Tesis:
MIDDLEWARE FOR HIGH-AVAILABLE AND SCALABLE MULTI-TIER AND SERVICE-ORIENTED ARCHITECTURES.
- Autor: PEREZ SORROSAL, Francisco
- Título: MIDDLEWARE FOR HIGH-AVAILABLE AND SCALABLE MULTI-TIER AND SERVICE-ORIENTED ARCHITECTURES.
- Fecha: 2009
- Materia: Sin materia definida
- Escuela: FACULTAD DE INFORMATICA
- Departamentos: LENGUAJES Y SISTEMAS INFORMATICOS E INGENIERIA DE SOFTWARE
- Acceso electrónico:
- Director/a 1º: PATIÑO MARTINEZ, Marta
- Resumen: Information systems constitute the backbone of modern companies. They support ebusiness applications and contain critical data on which depends the work of many people in all the levels of organizations. More precisely, the middleware that supports multi-tier and service-oriented applications is nowadays an important cornerstone for many businesses. There are many applications and services that require transactions running on top of this middleware.
Transactions are used to ensure consistency of critical data in the presence of concurrent accesses and failures. High availability and scalability are also two desirable properties for the middleware platforms where the applications are deployed. When a failure occurs in a system, clients may not access the applications deployed on it, compromising their availability. Thus, high availability is a key factor for applications because they enable e-businesses to run in a 24/7 fashion. On the other hand, a system is said to be scalable if its performance increases after adding more resources to it. Scalability indicates the capacity of a system of handling growing amounts of work. Thus, the inclusion of these properties in current information systems can make the difference between a competitive and a non-competitive organization. Providing high availability and scalability to current middleware platforms is not trivial. Fist of all, the inclusion should be transparent for the applications. This requires that the underlying middleware infrastructure provides and combines the adequate techniques, such as replication, failover and recovery. Moreover, data consistency of stateful applications must be guaranteed. Finally, when trying to get one of these two properties (e.g. high vailability),
the other (scalability) may be compromised and vice-versa. Most of the iddleware platforms that claim to provide these properties for stateful applications do not provide both features at the same time, presenting trade-offs. To the best of our knowledge, when a middleware platform claims to support both features, it only supports to scale out stateless applications.
Thus, new techniques and technology to overcome these trade-offs must be investigated. This Ph.D. Thesis aims to improve current middleware platforms. Our contributions allow to provide transparent high availability and scalability to stateful transactional applications and services. We have developed a set of replication protocols and services at middleware level,
which provide these two properties to multi-tier and service-oriented architectures (SOAs). Our protocols also guarantee the consistency of stateful applications and services running in the replicated infrastructure. Moreover, node failures are transparent to the clients. The replication protocols for multi-tier architectures are suitable for stateful applications running in clusters of application servers connected through LANs. In order to guarantee high availability, we have also developed recovery protocols that allow to recover failed nodes or add fresh ones to the cluster without stopping the system (online recovery). With regard to SOAs, we have developed a replication framework that provides high availability for critical web services. The web service replicas can be located in different nodes of LANs or WANs. The framework respects the web services autonomy, so the replicated web service is accessed in the same manner as the non-replicated one. Moreover, the framework
allows to easily deploy the web service replicas from the original non-replicated web service.