Article Preview
TopIntroduction
With the development of Service Oriented Computing (Papazoglou et al., 2008), web service technology has attracted much attention from industry and academia in recent years and achieved significant success. Service composition is one of the most important issues in SOC, and has become a main technology enabler for delivering cloud solutions with its ability of enabling the interoperation of heterogeneous systems, reuse of distributed functions in an unprecedented scale and creating value-added business applications (Jiang et al., 2012; Deng et al., 2013).
The explosion of data and information has been recognized in recent years and people have stepped into the age of big data. Generated data is growing too fast to store and handle as before. Emerging cloud-based infrastructures for storage have been widely accepted as the next-generation solution to address the data proliferation and the reliance on data (Kolodner et al., 2012). Data-intensive services are now being developed and utilized in more and more fields (Deng et al., 2013). The emergence of cloud computing and data-intensive services brings both huge changes and new challenges for Service Oriented Computing technology. Traditionally, service providers usually deployed services on their own local infrastructures. Due to the explosion of data and economic benefits, nowadays service providers prefer to deploy their services in cloud. Generally, these services are deployed in the data centers of cloud providers (e.g. Amazon, Google). For traditional service compositions, the time cost of data transmission among component services is negligible compared to the execute time of component services. So most research focuses more attention on how to effectively make service composition, but rarely considers the problem of deploying service composition. However, the data transmission among components in data-intensive service compositions is incredibly huge. If the component services are deployed independently and casually, the latency of the whole composition would be incogitable. Hence, it is essential to make an optimal deployment strategy for data-intensive service compositions for reducing latency of user requests.
At present, limited work has focused on the problem of deploying service compositions. The work in Kang et al. (2012) may be a little similar to ours, but they aimed at deploying multiple correlated services in cloud. However, service composition is much more complicated because the component services in compositions may be organized in different structures (Rao & Su, 2005). And there are complex dependency relationships (data/logic dependency) among component services. Dependency relationships, especially data dependency relationships, have a great influence on the overall performance of data-intensive service compositions. In this paper, we mainly consider the following factors to tackle the issue of deploying data-intensive service compositions:
- 1.
Mass Data Transmission: For a data-intensive service composition, data transmission across different data centers is inevitable. Besides, the amount of transmitted data is quite huge. Hence, it requires taking consideration of the network bandwidths between datacenters and the transmitted data size comprehensively to make an optimal deployment schedule and reduce the overall latency of the composition;
- 2.
Dependency Relationships: The component services in a data-intensive service composition may have multiple logic/data dependency relationships with each other. The components having dependency relationships should be deployed as close as possible. It can reduce the cost of data transmission;
- 3.
Load Balance: To make an optimal deployment schedule, we should also consider maintaining a relatively load balancing of data centers.