systems: Developer Needs and Tailored Tool Support

Slide 96

Slide 96 text

96 Overview of Ingredients Tailor and Evaluate Ingredients CONTRIBUTION User Study to Identify Information Needs CONTRIBUTION White-box Performance-In fl uence Modeling CONTRIBUTION

Slide 97

Slide 97 text

Compositionality Compression Insights!

Slide 98

Slide 98 text

Compositionality Performance-in fl uence models can be built by composing models built independently for smaller regions of code System Performance-In fl uence Model Local Performance-In fl uence Model Region Local Performance-In fl uence Model Local Performance-In fl uence Model Local Performance-In fl uence Model Region Region Region Measure System Decompose Measure Regions Compose 98

Slide 99

Slide 99 text

Performance-In fl uence Model Local Performance-In fl uence Model Region Local Performance-In fl uence Model Local Performance-In fl uence Model Local Performance-In fl uence Model Region Region Region Measure System Decompose Measure Regions Compose System 99 Compositionality Performance-in fl uence models can be built by composing models built independently for smaller regions of code

Slide 100

Slide 100 text

Performance-In fl uence Model Local Performance-In fl uence Model Region Local Performance-In fl uence Model Local Performance-In fl uence Model Local Performance-In fl uence Model Region Region Region Measure System Decompose Measure Regions Compose System 100 Compositionality Performance-in fl uence models can be built by composing models built independently for smaller regions of code

Slide 101

Slide 101 text

Performance-In fl uence Model Local Performance-In fl uence Model Region Local Performance-In fl uence Model Local Performance-In fl uence Model Local Performance-In fl uence Model Region Region Region Measure System Decompose Measure Regions Compose System 101 Compositionality Performance-in fl uence models can be built by composing models built independently for smaller regions of code

Slide 102

Slide 102 text

Performance-In fl uence Model Local Performance-In fl uence Model Region Local Performance-In fl uence Model Local Performance-In fl uence Model Local Performance-In fl uence Model Region Region Region Measure System Decompose Measure Regions Compose System 102 Compositionality Performance-in fl uence models can be built by composing models built independently for smaller regions of code

Slide 103

Slide 103 text

def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 103

Slide 104

Slide 104 text

Time TRUE FALSE Time TRUE TRUE TRUE FALSE def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 104

Slide 105

Slide 105 text

Time TRUE 5 FALSE 0 def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 105 Time TRUE TRUE TRUE FALSE

Slide 106

Slide 106 text

Time TRUE 5 FALSE 0 Tprocess = 5· def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 106 Time TRUE TRUE TRUE FALSE

Slide 107

Slide 107 text

Time TRUE 5 FALSE 0 Tprocess = 5· Time TRUE TRUE 3 TRUE FALSE 2 Tconvert = 2· + 1·. · def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 107

Slide 108

Slide 108 text

Tprocess = 5· Tconvert = 2· + 1·. · Compositionality + def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } 108

Slide 109

Slide 109 text

Tprocess = 5· Tconvert = 2· + 1·. · Compositionality + def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } T = 7· + 1· · 109

Slide 110

Slide 110 text

Compositionality System System 27 = 128 con fi gurations 22 + 24 + 22 + 24 + 22 = 38 con fi gurations 110

Slide 111

Slide 111 text

Compositionality System System 27 = 128 con fi gurations 22 + 24 + 22 + 24 + 22 = 38 con fi gurations 111

Slide 112

Slide 112 text

Compositionality System System 27 = 128 con fi gurations 22 + 24 + 22 + 24 + 22 = 38 con fi gurations 112

Slide 113

Slide 113 text

Compositionality Insights! Compression

Slide 114

Slide 114 text

Compression Compression allow us to simultaneously explore paths in multiple independent regions with a few con fi gurations 114

Slide 115

Slide 115 text

Compression if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds 115

Slide 116

Slide 116 text

Compression if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds Time TRUE FALSE Time TRUE FALSE Time TRUE FALSE 116

Slide 117

Slide 117 text

FALSE FALSE FALSE TRUE FALSE FALSE TRUE TRUE FALSE Compression if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds Time TRUE FALSE Time TRUE FALSE Time TRUE FALSE 117

Slide 118

Slide 118 text

FALSE FALSE FALSE TRUE FALSE FALSE TRUE TRUE FALSE if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds Time TRUE FALSE Time TRUE FALSE Time TRUE FALSE 118 Compression

