Prashant Doshi

Dynamic Workflow Composition: Using Markov Decision Processes

The advent of Web services has made automated workflow composition relevant to Web-based applications. One technique that has received some attention for automatically composing workflows is AI-based classical planning. However, workflows generated by classical planning algorithms suffer from the paradoxical assumption of deterministic behavior of Web services, then requiring the additional overhead of execution monitoring to recover from unexpected behavior of services due to service failures, and the dynamic nature of real-world environments. To address these concerns, we propose using Markov decision processes (MDPs) to model workflow composition. To account for the uncertainty over the true environmental model, and for dynamic environments, we interleave MDP-based workflow generation and Bayesian model learning. Consequently, our method models both the inherent stochastic nature of Web services and the dynamic nature of the environment. Our algorithm produces workflows that are robust to non-deterministic behaviors of Web services and that adapt to a changing environment. We use a supply chain scenario to demonstrate our method and provide empirical results.

Towards Automated RESTful Web Service Composition

Emerging as the popular choice for leading Internet companies to expose internal data and resources, Restful Web services are attracting increasing attention in the industry.While automating WSDL/SOAP based Web service composition has been extensively studied in the research community, automated RESTful Web service composition in the context of service-oriented architecture (SOA), to the best of our knowledge, is less explored. As an early paper addressing this problem, this paper discusses the challenges of composing RESTful Web services and proposes a formal model for describing individual Web services and automating the composition. It demonstrates our approach by applying it to a real-world RESTful Web service composition problem. This paper represents our initial efforts towards the problem of automated RESTful Web service composition.We are hoping that it will draw interests from the research community on Web services, and engage more researchers in this challenge.

Graphical models for interactive POMDPs: representations and solutions

We develop new graphical representations for the problem of sequential decision making in partially observable multiagent environments, as formalized by interactive partially observable Markov decision processes (I-POMDPs). The graphical models called interactive influence diagrams (I-IDs) and their dynamic counterparts, interactive dynamic influence diagrams (I-DIDs), seek to explicitly model the structure that is often present in real-world problems by decomposing the situation into chance and decision variables, and the dependencies between the variables. I-DIDs generalize DIDs, which may be viewed as graphical representations of POMDPs, to multiagent settings in the same way that I-POMDPs generalize POMDPs. I-DIDs may be used to compute the policy of an agent given its belief as the agent acts and observes in a setting that is populated by other interacting agents. Using several examples, we show how I-IDs and I-DIDs may be applied and demonstrate their usefulness. We also show how the models may be solved using the standard algorithms that are applicable to DIDs. Solving I-DIDs exactly involves knowing the solutions of possible models of the other agents. The space of models grows exponentially with the number of time steps. We present a method of solving I-DIDs approximately by limiting the number of other agents’ candidate models at each time step to a constant. We do this by clustering models that are likely to be behaviorally equivalent and selecting a representative set from the clusters. We discuss the error bound of the approximation technique and demonstrate its empirical performance.

Speeding up adaptation of web service compositions using expiration times

Web processes must often operate in volatile environments where the quality of service parameters of the participating service providers change during the life time of the process. In order to remain optimal, the Web process must adapt to these changes. Adaptation requires knowledge about the parameter changes of each of the service providers and using this knowledge to determine whether the Web process should make a different more optimal decision. Previously, we defined a mechanism called the value of changed information which measures the impact of expected changes in the service parameters on the Web process, thereby offering a way to query and incorporate those changes that are useful and cost-efficient. However, computing the value of changed information incurs a substantial computational overhead. In this paper, we use service expiration times obtained from pre-defined service level agreements to reduce the computational overhead of adaptation. We formalize the intuition that services whose parameters have not expired need not be considered for querying for revised information. Using two realistic scenarios, we illustrate our approach and demonstrate the associated computational savings.

Optimal Adaptation in Web Processes with Coordination Constraints

We present methods for optimally adapting Web processes to exogenous events while preserving inter-service constraints that necessitate coordination. For example, in a supply chain process, orders placed by a manufacturer may get delayed in arriving. In response to this event, the manufacturer has the choice of either waiting out the delay or changing the supplier. Additionally, there may be compatibility constraints between the different orders, thereby introducing the problem of coordination between them if the manufacturer chooses to change the suppliers. We focus on formulating the decision making models of the managers, who must adapt to external events while satisfying the coordination constraints, using Markov decision processes. Our methods range from being centralized and globally optimal in their adaptation but not scalable, to decentralized that is suboptimal but scalable to multiple managers. We also develop a hybrid approach that improves on the performance of the decentralized approach with a minimal loss of optimality

