Optimal software rejuvenation policies
Optimal software rejuvenation policies
复制标题
DOI：
10.1109/rams.2013.6517695
复制
发表时间：
2013-05
期刊：
影响因子：
--
通讯作者：
R. Agepati;N. Gundala;S. Amari
中科院分区：
文献类型：
--
作者：
R. Agepati;N. Gundala;S. Amari
对软件可靠性和性能的研究表明，长时间运行的软件系统显示出不断增加的故障率和/或其性能的逐步退化。这种现象被称为软件老化，它可能最终导致不可接受的系统性能下降和/或崩溃故障。一种叫做软件再生的技术可以用来对抗软件老化。这种技术包括偶尔终止应用程序或系统，清理其内部状态和/或环境，然后重新启动它。通过移除累积的错误条件并释放或整理操作系统资源，该技术可以主动防止未来意外的系统中断。与突然发生的故障导致的停机不同，与软件恢复相关的停机可以由用户或管理员自行安排，通常是在半夜或周末。本文提出了一种适用于广泛应用的基于状态的广义软件年轻化模型。复原模型包括一个随机退化过程，一组复原动作及其效果，以及识别系统退化的计划检查策略。利用马尔可夫决策过程获得了使系统总成本最小化的最优再生策略。稍加修改，该模型还可用于最大化系统可用性/容量。本文以一个受二维软件退化过程影响的web服务器为例，对所提出的模型和优化过程进行了验证。
Studies on software reliability and performance reveal that long-running software systems show an increasing failure rate and/or a progressive degradation of their performance. This phenomenon is referred to as software aging, and it may eventually lead to an unacceptable level of system performance degradation and/or crash failure. A technique called software rejuvenation can be used to counteract software aging. This technique involves occasionally terminating an application or a system, cleaning its internal state and/or its environment, and restarting it. By removing the accrued error conditions and freeing up or defragmenting operating system resources, this technique proactively prevents unexpected future system outages. Unlike downtime caused by sudden failure occurrences, the downtime related to software rejuvenation can be scheduled at the discretion of the user or administrator, typically during the middle of the night or over weekends. This paper presents a generalized condition-based software rejuvenation model that is applicable to a wide range of applications. The rejuvenation model includes a stochastic deterioration process, a set of rejuvenation actions and their effects, and a schedule inspection policy that identifies the system deterioration. The optimal rejuvenation policy that minimizes the overall cost associated with the system is obtained using Markov decision processes. With minor modifications, the model can also be used for maximizing the system availability/capacity. This paper demonstrates the proposed model and the optimization procedure using an example of a web server subject to a two-dimensional software degradation process.