Autonomic management of internet services: Experience with self-optimization

Autonomic computing is an appealing approach to facilitate the management of distributed systems. However, several challenges remain open among which: (i) How to manage legacy systems? (ii) How to manage complex distributed systems? In this paper, we precisely address these issues. The paper presents the design, implementation and evaluation of Jade, an environment for autonomic management of legacy computing systems. This environment relies on two main frameworks: a framework for administrable elements, which provides the administered legacy system’s elements with a uniform management interface; and a framework for autonomic managers, which regulates a set of managed elements for a specific management aspect. The aspect that we consider in this paper is self-optimization, i.e. preserving the system’s performance in response to wide variations of the load. The main contribution of this paper is a generic architectural model that eases the construction of autonomically managed systems. Thus, integrating a new managed legacy system only necessitates a few hundred lines of code. Another contribution of the paper is the application of this model to self-optimization of complex and multi-tier Internet services. Our experiments on an auction site show that Jade is able to maintain the client perceived performance stable when the workload varies, while, without Jade, the performance continuously decreases. Furthermore, in these experiments Jade allows twice as many client requests to be served as a system with no autonomic management