Dynamic workflow composition using Markov decision processes

The advent of Web services has made automated workflow composition relevant to Web based applications. One technique, that has received some attention, for automatically composing workflows is AI-based classical planning. However, classical planning suffers from the paradox of first assuming deterministic behavior of Web services, then requiring the additional overhead of execution monitoring to recover from unexpected behavior of services. To address these concerns, we propose using Markov decision processes (MDPs), to model workflow composition. Our method models both, the inherent stochastic nature of Web services, and the dynamic nature of the environment. The resulting workflows are robust to nondeterministic behaviors of Web services and adaptive to a changing environment. Using an example scenario, we demonstrate our method and provide empirical results in its support.

Inexact matching of ontology graphs using expectation-maximization

We present a new method for mapping ontology schemas that address similar domains. The problem of ontology matching is crucial since we are witnessing a decentralized development and publication of ontological data. We formulate the problem of inferring a match between two ontologies as a maximum likelihood problem, and solve it using the technique of expectation-maximization (EM). Specifically, we adopt directed graphs as our model for ontology schemas and use a generalized version of EM to arrive at a map between the nodes of the graphs. We exploit the structural, lexical and instance similarity between the graphs, and differ from the previous approaches in the way we utilize them to arrive at, a possibly inexact, match. Inexact matching is the process of finding a best possible match between the two graphs when exact matching is not possible or is computationally difficult. In order to scale the method to large ontologies, we identify the computational bottlenecks and adapt the generalized EM by using a memory bounded partitioning scheme. We provide comparative experimental results in support of our method on two well-known ontology alignment benchmarks and discuss their implications.

Selective Querying for Adapting Web Service Compositions Using the Value of Changed Information

Web service composition (WSC) techniques assume that the parameters used to model the environment remain static and accurate throughout the compositions execution. However, WSCs often operate in environments where the parameters of its component services are volatile. To remain optimal, WSCs must adapt to these changes. Adaptation requires up-to-date knowledge about the revised parameters of each of the services. One way of obtaining this knowledge is to query services for their revised parameters. Querying services for their parameters is time-consuming and expensive. We must therefore carefully manage how queries are conducted. Specifically, an adaptive WSC must know when to query for revised information, and from which service(s) to obtain information. We present a method to selectively query services using the value of changed information (VOC). VOC measures the value of the change that revised information may potentially introduce to the composition. We reduce the complexity of computing the VOC, first by anticipating values of the service parameters that do not change the WSC, and second by using parameter expiration times obtained from predefined service-level agreements. Using two scenarios, we illustrate our approach and demonstrate the computational savings theoretically and experimentally.

Interactive POMDPs: Properties and Preliminary Results

This paper presents properties and results of a new framework for sequential decision-making in multiagent settings called interactive partially observable Markov decision processes (I-POMDPs). I-POMDPs are generalizations of POMDPs, a well-known framework for decision-theoretic planning in uncertain domains, to cases when an agent needs to plan a course of action in an environment populated by other agents.

A hierarchical framework for logical composition of web services

Automatically composing Web services to form processes in the context of service-oriented architectures has attracted significant research. Prevalent approaches for automatically composing Web services predominantly utilize planning techniques to achieve the composition. However, classical planning based approaches face the following challenges: (i) difficulty in modeling the uncertainty of Web service invocations, (ii) inability to optimize the composition using non-functional parameters, and (iii) difficulty in scaling efficiently to large compositions. In order to address these issues, we present a hierarchical framework for logically composing Web services, which we call Haley. In comparison to classical planners, Haley utilizes decision-theoretic planning that is able to model and reason with the uncertainty inherent in Web service invocations and provides an expected cost-based optimization. Haley uses symbolic planning techniques that operate directly on first-order logic based representations of the state space to obtain the compositions. Consequently, it supports automated elicitation of the corresponding planning problem from Web service descriptions and produces a domain representation that is more compact than that of classical planners. Furthermore, it promotes scalability by exploiting the natural hierarchy found in real-world processes. Due to the limitations of the existing approaches and the complexity of the Web service composition problem, few implemented tools exist, although many approaches have been proposed in the literature. We have implemented Haley and provided a comprehensive tool suite for composing Web services. The suite operates on Web services described using well-known languages such as SAWSDL. It provides process designers with an intuitive interface to specify composition requirements, goals and a hierarchical decomposition if available, and automatically generates BPEL compositions while hiding the complexity of the planning and of BPEL from users.