next up previous
Next: Discussion Up: System Evaluation Previous: Local Desktop Client:

Results

The results for the test cases run on Platform A and B are shown in Table 1 and 2 respectively.


Table 1: Timing Results in Seconds for Experiments on Platform A
$n$ proc.: 1$^\ast$ 1 + 1 2 + 1 4 + 1 8 + 1 16 + 1
Load Op.: 41.00 16.63 8.64 4.20 2.17 1.09
Timeline: 0.35 0.07 0.07 0.06 0.07 0.07
Profile: 2.05 0.24 0.16 0.11 0.12 0.16
$^\ast$ sequential Vampir



Table 2: Timing Results in Seconds for Experiments on Platform B
$n$ proc.: 1$^\ast$ 1 + 1 2 + 1 4 + 1 8 + 1 16 + 1
Load Op.: 208.00 91.50 43.45 21.26 10.44 5.20
Timeline: 1.05 0.17 0.18 0.17 0.15 0.16
Profile: 7.86 0.82 0.44 0.25 0.16 0.14
$^\ast$ sequential Vampir


Figures 4 and 5, show the speedups derived from these timing results.

Figure 4: Speedups for the Parallel Analysis Server on Platform A
\includegraphics[width=\columnwidth]{fig_evaluation_speedups-cluster}

Figure 5: Speedups for the Parallel Analysis Server on Platform B
\includegraphics[width=\columnwidth]{fig_evaluation_speedups-o3k}


next up previous
Next: Discussion Up: System Evaluation Previous: Local Desktop Client:
2003-10-06