Catálogo de publicaciones - libros

Service Level Agreements (SLA) are needed to allow business interactions to rely on Internet services. Service Level Objectives (SLO) specify the committed performance level of a service. Thus, SLA compliance auditing aims at verifying these commitments. Since SLOs for various application services and end-to-end performance definitions vary largely, auditing of SLA compliances poses the challenge to an auditing framework. Moreover, end-to-end performance data are potentially large for a provider with many customers. Therefore, this paper presents a and auditing framework and a prototype, termed (diting Famework for nternet Servies), whose components can be distributed across different domains.

Experiences with the management of Grid specific monitoring and accounting data have shown that current approaches do not sufficiently support a distinction between providers, users and customers of a Grid. This gap can be filled by the use of Customer Service Management techniques which enable customers to individually monitor and control their subscribed services. We adapt a Customer Service Management scenario to Grid environments and outline an architecture dedicated to the management and visualization of monitoring and accounting data. To proof the concept, a prototype based on standard Grid components which manages user’s needs and interactions with the resource provider is presented.

Obtaining information about the usage of network ressources by individual users forms the basis for establishing network billing systems or network management operations. While there are already widely used accounting techniques available for measuring IP network traffic on a per-host basis, there is no adequate solution for accounting per-user network activities on a multiuser operating system. This work provides a survey on existing approaches to this problem and identifies requirements for a user-based IP accounting module. It develops a suitable software architecture LINUBIA and proposes a prototypical implementation for the Linux 2.6 operating system, which is capable of providing per-user accounting for both the IPv4 and the IPv6 protocol.

Web Services are an XML technology recently viewed as capable of being used for network management. A key aspect of WS in this domain is event reporting. WS-based research in this area has produced a collection of notification specifications, which consider even aspects such as filtering to reduce machine and network resource consumption. Still though, additional aspects need to be addressed if WS event reporting is to be used efficiently for network management. This paper borrows an idea in network management that of policy based task delegation and applies it in the context of WS-based management by using the WS-Notification standard messages, to increase event reporting efficiency. More specifically, we are adding functionality to the entity that produces events making it capable of performing a set of tasks apart from simple ones such as collecting and reporting notification data. This functionality allows an entity, such as a manager, capable of delegating tasks of various complexities to an event reporting entity where they can be performed dynamically. As a proof of concept that the approach is feasible and increases efficiency we analyze a complex event reporting scenario where task delegation is used. We compare this approach for performance to a plain WS-based event system and also to simple SNMP traps.

We present a transaction processing scheme for structured overlay networks and use it to develop a distributed Wiki application based on a relational data model. The Wiki supports rich metadata and additional indexes for navigation purposes.

Ensuring consistency and durability requires handling of node failures. We mask such failures by providing high availability of nodes by constructing the overlay from replicated state machines . Atomicity is realized using two phase commit with additional support for failure detection and restoration of the transaction manager. The developed transaction processing scheme provides the application with a mixture of pessimistic, hybrid optimistic and multiversioning concurrency control techniques to minimize the impact of replication on latency and optimize for read operations. We present pseudocode of the relevant Wiki functions and evaluate the different concurrency control techniques in terms of message complexity.