On Quality of Service Management

Abstract : A quality of service (QoS) management framework for systems is presented that satisfies application needs along multiple dimensions such as timeliness, reliability, cryptographic security and other application specific quality requirements. In this model, end users' quality preferences are taken into account when system resources are apportioned across multiple applications such that the net system utility accrued to the end users is maximized. The framework facilitates QoS tradeoff through a semantically rich (in terms of expressiveness and customizability) QoS specification interface that enables the end users to give guidance on the qualities they care about and the tradeoffs they are willing to make under potential resource shortages. The interface also allows the user or system administrator to define fine grained service requests easily for multi-dimensional complex QoS provisioning. Furthermore, by introducing the abstraction of Quality Index, which maps qualities to indices in a uniform way, and by the mathematical modeling of QoS Tradeoff and Resource Tradeoff, we transform the QoS management problem into a combinatorial optimization which ultimately enables us to quantitatively measure QoS, and to analytically plan and allocate resources. A series of optimization algorithms is developed that tackle the QoS management problem which is provably NP hard. The first set of algorithms treats the problem of maximizing system utility by allocating a single finite resource to satisfy the QoS requirements of multiple applications along multiple QoS dimensions. Two near optimal algorithms are developed to solve this problem. The first yields an allocation within a known distance from the optimal solution, and the second yields an allocation whose distance bound from the optimal solution can be explicitly controlled by a QoS manager.