Let's Talk a Bit About: Green Software
•Transferir como PPTX, PDF•
1 gostou•148 visualizações
Invited talk @ Talk a Bit'19
Recomendados
Recomendados
Mais conteúdo relacionado
Mais procurados
Mais procurados (14)
Semelhante a Let's Talk a Bit About: Green Software
Semelhante a Let's Talk a Bit About: Green Software (20)
Mais de GreenLabAtDI
Mais de GreenLabAtDI (9)
Último
Último (20)
Let's Talk a Bit About: Green Software
- 1. Let’s talk a bit about: Green Software greenlab.di.uminho.pt | ruipereira@di.uminho.pt Rui Pereira
- 3. Global energy system is unsustainable 3
- 4. Green Computing 4 Caught the attention of many companies allowing them to save: close to 50% of the energy costs of an organization can be attributed to the IT departments [PICMET, 2009] up to 90% of excessive energy used by hardware can be attributed to inefficient software [The GHG Protocol Report, 2013]
- 5. Green Computing 5 Everyday users also feel the impact 9 out of 10 users have low battery anxiety; long battery life was the most desired feature in 2017 [The Wall Street Journal, 2018]
- 6. Green Software Reducing energy consumption through software analysis and optimization Problem: How to analyze How to interpret How to improve 6
- 7. Green Software Problems (extended to programmers): How to analyze How to interpret How to improve 7 What do programmers know about software energy consumption? [IEEE Software’16] An empirical study of practitioners’ perspectives on green software engineering [ICSE’16] Mining questions about software energy consumption [MSR’14]
- 8. Green Software Two main problems for energy efficient software development: 8 Energy efficiency: a new concern for application software Developers [Communications of the ACM’17] Lack of knowledge Lack of tools
- 9. Energy vs. Power 9 Power (w) – rate (or effort) at which that work is done Energy (J) – amount of work done Power can change constantly while Energy is the accumulation Energy = Power x Seconds Power Energy 100W360,000 J = x 3,600s
- 10. Is faster greener? 10 slower Is faster greener? Energy-aware software: Challenges, opportunities, and strategies [Journal of Computational Science’13] An experimental survey of energy management across the stack [OOPSLA’14] Understanding energy behaviors of thread management constructs [OOPSLA’14] Power and energy implications of the number of threads used [PPMG’15] Using the greenup, powerup, and speedup metrics to evaluate software energy efficiency [IGSC’15] Haskell in green land: Analyzing the energy behavior of a purely functional language [SANER’16]
- 12. Energy Efficiency Across Programming Languages 12
- 13. 13
- 14. Understanding the energy efficiency of programming languages 14 Is a faster language always a more energy efficient one? Which are the most energy efficient languages? How much energy consumption is attributed to CPU? What paradigms are most energy efficient? How much energy does memory usage consume?
- 15. Comparing Programming Languages 15 C Java Haskell Fortran = = = * Formerly known as The Great Computer Language Shootout
- 17. 17 Normalized Results Average Joules ms Compiled 120 5103 VM 576 20623 Interpreted 2365 87614 Average Joules ms Imperative 125 5585 OO 879 32975 Functional 1367 42740 “Scripting” 2320 88322 57J 4604J 2s 167s
- 18. 18 Normalized Results Average Joules ms Compiled 120 5103 VM 576 20623 Interpreted 2365 87614 Average Joules ms Imperative 125 5585 OO 879 32975 Functional 1367 42740 “Scripting” 2320 88322
- 19. 19 Normalized Results 66Mb 1309Mb Average Joules ms Mb Compiled 120 5103 125 VM 576 20623 285 Interpreted 2365 87614 426 Average Joules ms Mb Imperative 125 5585 116 OO 879 32975 249 Functional 1367 42740 251 “Scripting” 2320 88322 421
- 21. 21 Energy vs. Memory (Peak Memory) Spearman rank-order correlation p = 0.2091 +1 = perfect positive -1 = perfect negative0 = no relationship 0.3 = weak uphill p = 0.2091 (Continuous Memory) Spearman rank-order correlation p = 0.744 +1 = perfect positive -1 = perfect negative0 = no relationship 0.7 = strong uphill p = 0.744
- 22. 22 Energy vs. Time vs. Memory
- 23. 23 Energy vs. Time vs. Memory (Pareto Optimization)
- 24. 24
- 25. Energy Efficiency Across Java Collections 25
- 26. Understanding the energy efficiency of Java collections 26 Which collections are the most energy inefficient? Can we use our data to suggest energy efficient alternatives? Can we quantify energy consumption of collections and their methods?
