Theory And Mechanism Of Graded Service Substitution In SOA

SOA (Service-Oriented Architecture), as an established paradigm for building software and one of the keystones of technologies in cloud computing, has obtained considerable attention from both research and practice area. As SOA matures, the field of available services providing the same and/or similar functions is growing and evolving at an ever increasing pace. With loosely-coupled feature, service substitu-tion materializes.Service substitution means, in the context of services composition, the original service is replaced with a new one without breaking some essential properties of service system, e.g., compatibility. Owing to the significance in repairing faults, improving performance and enhancing survivability, the theory and mechanism of service substitution has become a scientific problem in SOA.Facing the great and increasing number of available services with the same and/or similar functions, one of the key challenges is to make full use of them. Specifically, it is necessary, in the functional aspect, to classify them from weak to strong function:as for services with the same functions, to categorize them in terms of the non-functional properties in order to satisfy the different usage patterns.This paper focuses on the theory and mechanism of service substitution, and studies the following issues:Firstly, the primary problem is how to measure the correctness of graded ser-vice substitution. Service compatibility is considered as the criteria of correct service substitution. Process algebra (spec.Ï€-calculus) is used to describe service contract and model the messaging behaviours in services and their composition. The analysis of service composition and substitution is considered as a problem of model check-ing, and discussed from three perspectives, i.c., types of component services (isolat-ed/interactive), settings of service composition (open/self-contained) and focused problems (global/local properties). Further, the different specifications of service compatibility is discussed, and a spectrum of service compatibility is proposed.Secondly, one of the main challenges for service substitution, owing to au-tonomous services, is the opaque internal behaviour of services in the form of WSDL. In this setting, the key obstacle is to determine the correctness of service substitu-tion only depending on the limited information from WSDL. Based on the may-, should- and must- testing pre-orders, service compatibility is specified in terms of parts of computing paths as well as all of paths. Types of service interfaces are used to specify the patterns of exchanging massages. Equivalent interfaces have the same type of interfaces and the same set of outgoing and incoming messages, based on which equivalent contracts satisfy the equivalent service substitution. Subtype of interfaces has the more capacity of receiving messages and the less choice of sending messages, based on which subcontract ensures the graded service substitution.Thirdly, a handicap of successful service substitution in the case of SOA sur-vivability is the difficulty to rapidly identify a service which is full equivalent to the original one. Allowing degraded forms of substitutions is a possible way to over-come this obstacle. In this setting, the key of problem is the theory of substitution preserving service compatibility. Services advertising protocols (c.g., in the form of BPEL) are studied. Specifically, the following challenges are tackled:(1) Identifying changes in a component service that do not affect the essential service compatibility: (2) If service compatibility is affected, to determine whether these changes allow a given system, of which the service is a component, to continue offering at least a subset of its functionality (i.e., to continue working in a degraded form).Lastly, to aggregate QoS of service composition palys a key role in analysing non-functional properties in graded service substitution, in which the challenge is to compute QoS for composite services with complicated structures. The process structure tree is used to decompose and analyse orchestrations of service composition whose structures are specified by formal notation. Using the concepts of run and configuration, structure specified as directed acyclic graph is transformed into a choice one including several N-Structures. For structures of arbitrary loops, the average times of executing component is computed by using the theory of Markov chain. The overall QoS for a composite service is computed by aggregating QoS of its components according to their structure and relation, i.e., maximally-structured representation. To this end, the process structure tree is traversed in pre-order, i.e., computing the aggregate QoS from leaf nodes up to the root node.The main contributions of this research include:1. specifications of service compatibility and a spectrum of service compatibility from weak to strong.2. notions of service compatibility based on the may-, should- and must- testing pre-orders, and an extension of conventional substitution theory to deal with both strict and degraded substitution.3. theory and mechanism supporting both graded substitution for services in the form of WSDL by using the limited information of interfaces and degraded substitution for services in the form of BPEL to enhance the survivability.4. method for the aggregate QoS of composite services using process structure tree, which can terminate quasi-linearly.5. aggregation of QoS for two complicated structures of service composition, i.e., Parallel with Asymmetric Synchronous and Single-Entry-Multiple-Exit Loop.