Network services usually chain multiple types of Virtual Network Functions (VNFs) together in a specific order, known as service chain. One important issue of providing network services is reliability, which means that each type of VNF in a service chain acts properly on its function. The effective way to guarantee the reliability is to utilize sufficient redundancy through providing VNF backup instances beyond active primary ones. Software Defined Networking (SDN) enables the dedicate location pickup of primary and backup instances on switch-connected servers. The deployment of primary and backup VNF instances plays an important role of flow routing because of its influence on the transmission latency of the flow. In order to minimize the total latency of flows, we first formulate the reliable service chain deployment as a mathematical optimization problem. A detailed analysis of both software and hardware failure models is studied. Then we propose optimal solutions for the case of a single flow. Next, we prove the NP-hardness of our problem for the case with multiple flows and propose a heuristic strategy. Additionally, a performance-guaranteed solution is included if the server capacity is infinite. Extensive simulations are conducted to evaluate our proposed solutions compared to multiple algorithms.

中文翻译：

---

延迟高效的VNF部署和路径路由，可提供可靠的服务链

网络服务通常以特定顺序将多种类型的虚拟网络功能（VNF）链接在一起，称为服务链。提供网络服务的一个重要问题是可靠性，这意味着服务链中的每种类型的VNF都可以正确地发挥其功能。保证可靠性的有效方法是通过提供超出活动主实例的VNF备份实例来利用足够的冗余。软件定义网络（SDN）支持在交换机连接的服务器上对主实例和备份实例进行专用的位置提取。主VNF实例和备用VNF实例的部署在流路由中起着重要作用，因为它会影响流的传输延迟。为了最大程度地减少流的总延迟，我们首先将可靠的服务链部署公式化为数学优化问题。研究了软件和硬件故障模型的详细分析。然后，我们针对单流情况提出了最佳解决方案。接下来，我们针对多流情况证明了问题的NP难点，并提出了启发式策略。此外，如果服务器容量是无限的，则包括性能保证的解决方案。与多种算法相比，进行了广泛的仿真以评估我们提出的解决方案。如果服务器容量是无限的，则包括性能保证的解决方案。与多种算法相比，进行了广泛的仿真以评估我们提出的解决方案。如果服务器容量是无限的，则包括性能保证的解决方案。与多种算法相比，进行了广泛的仿真以评估我们提出的解决方案。