Hector A. Duran-Limon

A context-aware middleware for applications in mobile Ad Hoc environments

Novel ubiquitous computing applications such as intelligent vehicles, smart buildings, and traffic management require special properties that traditional computing applications do not support, such as context-awareness, massive decentralisation, autonomous behaviour, adaptivity, proactivity, and innate collaboration. This paper presents a new computational model and middleware that reflect support for the required the properties. The sentient object model is proposed by the CORTEX¹ project to support the construction of ubiquitous applications. A flexible, run-time reconfigurable component-based middleware has been built to provide run-time support to engineer the sentient object programming paradigm. An application infrastructure using sentient objects to enable cooperation between autonomous and proactive vehicles has been implemented to demonstrate the appropriateness of the computational model and the validity of the middleware for pervasive mobile ad hoc computing.

Reflection, self-awareness and self-healing in OpenORB

There is a growing interest in the area of self-healing systems. Self-healing does however impose considerable demands on system infrastructures---especially in terms of openness and support for reconfigurability. This paper proposes that the self-awareness inherent in reflective technologies lends itself well to the construction of self-healing systems. In particular, the paper examines the support provided by the Open ORB reflective middleware technology for the construction of this increasingly important class of system.

Adaptive resource management in middleware: a survey

Current middleware technologies cannot meet the demands of new application areas, such as embedded and mobile systems, that require mechanisms for dealing with a changing environment. This article reviews several approaches for providing adaptive resource management for middleware. Current middleware technologies, such as the Common Object Request Broker Architecture (CORBA) and .NET (http://msdn.microsoft.com/net), mask system and network heterogeneity problems and alleviate the inherent complexity of distributed systems in many application areas. However, the recent emergence of new application areas for middleware, such as embedded systems, real-time systems, and multimedia, imposes challenges that few existing middleware platforms can meet. In particular, because they impose greater resource-sharing and dynamism demands, these application areas require more complex and sophisticated middleware. Resource sharing must be controlled and predictable to ensure that activities running on the same middleware instance have adequate resources.

Reconfiguration of resources in middleware

The monolithic and inflexible nature of current middleware, has made it difficult to deal with emerging technologies such as multimedia. We believe that reflection provides a principled means to achieve the flexibility and adaptation required The main focus of this paper regards the reconfiguration of resources in middleware within the context of OpenORB, a reflective middleware architecture. A resource model is presented which provides a representation of the physical resources whereby various levels of abstraction are offered The approach is validated by a series of experimental results.

The Importance of Resource Management in Engineering Distributed Objects

Middleware technologies such as CORBA and DCOM have been developed as a means of tackling heterogeneity and complexity problems inherent in distributed systems. However, more work still need to be done to develop methodologies for the construction of distributed objects. In addition, little attention has been paid to the development of methodologies for the configuration of computational resources among distributed objects. This paper introduces a resource configuration description language (RCDL) for the specification of the resource management of distributed systems. This language is based on both a resource model and a task model. The former offers various levels of abstraction for resources, resource factories and resource mangers. The latter then provides a fine- and a coarse-grained approach to allocate resources to both application services and middleware services by breaking such services into task hierarchies. Finally, we use reflection as a principled means to obtain a clear separation of concerns between the functional and nonfunctional behaviour (e.g. resource management) of distributed systems.

Platform-independent modeling and prediction of application resource usage characteristics

Application resource usage models can be used in the decision making process for ensuring quality-of-service as well as for capacity planning, apart from their general use in performance modeling, optimization, and systems management. Current solutions for modeling application resource usage tend to address parts of the problem by either focusing on a specific application, or a specific platform, or on a small subset of system resources. We propose a simple and flexible approach for modeling application resource usage in a platform-independent manner that enables the prediction of application resource usage on unseen platforms. The technique proposed is application agnostic, requiring no modification to the application (binary or source) and no knowledge of application-semantics. We implement a Linux-based prototype and evaluate it using four different workloads including real-world applications and benchmarks. Our experiments reveal prediction errors that are bound within 6-24% of the observed for these workloads when using the proposed approach.

Using lightweight virtual machines to achieve resource adaptation in middleware

Current middleware does not offer enough support to cover the demands of emerging application domains, such as embedded systems or those featuring distributed multimedia services. These kinds of applications often have timeliness constraints and yet are highly susceptible to dynamic and unexpected changes in their environment. There is then a clear need to introduce adaptation in order for these applications to deal with such unpredictable changes. Resource adaptation can be achieved by using scheduling or allocation algorithms, for large-scale applications, but such a task can be complex and error-prone. Virtual machines (VMs) represent a higher-level approach, whereby resources can be managed without dealing with lower-level details, such as scheduling algorithms, scheduling parameters and so on. However, the overhead penalty imposed by traditional VMs is unsuitable for real-time applications. On the other hand, virtualisation has not been previously exploited as a means to achieve resource adaptation. This study presents a lightweight VM framework that exploits application-level virtualisation to achieve resource adaptation in middleware for soft real-time applications. Experimental results are presented to validate the approach.

QoS Management specification support for multimedia middleware

Middleware technologies are now widely used in order to provide support for the interaction of systems relying on different hardware and operating systems. At present middleware platforms, however, do not provide enough support for both the configuration and reconfiguration of quality of service (QoS) management aspects of real-time applications such as distributed multimedia systems. That is, current middleware only provides support for the low-level specification of QoS properties. This paper presents an architecture description language (ADL) called Xelha For the high-level specification of QoS management in multimedia middleware whereas lower-level aspects can be tuned by using an aspect-oriented suite of languages referred to as resource configuration description language (RCDL). Tool support is also provided for the interpretation of the Xelha and RCDL languages.

Using Lightweight Virtual Machines to Run High Performance Computing Applications: The Case of the Weather Research and Forecasting Model

There are many scientific applications that have high performance computing demands. Such demands are traditionally supported by cluster-or Grid-based systems. Cloud computing, which has experienced a tremendous growth, emerged as an approach to provide on-demand access to computing resources. The cloud computing paradigm offers a number of advantages over other distributed platforms. For example, the access to resources is flexible and cost-effective since it is not necessary to invest a large amount of money on a computing infrastructure nor pay salaries for maintenance functions. Therefore, the possibility of using cloud computing for running high performance computing applications is attractive. However, it has been shown elsewhere that current cloud computing platforms are not suitable for running this kind of applications since the performance offered is very poor. The reason is mainly the overhead from virtualisation which is extensively used by most cloud computing platforms as a means to optimise resource usage. In this paper, we present a lightweight virtualisation approach applied to WRF, a challenging communication-intensive, high performance computing application. Our experimental results show that lightweight virtualisation imposes about 5% overhead and it substantially outperforms traditional heavy-weight virtualisation such as VMware.

Performance prediction of weather forecasting software on multicore systems

Performance prediction is valuable in different areas of computing. The popularity of lease-based access to high performance computing resources particularly benefits from accurate performance prediction. Most contemporary processors are employing multiple computing cores, which complicates the task of performance prediction. In this paper, we describe the methodology employed for predicting the performance of a popular weather forecasting application on systems with between 4 and 256 processors. An average prediction error of less than 10% was achieved after testing on three different multi-node, multicore systems.