Universidad Nacional del Centro (UNICEN) Institute of Software Technology ISISTAN Research Institute Reliable Software Systems Software Engineering Group Using the CoCoME We already community case study attended for evaluating the SSP ’16... SQuAT architecture … and SSP ‘17. optimization approach: results and experiences Sebastian Frank Santiago Vidal
SQuAT Case Study Experiences Results Future Work SQuAT Approach Motivation The system We must process If the database must not fail if 10 million changes , we a server goes transactions per must be prepared • Difficulties to manage and optimize down sec. all QAs in a satisfying way, i.e., make Modifiability Reliability Performance Expert Expert Expert everybody happy • Monolithic analysis is unnatural • Trade-offs between different QA goals & negotiation among experts Product in software architecture Client Manager We have I want the system to be deadlines and fast , adaptable , we are running fail-proof , secure and out of money cheap Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 2
SQuAT Case Study Experiences Results Future Work SQuAT Approach Scenario (Goal) The Bots Abstract Model Analyzer Bot Model Optimizer Modifiability Bot Performance Bot Workload, Usage Behaviour, Add, Remove, Modify CPU Clock Rate Components & Interfaces Headless Palladio LQN KAMP Extension Solver (Change Impact Analysis) Headless PerOpteryx Henshin Rules (Evolutionary Algorithm) Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 3
SQuAT Case Study Experiences Results Future Work SQuAT Approach Local search The algorithm P M M P Next Level Palladio Component P M P M Model Initial Proposed Candidate Candidates Global search Analyze Optimize Negotiate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 4
SQuAT Case Study Experiences Results Future Work Case Study CoCoME See SSP ‘17 and/or “ Distributed Quality-Attribute Case Study: ST+ Case Study: CoCoME Optimization of Software Architectures ” (SBCARS) About Business Trip Management Trading System (supermarket) Size 5 Components 51 Components 4 Alternative Components 5 Alternative Components Scenarios 2 Performance Scenarios 4 Performance Scenarios 2 Modifiability Scenarios 4 Modifiability Scenarios Levels 2 5 Quality Fully Modeled Partially Modeled (~25%) “ Clean ” “ Dirty ” Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 5
SQuAT Case Study Experiences Results Future Work Case Study Scenarios in the CoCoME Case Study Name Stimulus Expected Response P1 Overall Usage Increase +10% < 1.2 seconds P2 Overall Usage Increase +50% < 1.4 seconds P3 Server Fault (1 out of 2 in a cluster) < 1.0 seconds P4 Usage Increase of Specific Service < 2.4 seconds M1 Add NFC Payment Method < 2270 (complexity) M2 Add Premium User Service < 750 (complexity) M3 Add Withdraw Money Function < 170 (complexity) M4 Add Logging Function < 2180 (complexity) Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 6
SQuAT Case Study Experiences Results Future Work Experiences Model: Preparations • Model Allocation, Usage Behaviour,... • Understanding CoCoME • Call Graph Tool • Side Effect: Bug Fixing Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 7
SQuAT Case Study Experiences Results Future Work Experiences Model: Limitations Replace What are “Events”? - Events - Composite Components What are “Composite Components”? Proposed Initial Candidates Candidate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 8
SQuAT Case Study Experiences Results Future Work Experiences Model: Bad Practices Empty I can replace this Interface component with an Component without arbitrary one. Interface Error Proposed Initial Candidates Candidate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 9
SQuAT Case Study Experiences Results Future Work Experiences Model: Validity Error Sometimes my transformations I will filter fail them out! Proposed Initial Candidates Candidate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 10
SQuAT Case Study Experiences Results Future Work Experiences Seed Levels: 5 Selection Scalability Partial Parallelization Wait some minutes ! See you in Size: Big 3-5 days ! Filter population. See you in 25 Wait some days ! seconds ! Proposed Initial Candidates Candidate Model Caching SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 11
SQuAT Case Study Experiences Results Future Work Experiences Again! CPU Clock Rate Ripple Effect Let’s use it! Again! Max. CPU Clock Rate Factor: 2x 4 GHz >1000 GHz I like your I don’t candidates! care Proposed Initial Candidates Candidate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 12
SQuAT Case Study Experiences Results Future Work Experiences CPU Clock Rate Ripple Effect Model Optimization 1. Step: Automatically Generate Headless Call Configuration PerOpteryx & Optimize Candidate Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 13
SQuAT Case Study Experiences Results Future Work Experiences CPU Clock Rate Ripple Effect Model Optimization I have a new parameter 1. Step: Automatically Call Generate Configuration 2. Step: Headless Modify Configuration PerOpteryx 3. Step: Optimize Call Candidate Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 14
SQuAT Case Study Experiences Results Future Work Experiences New Utility Function Utility Function My scenarios are satisfied Worsen performance is more beneficial than improving modifiability What about me? Proposed Initial Candidates Candidate SQuAT Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 15
SQuAT Case Study Experiences Results Future Work Results M1 - P1 Initial • Many “good” agreements Candidate • satisfy Performance Scenario 1 • satisfy Modifiability Scenario 1 • Higher levels are useful Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 16
SQuAT Case Study Experiences Results Future Work Results Initial Candidate More Results M2 - P2 M3 - P3 M4 - P4 Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 17
SQuAT Case Study Experiences Results Future Work Future Work Lessons Learned • Model • Quality (“Bad Practices”) • Exception Handling Strategies • Filters, ... • Visualization • Understand Changes • Compare Models • Interaction • New Insights at Runtime • Human Knowledge Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 18
SQuAT Case Study Experiences Results Future Work Future Work Visualization & Interaction Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 19
SQuAT Case Study Experiences Results Future Work Future Work More Ideas • Personalization of Bots • Own Utility Function, ... • Scalability • Microservices • Further Parallelization, Caching, ... • Usability • Natural Language • Scenarios & Results • Machine Learning Using the CoCoME community case study for evaluating the SQuAT architecture optimization approach: results and experiences 11/08/2018 20
Recommend
More recommend
