Abstract

Evaluation of non-functional properties of a design (such as performance, dependability, security, etc.) can be enabled by design annotations specific to the property to be evaluated. Performance properties, for instance, can be annotated on UML designs by using the "UML Profile for Schedulability, Performance and Time (SPT)". However the communication between the design description in UML and the tools used for non-functional properties evaluation requires support, particularly for performance where there are many alternative performance analysis tools that might be applied. This paper describes a tool architecture called PUMA, which provides a unified interface between different kinds of design information and different kinds of performance models, for example Markov models, stochastic Petri nets and process algebras, queues and layered queues.The paper concentrates on the creation of performance models. The unified interface of PUMA is centered on an intermediate model called Core Scenario Model (CSM), which is extracted from the annotated design model. Experience shows that CSM is also necessary for cleaning and auditing the design information, and providing default interpretations in case it is incomplete, before creating a performance model.

Distributed or parallel software with synchronous communication via rendezvous is found in client-server systems and in proposed Open Distributed Systems, in implementation environments such as Ada, V, Remote Procedure Call systems, in Transputer systems, and in speci cation techniques such as CSP, CCS and LOTOS. The delays induced by rendezvous can cause serious performance problems, which are not easy to estimate using conventional models which focus on hardware contention, or on a restricted view of the parallelism which ignores implementation constraints. Stochastic Rendezvous Networks are queueing networks of a new type which have been proposed as a modelling framework for these systems. They incorporate the two key phenomena of included service and a second phase of service. This paper extends the model to also incorporate di erent services or entries associated with each task. Approximations to arrival-instant probabilities are employed with a Mean-Value Analysis framework, to give approximate performance estimates. The method has been applied to moderately large industrial software systems.

The importance of assessing software non-functional properties (NFP) beside the functional ones is well accepted in the software engineering community. In particular, dependability is a NFP that should be assessed early in the software life-cycle by evaluating the system behaviour under different fault assumptions. Dependability-specific modeling and analysis techniques include for example Failure Mode and Effect Analysis for qualitative evaluation, stochastic Petri nets for quantitative evaluation, and fault trees for both forms of evaluation. Unified Modeling Language (UML) may be specialized for different domains by using the profile mechanism. For example, the MARTE profile extends UML with concepts for modeling and quantitative analysis of real-time and embedded systems (more specifically, for schedulability and performance analysis). This paper proposes to add to MARTE a profile for dependability analysis and modeling (DAM). A case study of an intrusion-tolerant message service will offer insight on how the MARTE-DAM profile can be used to derive a stochastic Petri net model for performance and dependability assessment.

The Object Management Group (OMG) is in the process of defining a UML Profile for Schedulability, Performance and Time that will enable the construction of models for making quantitative predictions regarding these characteristics. The paper proposes a graph-grammar based method for transforming automatically a UML model annotated with performance information into a Layered Queueing Network (LQN) performance model. The input to our transformation algorithm is an XML file that contains the UML model in XML format according to the standard XMI interface. The output is the corresponding LQN model description file, which can be read directly by existing LQN solvers. The LQN model structure is generated from the high-level software architecture and from deployment diagrams indicating the allocation of software components to hardware devices. The LQN model parameters are obtained from detailed models of key performance scenarios, represented as UML interaction or activity diagrams.

Performance analysis of a software specification in a language such as UML can assist a design team in evaluating performance-sensitive design decisions and in making design trade-offs that involve performance. Annotations to the design based on the UML Profile for Schedulability, Performance and Time provide necessary information such as workload parameters for a performance model, and many different kinds of performance techniques can be applied. The Core Scenario Model (CSM) described here provides a metamodel for an intermediate form which correlates multiple UML diagrams, extracts the behaviour elements with the performance annotations, attaches important resource information that is obtained from the UML, and supports the creation of many different kinds of performance models. Models can be made using queueing networks, layered queues, timed Petri nets, and it is proposed to develop the CSM as an intermediate language for all performance formalisms. This paper defines the CSM and describes how it resolves questions that arise in performance model-building. Performance analysis of software specificationsPreliminary performance analysis can be effective not only in avoiding performance disasters in software projects [24], but in understanding the many design tradeoffs that involve performance. There are many suitable modeling techniques, which are surveyed in [2], however, the time and effort necessary to create the performance models may be prohibitive. The UML Profile for schedulability, performance and time (SPT) [14] was developed to assist in the capture of performance data, and in the automation of the model-building step. This should make the analysis more accessible to developers who are concerned about performance issues in their designs. Fig. 1 illustrates the type of processing that is envisaged by the SPT Profile.The SPT Profile addresses a broad range of applications, from embedded systems with schedulability concerns (as described by Liu [11]), to applications with statistical performance requirements, such as telecommunications, business systems and web services. The Core Scenario Model (CSM) defined here supports building performance models from the specifications of such systems.The change of perspective from a functional specification model in UML to a performance model is profound and requires a re-orientation of the model information. Resources, which are peripheral in functional specification, are central for performance. Performance is determined by the way operations use resources (which resources, for how long, and in what order). The CSM metamodel expresses resource-centric models, which may be derived from UML via the SPT Profile, in a precise way which supports the generation of different kinds of performance models.

No abstract

A reengineered approach to the early prediction of preterm birth is presented as a complimentary technique to the current procedure of using costly and invasive clinical testing on high-risk maternal populations. Artificial neural networks (ANNs) are employed as a screening tool for preterm birth on a heterogeneous maternal population; risk estimations use obstetrical variables available to physicians before 23 weeks gestation. The objective was to assess if ANNs have a potential use in obstetrical outcome estimations in low-risk maternal populations. The back-propagation feedforward ANN was trained and tested on cases with eight input variables describing the patient's obstetrical history; the output variables were: 1) preterm birth; 2) high-risk preterm birth; and 3) a refined high-risk preterm birth outcome excluding all cases where resuscitation was delivered in the form of free flow oxygen. Artificial training sets were created to increase the distribution of the underrepresented class to 20%. Training on the refined high-risk preterm birth model increased the network's sensitivity to 54.8%, compared to just over 20% for the nonartificially distributed preterm birth model.