SLIDE 1
Is the PCM Ready for ACTORs and Multicore CPUs? A Use Case-based - - PowerPoint PPT Presentation
Is the PCM Ready for ACTORs and Multicore CPUs? A Use Case-based - - PowerPoint PPT Presentation
Faculty of Computer Science Software Engineering Chair Is the PCM Ready for ACTORs and Multicore CPUs? A Use Case-based Evaluation Stefan Staude, Markus Frank, Marcus Hilbrich TU Chemnitz Faculty of Computer Science Software Engineering
SLIDE 2
SLIDE 3
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Research Questions
Q1: Is it possible to model a parallel system following the ACTORS approach with the PCM? Q2: How accurate are the simulated predictions compared to the real execution?
3/12
SLIDE 4
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Methodology Evaluation Measurement
Bank Transactions
Palladio-Model
1 2 3 4 4
Implementation Step Single-Core Multi-Core
4/12
SLIDE 5
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Experimentsetup
Bank Transactions
- Multiple Accounts with a positiv balance
- Transactions could fail and block each other
- Differing outcomes depending on the execution order
- Common use case for ACTORs
Accounts
- Acc. Actor
Transaction
Experiment Manager
- Src. Acc.
- Trg. Acc.
5/12
SLIDE 6
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Palladio-Model
Deployment Diagram:
6/12
Experiment Manager
- Src. Acc.
- Trg. Acc.
SLIDE 7
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Palladio-Model
SEFF ExperimentHandler:
7/12
[...]
SLIDE 8
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Evaluation
Reseach Question Q1:
- Each Actor must be modeled individually
- High manual effort
- No asynchrone message passing
Our Model:
- Very abstract
- Loses ACTORs characteristics
- Low homomorphism
7/12
SLIDE 9
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Measurement Execution Simulation Worker Threads Mean Time Speedup Mean Time Accuracy 1 33.88 s 1.00 33.22 s 0.98 2 15.99 s 2.12 16.71 s 0.96 4 7.48 s 4.53 8.41 s 0.86 8 6.01 s 5.63 4.25 s 0.70 16 5.89 s 5.75 2.18 s 0.37
Performance Results:
9/12
SLIDE 10
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Measurement Execution Simulation Worker Threads Mean Time Speedup Mean Time Accuracy 1 33.88 s 1.00 33.22 s 0.98 2 15.99 s 2.12 16.71 s 0.96 4 7.48 s 4.53 8.41 s 0.86 8 6.01 s 5.63 4.25 s 0.70 16 5.89 s 5.75 2.18 s 0.37
Performance Results:
9/12
SLIDE 11
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Evaluation
Reseach Question Q2:
10/12
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 5 10 15 20 25 30 35 40
Execution Simulation
Threads Mean Time in s
SLIDE 12
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Evaluation
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 2 4 6 8 10 12 14 16 18
Execution Simulation
Threads Speedup
Reseach Question Q2:
11/12
SLIDE 13
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Prediction Multi-Core Software-Performance-Prediction Software Engineer Model
12/12
Questions? Questions?
SLIDE 14
www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude
Evaluation
Reseach Question Q2:
1 2 4 8 16 0.2 0.4 0.6 0.8 1 1.2
Accuracy
Accuracy
Threads
SLIDE 15