中文摘要：

随着人们对网络服务的依赖性日益增强，网络服务系统的服务质量和可用性变得至关重要．然而，服务系统的日益复杂化及第三方软件和异构中间件系统的大量部署，为各种软件缺陷提供更多隐蔽空间．这对保证服务质量和系统可用性提出挑战．文中提出一个自适应的选择性再生框架，以保证系统的服务质量．选择性再生框架，把基于请求处理路径的性能故障分析诊断技术和软件再生技术相结合，通过及时诊断和再生系统中性能故障的部件，改善请求的响应时间，保证系统服务质量．文中提出了一个新的基于请求路径的性能故障的诊断方法．同时，文中实现了一个层次化的选择性的再生策略．基于RUBiS系统的实验结果表明文中方法可以有效保证多层服务系统的服务质量．

英文摘要：

Quality of service and availability are crucial for Internet service system. However, service systems become increasingly complex, and deployed with various heterogeneous middle- ware and commercial off-the self software without source code. These introduce more hidden spaces for software defects, which challenge the traditional approaches for guarantee system＇s Service Level Objectives, or SLOs. In this paper, the authors propose an Adaptive Selective Re- juvenation Framework to guarantee SLOs in multi-tier service systems. The selective rejuvenation framework combines software rejuvenation technique with request tracing technique to identify and selectively rejuvenate the performance faulty components. The authors implement a novel re- quest tracing technique in the rejuvenation framework, which reconstruct request causal paths precisely, and analyze request causal paths by clustering similar paths into path patterns and se- lecting relevant path patterns. Meanwhile, the authors implement a hierarchical rejuvenation scheme, which identify and do proactive rejuvenation of performance faulty components, before performance degradation becomes severe. The evaluations are given with RUBiS benchmark, and the results show that the request path analysis approach is effective, and selective rejuvenation framework can effectively improve system＇s Quality of Services.