Performability and Dependability Evaluation
To stay in business, organizations must develop and maintain a value proposition that is compelling, whether you are a high-volume or a low-volume provider of products or services. Customers expect both receiving products that work properly for long and expect continuous delivering of contracted services. Over the last decade, performability (performance and reliability) issues have become a critical aspect of that proposition. In particular, such an evaluation may be carried out through either measuring or considering system’s models. In many situations modeling is the method of choice either because the system might not yet exist or due to the inherent complexity of creating specific scenarios under which the system should be evaluate. However, one of the main modelers concerns whenever evaluating systems’ performability is choosing the most suitable modeling technique and the evaluation method.
The successful implementation and tuning of systems encompasses several aspects within an organization: from business infrastructure to continuous personal educational capacity programs; from quality assurance management methods to tools and models for evaluation of quality issues related to the products or services delivered. Nevertheless, the return of such investment is immeasurable, since it might mean the real permanence of the organization in market.
Performability evaluation may be carried out through either measuring or considering system’s models. In many situations modeling is the chosen method either because the system might not yet exist or due to the inherent complexity of creating specific scenarios under which the system should be evaluate. However, one of the main modeler’s concerns, whenever evaluating systems’ performability, is to decide the most suitable modeling technique and the evaluation method. Several modeling formalisms are available, each one with its advantages and drawbacks. Closed-form models such as queuing networks, reliability block diagrams and fault trees are some of most widely adopted models for performance, availability and reliability evaluation. These models represent the systems’ structures and their evaluation are based on the interaction of systems’ components concerning system’s structure. They enable us to represent component’s networks through serial and parallel compositions and provide closed-form formulas. Nevertheless, such models cannot thoroughly handle event dependencies that are often faced when representing resource sharing, maintenance policies, redundant mechanism and fault dependencies. State based methods, on the one other hand, can easily consider those dependencies, so allowing representing complex redundant mechanisms as well as sophisticated maintenance policies. On the other hand, they suffer from the state space explosion.
Many academic and practical studies have been conducted in our group for performance and reliability evaluation of systems. This works encompasses applications on computer architecture tuning, distributed systems (middleware and web-services evaluation), computer network, data-base server systems, manufacturing systems and logistics