TC: Small: Exploiting Software Elasticity for Automatic Software Self-Healing

服务器应用程序中的软件故障是保持系统可用性的重要问题。 在没有完美的软件，本研究的重点是容忍和恢复错误，利用软件的弹性：正常的代码恢复某些故障时，低级别的故障被操作系统或适当的仪器屏蔽的能力。 软件弹性是利用引入救援点，位置在应用程序代码中处理程序员预期的故障，这是自动重新利用和测试，安全地使故障恢复从一个更大的类的意外故障。 救援点从未知故障中恢复软件，同时通过模仿已知错误条件下的系统行为来维护系统完整性和可用性。 它们使用模糊识别，使用检查点重新启动机制创建，并进行测试，然后使用二进制补丁注入生产代码。 这种方法掩盖了失败，允许继续执行程序，同时最大限度地减少不希望的副作用，使应用程序恢复和软件自我修复。

Software failures in server applications are a significant problem for preserving system availability. In the absence of perfect software, this research focuses on tolerating and recovering from errors by exploiting software elasticity: the ability of regular code to recover from certain failures when low-level faults are masked by the operating system or appropriate instrumentation. Software elasticity is exploited by introducing rescue points, locations in application code for handling programmer-anticipated failures, which are automatically repurposed and tested for safely enabling fault recovery from a larger class of unanticipated faults. Rescue points recover software from unknown faults while maintaining system integrity and availability by mimicking system behavior under known error conditions. They are identified using fuzzing, created using a checkpoint-restart mechanism, and tested then injected into production code using binary patching. This approach masks failures to permit continued program execution while minimizing undesirable side-effects, enabling application recovery and software self-healing.