Slide 119

Slide 119 text

FALSE FALSE FALSE TRUE FALSE FALSE TRUE TRUE FALSE if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds Time TRUE FALSE 0 Time TRUE FALSE 0 Time TRUE FALSE 0 119 Compression

Slide 120

Slide 120 text

FALSE FALSE FALSE TRUE FALSE FALSE TRUE TRUE FALSE if(a) ... // execution time: 1 second if(b) ... // execution time: 2 seconds if(c) ... // execution time: 3 seconds Time TRUE 1 FALSE 0 Time TRUE 3 FALSE 0 Time TRUE 2 FALSE 0 120 Compression

Slide 121

Slide 121 text

System 22 + 24 + 22 + 24 + 22 = 38 con fi gurations System 4 con fi gurations 121 Compression

Slide 122

Slide 122 text

System 22 + 24 + 22 + 24 + 22 = 38 con fi gurations System 4 con fi gurations 122 Compression

Slide 123

Slide 123 text

System 22 + 24 + 22 + 24 + 22 = 38 con fi gurations System 4 con fi gurations 123 Compression

Slide 124

Slide 124 text

System 4 con fi gurations FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE TRUE TRUE FALSE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE FALSE TRUE FALSE 124 Compression

Slide 125

Slide 125 text

System 4 con fi gurations FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE TRUE TRUE FALSE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE FALSE TRUE FALSE 125 Compression

Slide 126

Slide 126 text

System 4 con fi gurations FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE TRUE TRUE FALSE TRUE TRUE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE TRUE FALSE FALSE TRUE FALSE 126 Compression

Slide 127

Slide 127 text

Compression Measure Performance Build Model Compositionality System 127

Slide 128

Slide 128 text

Taint Analysis

Slide 129

Slide 129 text

def main() { boolean a = ... process(a, b); } def process(boolean x, boolean y) { if(x) convert(y); ... // execution time: 5 seconds } def convert(boolean x) { if(x) ... // execution time: 3 seconds else ... // execution time: 2 seconds } In fl uencing options 129

Slide 130

Slide 130 text

Analyze Regions Measure Performance Build Model Taint Analysis Compression Compositionality System 130

Slide 131

Slide 131 text

ConfigCrusher Comprex Static Taint Analysis Regions: Control-flow statements Performance measurement: Instrumentation Dynamic Taint Analysis Regions: Methods Performance measurement: Off-the-shelf profiler 131

Slide 132

Slide 132 text

ConfigCrusher Comprex Static Taint Analysis Regions: Control-flow statements Dynamic Taint Analysis Regions: Methods 132 Both prototypes ef fi ciently build accurate and interpretable models Dynamic taint analysis scales to larger systems Method-level granularity does not sacri fi ces compression potential

Slide 133

Slide 133 text

Evaluation

Slide 134

Slide 134 text

134 ConfigCrusher Comprex Evaluate cost, accuracy, interpretability 13 open-source Java systems Evaluation 50 combinations of Sampling and Machine Learning

Slide 135

Slide 135 text

Results: Density Converter 0 Cost (minutes) Error (MAPE) 0 40 20 Comprex x 200 random con fi gurations & Random Forest 5 10 200 random con fi gurations & Stepwise linear regression 200 400 100 300 x x 60 Con fi gCrusher x 135

Slide 136

Slide 136 text

x + + + Results: Density Converter 0 Cost (minutes) Error (MAPE) 0 40 20 Comprex 200 random con fi gurations & Random Forest 5 10 200 random con fi gurations & Stepwise linear regression 200 400 100 300 60 Con fi gCrusher + Interpretable x Not interpretable 136

Slide 137

Slide 137 text

x + + + Results: Density Converter 0 Cost (minutes) Error (MAPE) 0 40 20 Comprex 200 random con fi gurations & Random Forest 5 10 200 random con fi gurations & Stepwise linear regression 200 400 100 300 60 Con fi gCrusher + Interpretable x Not interpretable 137 Ef fi ciently build accurate performance- in fl uence models Models are interpretable for debugging purposes

Slide 138

Slide 138 text

Influencing Options Option Hotspots White-box Modeling Global Local 138 Help developers debug the performance of configurable software systems White-box Performance-In fl uence Modeling CONTRIBUTION

Slide 139

Slide 139 text

Thesis Statement Tailoring speci fi c white-box analyses to track how con fi guration options in fl uence the performance of code-level structures in con fi gurable software systems helps developers to (1) ef fi ciently build accurate and interpretable global and local performance-in fl uence models (2) more easily inspect, trace, understand, and debug con fi guration-related performance issues. 139