Database Reference
In-Depth Information
300
Test dataset 1
Test dataset 2
250
200
150
100
50
0
0
50
100
150
200
250
300
350
400
Number of concrete services for each abstract service
Fig. 8.10
Time vs Concrete services
When the number of abstract services increases, the optimal fitness value obtained
from GA based approach also increases. This is because population size and other
related variables stay the same when the number of the abstract services varies.
Hence, GA based approach are more scalable and efficient than IP approaches.
8.4
Related Work
Most composition approaches in SOC use linear programming methods. [
238
]
presents two approaches: one focuses on local optimization, the other on global
optimization. They use integer programming to solve the global optimization
problem. The limit of this approach is that all QoS attributes need to be linearized as
integer programming is a linear programming approach. [
66
] proposes an improved
approach based on [
238
], using Mixed Linear Programming (MILP) approach. They
also introduce several concepts such as loop peeling and negotiation mechanisms to
address situation where no feasible solution can be found. [
63
] proposes an approach
to decompose global QoS constraints into local constraints with conservative upper
and lower bounds. These local constraints are resolved by using an efficient
distributed local selection strategy.
All of the aforementioned approaches only consider the service composition
problems in small-scale scenarios. These linear programming approaches are not
suitable to handle large-scale scenarios problems, e.g. service composition in cloud
