Gao Xianming

Software Data Plane and Flow Switching Plane Separation in Next-Generation Router Architecture

Routers have traditionally been architected as two elements: forwarding plane and control plane through ForCES or other protocols. Each forwarding plane aggregates a fixed amount of computing, memory, and network interface resources to forward packets. Unfortunately, the tight coupling of packet-processing tasks with network interfaces has severely restricted service innovation and hardware upgrade. In this context, we explore the insightful prospect of functional separation in forwarding plane to propose a next-generation router architecture, which, if realized, can provide promises both for various packet-processing tasks and for flexible deployment while solving concerns related to the above problems. Thus, we put forward an alternative construction in which functional resources within a forwarding plane are disaggregated. A forwarding plane is instead separated into two planes: software data plane (SDP) and flow switching plane (FSP). SDP is responsible for packet-processing tasks without its expansibility restricted with the amount and kinds of network interfaces. FSP is in charge of packet receiving/transmitting tasks and can incrementally add switching elements, such as general switches, or even specialized switches, to provide network interfaces for SDP. At last, we make an experiment on our platform in terms of bandwidth utilization rate, configuration delay, and the processing time of a simple router. Our experimental results show that the separation of SDP and FSP brings greater modularity to router architecture, allowing operators to optimize their deployments.

Two-Tier Model for Supporting Network Functions Virtualization with FoCES

Network functions virtualization (NFV) is used to leverage standard virtualization technologies to consolidate multiple network functions onto industry standard equipment. And most network functions consist of two layers including control plane and data plane. So we put forward a two-tier model for supporting network function virtualization based on ForCES framework. It includes three key components: CP, NF, and NF Manager. NF Manager judges which CP is active and establishes communication channel between NF and CP. NF is regarded as one of network functions such as switching elements, mobile network nodes, and other network-wide functions. CP runs the NF terminals to manage corresponding NFs and configures NF topology to customize service chaining. Besides, this model should include authentication and authorization mechanism to ensure communication security and NF security, message priority to provide different service of quality for different network functions, and CP redundancy or CP failover to reduce the affect introduced by a single point of failure. This paper mainly puts forward a general model for network functions virtualization, points out key-components, and lists current technologies related with NFV. At last, it designs two typical frameworks based on our proposed two-tier model. We hope our analysis about device framework for supporting network functions virtualization can enhance the speed of deployment of network functions virtualization.

Research of Packet Forwarding among Virtual Routers in Identical Logic Network

There are two mapping relationships between virtual routers and physical platform: many-to-one and one-to-one. When multiple virtual routers that are in identical logic network cohabit a shared router platform of physical network, we propose an Interiorly Fast Forwarding Mechanism (IFFM) based on study of Open Shortest Past First (OSPF) and investigation in generation of FIB. This mechanism is proved in which reducing delay is the yardstick of solving fast forwarding among multiple virtual routers. Its purpose is to eliminate the hops of packet forwarding and to advance the speed of packet transmitting. Meanwhile, this mechanism enhances network robustness by reducing network loss ratio that is caused by single point of failure.