Summary form only given, as follows. Today's organizations consume an increasing share of their computing resources as computing services, using an ondemand model and paying only for what they use. They consume cloud business applications and processes, and store and analyze vast amounts of information on cloud provisioned resources. Cloud computing is dramatically reducing the capital required to support IT operations, while bringing unprecedented openness, flexibility and access to enterprise computing. One of the paradoxes of this new environment is that openness and easy access to services and resources results in the creation of new barriers and computing silos. New 'born in the cloud' enterprises are quickly discovering the difficulty of integrating SaaS services into their daily operations in efficient, secure and reliable ways. SaaS APIs are as siloed as traditional enterprise applications, if not more. Data is still represented in heterogeneous formats and kept in multiple application stores. Thus, while access and delivery have fundamentally improved, the core problems of process and data integration remain. If anything, they are getting harder. Successful technologies for process and data integration in the cloud will differ fundamentally from existing approaches for enterprise integration. While existing approaches focus on eliminating silos and leveling access to process and information, cloud integration will acknowledge barriers as the price of extending our reach beyond the limits a single organization and focus on linking capabilities and information. Instead of reengineering processes, new process integration technology will link processes and application services offered by specialized providers, leveraging instead of replacing available processes. Instead of coercing heterogeneous information into unified schemas, the new approach to information integration will link enterprise data and open data to offer virtual access to much richer information set. New cloud integration services are already being built leveraging a collection of technologies born on the Web and also within the enterprise. They include "low touch" technologies such as process correlation and mining, process analytics, linked and open data, and provenance among others. This keynote will discuss trends in SaaS and cloud-based integration, and review some of the foundational technologies that can support a new cloud centric data and application integration paradigm.

In today's Web, many functionality-wise similar Web services are offered through heterogeneous interfaces (operation definitions) and business protocols (ordering constraints defined on legal operation invocation sequences). The typical approach to enable interoperation in such a heterogeneous setting is through developing adapters. There have been approaches for classifying possible mismatches between service interfaces and business protocols to facilitate adapter development. However, the hard job is that of identifying, given two service specifications, the actual mismatches between their interfaces and business protocols. In this paper we present novel techniques and a tool that provides semi-automated support for identifying and resolution of mismatches between service interfaces and protocols, and for generating adapter specification. We make the following main contributions: (i) we identify mismatches between service interfaces, which leads to finding mismatches of type of signature, merge/split, and extra/missing messages; (ii) we identify all ordering mismatches between service protocols and generate a tree, called mismatch tree, for mismatches that require developers' input for their resolution. In addition, we provide semi-automated support in analyzing the mismatch tree to help in resolving such mismatches. We have implemented the approach in a tool inside IBM WID (WebSphere Integration Developer). Our experiments with some real-world case studies show the viability of the proposed approach. The methods and tool are significant in that they considerably simplify the problem of adapting services so that interoperation is possible.

Web services and messaging, as application-to-application communication paradigms, have so far been considered separately, with independent programming models and supporting middleware. Different efforts are now introducing messaging notions such as asynchrony, greater consumer cardinality, and looser coupling between Web services. This trend will likely result in an extension of the Web services programming model. It is not clear, however, that this extension will adhere to a pre-planned approach. A coherent approach requires a thorough integration of the Web services and messaging paradigms. This paper proposes one such approach which, in addition to supporting the current style of Web services interactions, allows the incorporation of messaging-style interactions under a common programming model. These messaging-style interactions include asynchronous request-response, oneway multi-consumer interactions, and even multiple-choice point-to-point interactions, common in message queuing systems. This paper also elaborates on a model for oneway multi-consumer interactions that integrates the publish/subscribe mode of messaging into the Web services programming model. A primary motivation for our approach is to take advantage of key messaging features, while exerting as small an impact as possible on the Web services programming model

Abstract BPEL4WS (BPEL in short) is a business process definition language built natively on top of the Web services application model. BPEL provides a workflow‐oriented composition model for Web services applications, and is thus a central piece in the heavily componentized service‐oriented computing model. BPEL results from the merger of two distinct process metamodels (the process algebra model of XLANG and the graph‐oriented model of WSFL) into a coherent and powerful framework. Implementing BPEL thus presents significant challenges to middleware developers. This paper discusses those challenges and describes the design and architecture of the BPWS4J runtime, and a full implementation of the BPELWS 1.1 specification. Copyright © 2005 John Wiley & Sons, Ltd.

Architectural mismatch; the semantic and syntactic gap preventing component composition, can be partially alleviated through the ubiquity of the protocols for service composition (syntax). And yet, the semantics of service or component composition have remained elusive. The contribution of this paper builds on the notions introduced by grammar-oriented object design (GOOD) using the notion of manners to specify the semantics of services and drive their automated composition through dynamically reconfigure architectures enabled through GOOD. This leads to the use of manners for explicit service semantics representation in a re-configurable architectural style. We describe the base requirements for automated component and services composition and show how automated assembly of components and services can be accomplished through a dynamically reconfigurable (DyRec) architectural style. In addition, we show how this vision of dynamic reconfiguration is naturally supported by the Web services framework, because it derives from some of the core requirements of the service oriented architectures (SOA) style embodied by Web services. Further, we discuss how such a DyRec architectural style can be implemented using grammar-oriented object design (GOOD) and describe a project that led to a tool called the Business Compiler, to accomplish dynamic composition and collaboration.

Service metadata plays a central role in service oriented architecture (SOA)s. Explicit service metadata makes possible the decoupling of applications and supports dynamic selection and binding. Moreover it extends the basic interoperability model beyond the implicit agreement to use a set of common protocols, making possible advanced collaboration models based on the dynamic discovery of partner capabilities. Web services provide the beginning of an XML based instantiation of the SOA model, with major middleware vendors providing implementations on top of their current platforms. In this paper, we discuss how the use of middleware metadata in SOA environments leads to optimized interaction and increase application decoupling.

We study Gaussian integration of Lipschitz functions in the ℓp-norm, p∈[1, ∞], defined over the d dimensional space Rd. It is assumed that the Gaussian weight has a diagonal covariance matrix with variances σi. We consider the problem of how the worst case complexity of computing an ε-approximation, comp(ε), depends on the variances σi. If ε goes to zero then comp(ε) is of order ε−d and we have the curse of dimension. For a fixed ε we provide lower and upper bounds on comp(ε) of the form ε−t where the exponent t is small for small variances σi. We call this moderation of the exponential growth of the complexity, the delayed curse of dimension. We prove similar upper and lower bounds for the integration of Lipschitz functions on sequence spaces ℓp, p∈[1, ∞), which is essentially the problem of path integration with a Gaussian measure for which ∑∞i=1 σp/2i<∞. Finally, we show how the same proof technique yields an upper bound when the integrand belongs to the Sobolev class Wr, ∞(Rd), d<∞, of functions with bounded (generalized) derivatives of order r, and when Hermite information is used. This upper bound shows that the moderating effect of the variances becomes much more essential for large r.

We study optimal integration over the infinite interval (−∞, +∞) for a Gaussian weight with variance σ. We consider functions satisfying a Lipschitz condition with constantL. We characterize the optimal information by a system of nonlinear equations, and show that (asymptotically) the solution is defined by the quantiles of the Gaussian weight with double variance 2σ. We provide an optimal algorithm and prove that fornoptimal sample points the worst case error of the optimal algorithm is asymptotically equal toπσ/2L/n. Finally, we show that the worst case error for the set of sampling points defined by the zeros of Hermite polynomials is quadratically worse than for optimal sample points.