Orchestration of resources in distributed, heterogeneous grid environments using dynamic service level agreements
Loading...
Date
2011-12-19
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Die Akzeptanz des Internets und der zunehmende Ausbau von Netzwerkkapazitäten ermöglichen bereits heute einen effizienten und zuverlässigen Austausch riesiger Datenmengen zwischen verschiedenen Rechensystemen weltweit. Hieraus resultieren neue Paradigmen bei der Bereit-stellung und Nutzung verteilter IT-Ressourcen wie zum Beispiel das Grid-Computing. Im Grid-Computing werden Rechenressourcen verschiedener Institutionen bzw. Organisationen koordiniert zur Lösung wissenschaftlicher und wirtschaftlicher Problemstellungen genutzt. Neben Rechenressourcen werden dabei auch Daten, Datenspeicher oder Software bereitgestellt. Die Qualität mit der diese Ressourcen bereitgestellt werden gewinnt dabei zunehmend an Bedeutung. Qualitätseigenschaften sind zum Beispiel die minimale Verfügbarkeit von Rechenressourcen, die maximale Zugriffszeit eines Datenspeichers oder die maximale Antwortzeit einer web-basierten Anwendung. Für Ressourcenanbieter bedeutet dies dass spezifische Prozesse implementiert werden müssen um qualitativ hochwertige IT-Dienste bereitzustellen. Zudem können Dienste mit unterschiedlichen Dienstqualitäten bereitgestellt werden, wobei Dienste mit geringerer Qualität preiswerter angeboten werden als solche mit hoher Qualität. Anwender hingegen können den für sie passenden Dienst hinsichtlich ihrer Anforderungen und ihres Budgets auswählen.
Service Level Agreements (SLAs) sind ein akzeptierter Ansatz um Verträge über IT-Dienste und Dienstqualitäten zu realisieren. SLAs beschreiben sowohl die funktionalen als auch die nicht-funktionalen Anforderungen von IT-Diensten als auch Vergütung und Strafen für Erfüllung bzw. Nichterfüllung der definierten Anforderungen. Diese Arbeit behandelt Methoden zur Verhandlung und Verwaltung von dynamischen SLAs in verteilten Systemen auf Basis des WS-Agreement Standards. Im Fokus steht hierbei die Deklaration von SLAs, deren automatisierte Verhandlung und Erstellung, das Monitoring von SLA Garantien, sowie die Verwendung von SLAs zur koordinierten Nutzung von IT-Ressourcen. Zu diesem Zweck wurde aufbauend auf die WS-Agreement Spezifikation ein Protokoll zur dynamischen Verhandlung bzw. Neuverhandlung von SLAs entwickelt. Dies beinhaltet die Definition eines Verhandlungsmodells zum Austausch von Angeboten zwischen den Verhandlungspartnern. Die anschließende Erstellung der SLAs basiert auf dem WS-Agreement Standard stellt einen automatisierter Prozess dar. Da es sich bei SLAs um elektronische Verträge handelt wurden Mechanismen zur Validierung von SLA Angeboten entwickelt und im Detail vorgestellt. Darüber hinaus werden Methoden zur automatisierten Evaluation von SLA Garantien beschrieben. Abschließend wird die Architektur und Implementierung eines Orchestrierungsdienstes zur Co-Allokation beliebiger Ressource wie z.B. Rechen- und Netzwerkressourcen vorgestellt. Die Ressourcenorchestrierung wurde hierbei mittels SLAs realisiert.
