An important approach to design a composition of services is to do so not from the perspective of one party, but from a representation of interactions between two or more autonomous parties. Protocols are specifications of interactions that assiduously disregard the implementations of the services, but describe the interactions: protocols thus promote interoperability among heterogeneous and autonomous services. 23 Newer approaches for protocols accommodate data representations, but additional research is needed to bridge the gap between interaction and data. An important direction is to formalize sociotechnical interactions as a basis for governance of autonomous services. 20

Service composition. Future research in service composition should target the following topics:

Large-scale Web and cloud service composition. Service composition research should extend to non-WSDL-described services or services described by plain text. As an example, ProgrammableWeb.com hosts over 10,000 API services and over 6,000 service mashups, with a great majority described in plain text. Web information extraction, natural language processing, data and text mining, collaborative tagging, and information retrieval technologies can be used to extract useful semantics, group relevant services, and detect novel composition patterns. Nevertheless, short service descriptions comprised of limited terms coupled with the diverse naming conventions used by service providers pose novel challenges that demand technological innovations to advance the state of the art in both service computing and all relevant domains. Furthermore, a software system may need to evolve during its life cycle and be able to handle the changes in its running environment and the increasing complexity of its workflows. Self-adaptive software evaluates its behavior and makes adjustment according to the evaluation result to address issues and improve performance.

A cloud-computing environment
provides an attractive option for de-
ploying services, because of the poten-
tial scalability and accessibility it of-
fers. However, it introduces problems
related to:
incorporating possibly billions of
things, and harnessing their data and
functionality to provide novel smart
services that benefit enterprises, in-
dustries, and our society. The funda-
mental IoT challenges as related to
service computing are:

1. Continuously maintaining cyber personalities and contexts for Io T devices. In particular, Io T things need to have Web identities and Web representations (for example, Web proxies) that reflect their physical spaces. They also need to connect and communicate within social, environmental, user-centric, and application contexts, and such contexts need to be maintained and managed.

2. Continually discovering, integrating, and (re)using Io T things and their data. Specifically, the Io T environment is a federated environment where things and their data, cloud services, and IT services (for example, for data analysis and visualization) are often provided by independent providers with diverse interfaces, as well as business, cost, and QoS models. To provide new Internet-scale services, the IoT must (re)use IoT things deployed by others and data collected by others for their own purpose.

Service Computing
Research Roadmap

Following the major challenges identified earlier, we now propose a research roadmap for service computing. The roadmap focuses on four emerging research areas, namely service design, service composition, crowdsourcing-based reputation, and Internet of Things.

Service design. Research should focus on laying the foundation for service design drawing from similar research in databases, software engineering, and distributed systems.

The design of service systems should
build upon a formal model of services
that enables efficient access to a large
service space with diverse function-
alities. The service model can sup-
port accessing services much as the
relational model supports accessing
structured data. As service composi-
tion is usually needed to fulfill com-
plex user requests, it is important for
the service model to support depen-
dency relationships among different
services and their operations. Similar
to the relational model, the service
model should enable the design of ser-
vice query algebra and calculus that
allow general users to declaratively
query multiple services in a transpar-
ent and efficient manner. In addition,
service query optimization should go
beyond just generating efficient query
execution plans. As a large number of
service providers may compete to of-
fer similar functionalities, the design
of the service model should support
the implementation of optimization
strategies for finding the “best” Web
services or composition thereof with
respect to the expected user-supplied
quality. In sum, a formal service mod-
el that meets the requirements noted
here will take center stage to provide
efficient and transparent access to
computing resources through servic-
es, which is a critical step to reach the
full potential of service computing.

Recent work25 has shown great promise in designing and using a formal service model to help users access service. A set of algebraic operators, defined based on a formal service model, provides standard ways to manipulate services as first-class objects. These include choosing specific operations from a service as well as quality-based service selection and mashup operators across multiple services to form a composed service. The implementation of these operators enables the generation of service execution plans that can be directly used by users to access services. Due to functional dependency and quality dimensions, a service model is more complex than a relational data model.

In particular, three key features of services are crucial: functionality, behavior, and quality. Functionality is specified by the operations offered by a service; behavior reflects how the service operations can be invoked, and is decided by the dependency constraints between service operations; quality determines the non-functional properties of a service. A viable service query language should allow users to locate and invoke their desired functionalities, select the best service providers that meet their quality requirements, and generate service compositions when functionalities from multiple services are required.