Development of Cloud Applications in Hybrid Clouds with Support for Multi-scheduling

Development of cloud applications must consider many aspects inherent in the distributed nature of clouds, mainly those related to elasticity, high access level to computational resources, multi-tenant behavior, transparency, pay-per-use model, and resource scalability. In addition, portability is a key feature that must be present in any development framework to allow extensions and simplify resource sharing by standardized interfaces. Open source approaches can be used, but the model must be composed of independent parts to optimize the availability of active components in the infrastructure. Hybrid cloud models are interesting because widely acceptable solutions can be developed without “reinventing the wheel.” Private clouds are more suitable for keeping restricted data or supporting services of small enterprises or institutions. However, their infrastructure must offer alternatives to provide services outside their own domain. In this context, a private cloud can use frameworks of public clouds and aggregate services to support the development of new applications. This generally occurs in PaaS models, where the platform offers pre-configured tools to interact with services of other domains. Security issues must also be considered at all stages of development, as most of the communication takes place among services located in different domains, linked by Internet connections. Solutions such as OpenID guarantee that public cloud services are used for the purpose of authentication, but additional security features in the source domain must be assured. In this chapter, a development framework is presented to guide the development of widely acceptable cloud applications, following standardized open source solutions. This framework, originally developed for a robotic environment, can be extended to support other cloud environments. The study presents aspects related to multi-scheduling of virtual machines and suggests how virtualized applications can be developed with different methodologies, such as dynamic IP, Web service with SOAP communication, MapReduce approach, and OCCI-based infrastructure.