Database Reference
In-Depth Information
computing. [
93
] was the first to use GA for optimization of QoS-aware compositions
in SOC. The results show that their GA implementation scales better than linear
programming. [
135
] presents a GA and a Culture Algorithm (CA) for Web service
compositions. The first algorithm is similar to [
93
], the latter uses a global belief
space and an influence function that accelerate the convergence of the population.
[
202
] presents a mutation operator which consider both the local and global
constraints to accelerate the converge of the population.
Existing GA-based approaches are solely focus on service composition in
application level, which do not consider the computing resources composition.
Service composition in cloud computing involves application service composition
and computing resources matching and scheduling. In this chapter, a genetic
algorithm based approach is proposed to compose services in cloud computing, by
combining QoS-aware service composition approaches and resources matching and
scheduling approaches.
8.5
Conclusion
A genetic algorithm based approach is presented for service compositions in cloud
computing. Service compositions in cloud computing involve the selections of
application services and utility computing services. The chromosome size is bound
to the number of n of abstract services. The number of possible application services
and utility computing services only augments the search space. For small-scale
scenarios, the proposed approach finds optimal solutions. For larger-scale problems,
it outperforms the integer programming approach. This is a beginning to propose
robust service composition approaches in cloud computing. Future work may focus
to eliminate several assumptions: (1) QoS values for each component are known in
this research. Calculating the QoS values at runtime is one direction; (2) penalty
factor in the fitness function is static. More dynamic fitness functions can be used
to improve the performance of the approach. (3) novel crossover and mutation
operators may accelerate the converge.
