Interaction patterns based checkpointing of choreographed web services

Abstract

Complex business processes can be realised by composing web services already available on the web. Choreography of web services describes how its constituent services have to interact with other to achieve business goals. In order to make a choreographed web service resilient to transient faults, we investigate in this paper the idea of applying checkpoints to web services. We do this in two steps: firstly, we identify patterns in which choreographed web services interact with each other and propose complexity metrics for these interaction patterns. Next, we propose pattern based checkpointing policy that identifies checkpointing locations in the given choreography. We also provide formal specification to the proposed patterns. We demonstrate the usability of our approach through the development of a tool that automatically inserts checkpointing locations in the given choreography specification.

Survey on Web Services Fault Tolerance Approaches Based on Checkpointing Mechanisms

“…• At design time, they use the choreography document and introduce checkpoint locations at places where non-repeatable actions take place [54].…”

“…Vathsala et al identify the most appropriate checkpointing locations by means of their model, where they model the choreography composition as a set of interaction patterns. An introduction to this approach can be found in [20] and details can be found in [54]. Therefore, they stipulate that by making use of QoS values of services, they always met the execution times and cost of service constraints.…”

Correctness Control of Privacy-Preserving Decentralized Constrained Web Service Compositions Through Auditing

“…This makes it hard to maintain a reliable environment in the sense that colluding web services would continue their dishonest behavior, and possibly enlarge its scope, as this will have no impact on their functionality or reputation. There are some proposals to enforce control in choreographed compositions, but only regarding some nonfunctional requirements such as execution time and cost [4], or ensuring resilience to faults by detecting and correcting them during the deployment process using pattern based check-pointing [20]. Moreover, there are some works that considered the computation of reputation scores for web services, but mainly through feedback from end users, either explicitly solicited or predicted based on historical rating (e.g., the work in [23]).…”