Weihai Yu

Scalable Services Orchestration with Continuation-Passing Messaging

Services orchestration is typically supported with the traditional centralized workflow technology. The central engine becomes an intensive resource when the number of services to be orchestrated becomes large. Existing decentralized approaches, being primarily based on static process partitioning, exhibit some other shortcomings such as in resource utilization and fault handling. We present an approach to decentralized services orchestration that does not have these shortcomings and scales well for a large number of services. The approach is based on a new messaging scheme, continuation-passing messaging, where continuations, or the reminder of executions, are passed along with messages for services orchestration. The approach also supports services recovery, by automatically generating recovery plans into a second continuation called failure continuation. Our experimental results show that the new approach provides much better scalability than traditional centralized ones.

Continuation-passing enactment of distributed recoverable workflows

Scalability, reliability and adaptability are among the key requirements for the enactment of distributed workflows. In addition, system resources should be efficiently utilized. Central workflow engines and static workflow instantiation are some of the important obstacles to meeting these requirements. We propose a fully decentralized approach to workflow enactment that is not subject to these obstacles. In addition, it supports automatic recovery. The approach is of continuation-passing style, where continuations, or the reminder of the executions, are passed along with asynchronous messages for workflow enactment. Two continuations are associated to an execution: a success continuation and a failure continuation. Recovery plans for workflows are automatically generated and included in failure continuations at runtime. A prototype is implemented.

A Dynamic Two-Phase Commit Protocol for Adaptive Composite Services

Next-generation applications based on Web services impose additional requirements on the use of coordination protocols with various optimizations, such as the two-phase commit protocol (2PC). This article analyses the well-known 2PC optimizations “presumed commit†and “presumed abort,†and presents an improved 2PC that is suitable for Web-service-based applications. More specifically, the protocol allows every individual service provider to choose dynamically the most appropriate presumption for any distributed transaction. This new capability is especially useful when a composite Web service is integrating component services that make different presumptions in their commit protocols. The protocol does not introduce extra overhead to the previous 2PC variants in terms of number of messages and log records, and it is easy to understand and realize. Our simulation shows that the choice of appropriate presumption has significant influence on system performance, and that in some heterogeneous settings, combining different presumptions in individual transactions outperforms adopting only one single presumption.

A string-wise CRDT for group editing

Real-time group editing has been envisioned as an important application for group collaboration. Operational transformation (OT) has been the concurrency control mechanism for group editing, due to its potential for high responsiveness to local editing operations. OT algorithms are generally very sophisticated and computation intensive. Recently, commutative replicated data types (CRDT) have appeared as viable substitutes of OT. Existing OT and CRDT work suffers from serious limitations. This note presents a CRDT that addresses some of these limitations.

Decentralized Orchestration of BPEL Processes with Execution Consistency

Scalability, consistency and reliability are among the key requirements for orchestration of BPEL processes. We present a fully decentralized approach to orchestration of BPEL processes that is of continuation-passing style, where continuations, or the reminder of the executions, are passed along with asynchronous messages for process orchestration. Furthermore, we identify and address some consistency issues that are more challenging for decentralized orchestrations.

A CRDT Supporting Selective Undo for Collaborative Text Editing

Undo is an important feature of editors. However, even after over two decades of active research and development, support of undo for real-time collaborative editing is still very limited. We examine issues concerning undo in collaborative text editing and present an approach using a layered commutative replicated data type CRDT. Our performance study shows that it provides sufficient responsiveness to the end users.

Consistent and decentralized orchestration of BPEL processes

Scalability, consistency and reliability are among the key requirements for orchestration of BPEL processes. In addition, system resources should be efficiently utilized. We present a fully decentralized approach to orchestration of BPEL processes that achieves high scalability and supports automatic process recovery. The approach is of continuation-passing style, where continuations, or the reminder of the executions, are passed along with asynchronous messages for process orchestration. Furthermore, we identify and address two consistency issues that are more challenging for decentralized orchestrations.

Decentralised web-services orchestration with continuation-passing messaging

Traditional centralised approaches to orchestrating composite services are subject to serious limitations for orchestrating services in open web applications. Dealing with these limitations calls for decentralised approaches, which are themselves faced with a number of technical challenges, due primarily to lack of overview of dynamic process status. We present a new decentralised approach that meets some of these technical challenges. The approach is based on continuation-passing messaging. It facilitates dynamic runtime tasks like fault handling and process recovery. Our experimental results show performance advantage of the approach, both in normal executions and in case of service failures.

Supporting String-Wise Operations and Selective Undo for Peer-to-Peer Group Editing

Real-time group editing has been envisioned as an effective manner of collaboration. For years, operational transformation (OT) has been the standard concurrency control mechanism for real-time group editing, due to its potential for high responsiveness to local editing operations. OT algorithms are generally non-trivial to be error-free and are computation intensive. Recently, commutative replicated data types (CRDT) have appeared as an alternative to OT. The state-of-the-art OT and CRDT work still lacks the basic functionality found in single-user text editors. In particular, there is no published work that supports both string-wise operations and selective undo. This paper presents an approach that combines and extends OT and CRDT strengths. It is fully decentralized and supports string-wise editing operations and selective undo. Our performance study shows that it provides sufficient responsiveness to the end-users.

Fault handling and recovery in decentralized services orchestration

Today, orchestration of composite services is typically carried out by dedicated central engines. A central engine, however, can easily become a performance bottleneck. Furthermore, centralized orchestration is infeasible for the composite of services beyond enterprise boundaries. Decentralized orchestration approaches are therefore devised to overcome these drawbacks. Decentralized orchestration, however, is generally regarded as more complex for certain orchestration management tasks due to absence of centralized states. Among the challenging tasks are fault handling and recovery. This paper presents fault handling and recovery of composite services in a decentralized orchestration approach that is based on continuation-passing messaging. The experimental results show performance advantage of the approach even in presence of service faults.