Information Technology Reference
In-Depth Information
Table 2.
Improved algorithm parallelized and distributed
case
processes time[s] speedup eciency
sisim
single node
1
9240.11
1.00
1
7522.86
1.23
123%
sisim-gfs
single node
2
3909.90
2.36
118%
3
2636.28
3.50
117%
4
2051.63
4.50
113%
4
1994.82
4.63
116%
sisim-gfs
distributed
8
1005.58
9.19
115%
12
674.75
13.69
114%
16
516.15
17.90
112%
24
388.29
23.80
99%
32
307.39
30.06
94%
32 x
sisim
sisim-gfs
24 x
One
workstation
Multiple
workstations
16 x
8 x
2
4
8
12
16
24
32
Available resources [cores]
Fig. 6.
Speedup test of
sisim
and
sisim-gfs
with available resources
1 and 2). Using the described strategy is possible to run the same number of
simulationsinlessthan6minutes.
5 Conclusions and Future Work
The proposed strategy has successfully fulfilled the requirements presented in
section 2.1 and has been flexible enough to fit many embarrassingly parallel algo-
rithms. Results show that using the proposed software architecture and available
computational resources (local computer network, clusters or cloud services),
