Cloud services have been widely used to support tasks in business processes. A variety of services with differing types, brands, and quality of service (QoS) characteristics are available from various vendors. Additionally, companies also build their own private clouds to meet specific business requirements related to performance, privacy, and security. The problem of selecting and assembling appropriate services to support an organization’s multiple related business processes is very challenging. This problem also differs from traditional product/service selection problems because of the presence of business processes with non-sequential tasks and multiple, related business processes. The various QoS characteristics of services, the special requirements of some subtasks in the business processes, compatibility between cloud services, and the coordination of multiple business processes need to be considered when selecting appropriate services. This paper develops a multi-factor cloud service composition optimal selection (CSCOS) model to formalize the constrained combinatorial optimization problem and designs an improved differential evolution algorithm based on a constructive cooperative coevolutionary framework (C3IMDE) for solution. Experiments on synthetic data demonstrate that C3IMDE has better efficiency and stability than benchmark algorithms, especially for large-scale, multi-process collaborative optimization.

云服务已被广泛用于支持业务流程中的任务。不同的供应商提供了具有不同类型、品牌和服务质量 (QoS) 特征的各种服务。此外，公司还构建自己的私有云，以满足与性能、隐私和安全相关的特定业务需求。选择和组装适当的服务来支持组织的多个相关业务流程的问题非常具有挑战性。该问题也不同于传统的产品/服务选择问题，因为存在具有非顺序任务的业务流程和多个相关的业务流程。选择合适的服务时需要考虑服务的各种QoS特性、业务流程中某些子任务的特殊要求、云服务之间的兼容性以及多个业务流程的协调。本文开发了多因素云服务组合优化选择（CSCOS）模型来形式化约束组合优化问题，并设计了一种基于构造性合作协同进化框架（C3IMDE）的改进差分进化算法来求解。合成数据实验表明，C3IMDE比基准算法具有更好的效率和稳定性，特别是对于大规模、多进程协同优化。