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
SMART_READER_LITE
LIVE PREVIEW

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-1
SLIDE 1

Faculty of Computer Science Software Engineering Chair

www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude

Stefan Staude, Markus Frank, Marcus Hilbrich

TU Chemnitz Faculty of Computer Science Software Engineering Chair

Is the PCM Ready for ACTORs and Multicore CPUs?

A Use Case-based Evaluation

slide-2
SLIDE 2

www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude

Prediction Multi-Core Single-Core Software-Performance-Prediction Software Engineer Model

2/12

[Frank16] [Lehrig16] [Becker14]

slide-3
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
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
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
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
SLIDE 7

www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude

Palladio-Model

SEFF ExperimentHandler:

7/12

[...]

slide-8
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
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
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
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
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
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
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
SLIDE 15

www.tu-chemnitz.de SSP Karlsruhe ∙ 15.11.17 ∙ Stefan Staude

[Frank16] – M. Frank and M. Hilbrich „Performance Prediction for Multicore Environments – An Experiment Report“ [Lehrig16] – S. Lehrig and S. Becker „Using Performance Models for Planning the Redeployment to Infrastructure-as-a-Service Environments: A Case Study“ [Becker14] – M. Becker, M. Platenius and S. Becker „Cloud Computing Reduces Uncertainties in Quality-of-Service Matching!“