20.01.2015 Views

Performance Modeling and Benchmarking of Event-Based ... - DVS

Performance Modeling and Benchmarking of Event-Based ... - DVS

Performance Modeling and Benchmarking of Event-Based ... - DVS

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

5.2. CASE STUDY I: SPECJMS2007 103<br />

CPU Utilization<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

CPU Time Per Message (ms)<br />

0.4<br />

0.35<br />

0.3<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

BASE (Vertical)<br />

BASE (Vertical)<br />

Figure 5.21: Measurement Results for Vertical Experiments<br />

100<br />

12000<br />

CPU Utilization<br />

80<br />

60<br />

40<br />

20<br />

Messages Sent Per Sec<br />

11000<br />

10000<br />

9000<br />

8000<br />

7000<br />

6000<br />

0<br />

500 550 600 650 700 750 800 850 900 950<br />

BASE (Customized Vertical)<br />

5000<br />

500 550 600 650 700 750 800 850 900 950<br />

BASE (Customized Vertical)<br />

Messages Received Per Sec<br />

12000<br />

11000<br />

10000<br />

9000<br />

8000<br />

7000<br />

6000<br />

5000<br />

500 550 600 650 700 750 800 850 900 950<br />

BASE (Customized Vertical)<br />

CPU Time Per Message (ms)<br />

0.4<br />

0.35<br />

0.3<br />

0.25<br />

0.2<br />

0.15<br />

0.1<br />

0.05<br />

0<br />

500 550 600 650 700 750 800 850 900 950<br />

BASE (Customized Vertical)<br />

Figure 5.22: Measurement Results for Customized Vertical Experiments with P2P Messaging<br />

1, 4 <strong>and</strong> 5), whereas the second one runs only pub/sub interactions (i.e., 3, 6 <strong>and</strong> 7) 5 . In both<br />

cases, the relative interaction mix for the considered interactions is the same as for the st<strong>and</strong>ard<br />

vertical topology. Figures 5.22 <strong>and</strong> 5.23 show the measurement results. We can see that, as<br />

expected, the pub/sub portion <strong>of</strong> the workload is by far much more light-weight than the P2P<br />

portion. This is due to two reasons. On the one h<strong>and</strong>, for a given value <strong>of</strong> the BASE, the P2P<br />

message traffic injected is much larger than the pub/sub traffic according to the definition <strong>of</strong><br />

the vertical topology presented in Section 5.1.4. On the other h<strong>and</strong>, the server overhead per<br />

5 Note that Interaction 2 is not part <strong>of</strong> these workloads since it contains a mix <strong>of</strong> both P2P <strong>and</strong> pub/sub<br />

messaging.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!