In recent decades the acceptance of the internet and the increase of network capacity have resulted in a situation in which it is now possible to transfer huge amounts of data efficiently and reliably between different computing systems worldwide. This enables new paradigms in provision and use of distributed IT resources. Grid computing is such a well-known paradigm where computing resources owned by various institutions and organizations are used in a coordinated way in order to solve scientific and economic problems. Besides computing resources also data, storage or software resources are provided. Today it becomes more and more important with which quality the different resources are provided. This may be, for example, the minimal availability of computing resources, the maximum access time of a data storage or the maximum response time of a web-based application. Offering resources with a defined quality means for resource providers that they need to implement specific processes to assert the quality of the provisioning process. On the other hand, resource providers can offer their services at different quality levels. Services with a lower quality can be offered cheaper than those with a higher quality. Service consumers can therefore select the service with the appropriate service level in terms of their requirements and budget. This provides both parties, service provider and consumer, with more flexibility during the service provisioning process. Service level agreements (SLAs) are an accepted approach to realize contracts for IT services and service qualities. They describe the functional and the non-functional requirements of IT services. Additionally, they define compensation and penalties for delivering services with the defined requirements respectively for failing to meet these quality criteria. This thesis examines methods for negotiation and management of SLAs in distributed systems based on the WS-Agreement standard. The focus is on methods for SLA declaration, automated SLA negotiation and creation processes, monitoring of SLA guarantees, and the application of SLAs for coordinated IT resource provisioning. Therefore, a protocol for dynamic negotiation or renegotiation of SLAs is developed as an extension to the WS-Agreement specification. This includes the definition of a negotiation model for the exchange of offers between the negotiating partners. The subsequent SLA creation process is an automated process in distributed systems. Since SLAs are a kind of electronic contracts a mechanism for validating the integrity of SLA offers was developed and is presented in detail. In addition, automatic methods for SLA guarantee evaluation are described. Finally, an orchestration service for co-allocating arbitrary resources such as computing and network resources is presented. The resource orchestration process has been realized using SLAs. The architecture of this service is evaluated and based on the evaluation result an advanced orchestration service architecture is conceived.
In recent decades the acceptance of the internet and the increase of network capacity have resulted in a situation in which it is now possible to transfer huge amounts of data efficiently and reliably between different computing systems worldwide. This enables new paradigms in provision and use of distributed IT resources. Grid computing is such a well-known paradigm where computing resources owned by various institutions and organizations are used in a coordinated way in order to solve scientific and economic problems. Besides computing resources also data, storage or software resources are provided. Today it becomes more and more important with which quality the different resources are provided. This may be, for example, the minimal availability of computing resources, the maximum access time of a data storage or the maximum response time of a web-based application. Offering resources with a defined quality means for resource providers that they need to implement specific processes to assert the quality of the provisioning process. On the other hand, resource providers can offer their services at different quality levels. Services with a lower quality can be offered cheaper than those with a higher quality. Service consumers can therefore select the service with the appropriate service level in terms of their requirements and budget. This provides both parties, service provider and consumer, with more flexibility during the service provisioning process. Service level agreements (SLAs) are an accepted approach to realize contracts for IT services and service qualities. They describe the functional and the non-functional requirements of IT services. Additionally, they define compensation and penalties for delivering services with the defined requirements respectively for failing to meet these quality criteria. This thesis examines methods for negotiation and management of SLAs in distributed systems based on the WS-Agreement standard. The focus is on methods for SLA declaration, automated SLA negotiation and creation processes, monitoring of SLA guarantees, and the application of SLAs for coordinated IT resource provisioning. Therefore, a protocol for dynamic negotiation or renegotiation of SLAs is developed as an extension to the WS-Agreement specification. This includes the definition of a negotiation model for the exchange of offers between the negotiating partners. The subsequent SLA creation process is an automated process in distributed systems. Since SLAs are a kind of electronic contracts a mechanism for validating the integrity of SLA offers was developed and is presented in detail. In addition, automatic methods for SLA guarantee evaluation are described. Finally, an orchestration service for co-allocating arbitrary resources such as computing and network resources is presented. The resource orchestration process has been realized using SLAs. The architecture of this service is evaluated and based on the evaluation result an advanced orchestration service architecture is conceived.
Description
Table of contents
Keywords
Cloud, Distributed systems, Grid, Monitoring, Service level agreement, SLA, SOA, WS-agreement, WS-agreement negotiation