- 27. Java Collections Framework (JCF) 27 ConcurrentSkipListSet ArrayList ConcurrentHashMap CopyOnWriteArraySet AttributeList ConcurrentSkipListMap HashSet CopyOnWriteArrayList HashMap LinkedHashSet LinkedList Hashtable TreeSet RoleList IdentityHashMap RoleUnresolvedList LinkedHashMap Stack Properties Vector SimpleBindings TreeMap UIDefaults WeakHashMap Sets Lists Maps Design – Data Structures Population sizes: 25,000 250,000 1,000,000
- 28. Java Collections Framework (JCF) 28 Design – Methods add add clear addAll addAll containsKey clear add(index) containsValue contains addAll(index) entrySet containsAll clear get iterateAll contains iterateAll iterator containsAll keySet remove get put removeAll indexOf putAll retainAll iterator remove toArray lastIndexOf values listIterator listIterator(index) remove removeAll remove(index) retainAll set sublist toArray Sets Lists Maps
- 29. Results (Lists - 25k population) 29
- 30. jStanley: Placing a Green Thumb on Java Collections 30 Between 2% - 17% energy reduction Average of 6.7% energy reduction https://greensoftwarelab.github.io/jStanley/
- 32. Energy efficiency of apps is important 32 Analyzed 1,780+ Android and iOS open source applications Reviewed Git commits, issues, and PR for “energy”, “power”, or “battery” (6,000+) Manual analysis of the changes Catalog of 22 Energy Patterns https://tqrg.github.io/energy-patterns/
- 33. Mobile Energy Patterns (3 of 22) 33
- 34. Where do I begin? 34
- 35. Quick changes 35
- 36. How do I measure energy? 36 Qualcomm Trepn Profiler Monsoon ($$$) ODroid XU-3 ($) Intel Running Average Power Limiter (RAPL) ODroid XU-3 ($)
- 38. 38 Media coverage: Featured in more than 30+ websites
- 39. 39 28,000+ unique devices 1,000M+ samples
- 40. 40 Explore the dataset: https://greenhub.di.ubi.pt/storage/dataset.zip
Notas do Editor
- Patterns, habits, structures algorithms
- Aware of the energy consumption problem, many times seeking help in resolving this, misconceptions within the programming community as to what causes highenergy consumption, how to solve them a heavy lack of support and knowledge for energy-aware development
- Most comment perception to the EC problem = performance issue (reduction in time = reduction in energy) I would suggest this as a good starting point if someone were to ask me how to optimize, as do others. But the opposite has been also observed, where an increase in time brought about a decrease in energy.
- Developing comparable programs written in different languages is both complex and time consuming. 27 languages and 10 tasks
- Top 5 keep their ranks. OCAML HASKELL RACKET PYTHON Scriptings
- OCAML HASKELL RACKET PYTHON
- optimizing for energy is not necessarily the same as optimizing for performance. Pascal is 10% more energy efficient than Chapel, yet Chapel is 55% faster.
- Energy harvesting device Wearable device we use a multi-objective optimization algorithm to sort these languages, known as Pareto optimization
- we use a multi-objective optimization algorithm to sort these languages, known as Pareto optimization