For service providers, this means faster time-to-market and lower capital and operational costs. Traditional methods of extending geographic reach through physical infrastructure buildout have failed, particularly in the broadband space--leaving service providers and their customers with nowhere to go. Meanwhile enterprises continue to spend millions of dollars maintaining national networks that connect remote offices and partners to complex private corporate network infrastructures. New multirouter technology promises to solve these problems, allowing companies to rethink the fundamental role of the router in their networks to save time and money.
The Legacy
Traditional routers have evolved in terms of performance and functionality, but their basic architecture has remained the same. Router vendors typically design their products to be deployed and managed by the organization that buys them (be it a service provider or enterprise). The smallest CPE routers and the largest core routers are based on a single centralized processor that performs routing calculations and recalculations, exception processing, and router management (see Figure 2a). The edge of the network is experiencing exponential growth, requiring a new class of router with high port densities and scalable processing power.
Monolithic routers don't scale control plane processing, (e.g., the ability to compute and recompute routes) because of their single route processor limitation. While newer routers deliver incremental scalability from a pure I/O perspective, they only address part of the problem. The more pressing issue is the routing horsepower--CPU and memory--available to each interface. Carriers need to support dynamic routing protocols running across a large number of the access links to meet increasing demands for dynamically routed customer traffic. Traditional router design forces them to deploy large numbers of monolithic routers at the network edge, at great capital and operational expense.
Virtual Routing
Virtual routing was conceived as a better way to support multiple customers concurrently using VPNs.
Virtual routing is a logical subdivision of a single route processor permitting the calculation of multiple routing tables (see Figure 2b). This was intended to create economies of scale and reduce network complexity. With virtual routers, carriers can theoretically offer private routing services simultaneously to multiple customers from a single platform.
But there are two fundamental flaws with virtual routers: reliability and scalability. Since all customers within a virtual router share a single route processor, associated memory, and a single operating system image, isolation between customers is virtual. Any underlying problem with the software or hardware could adversely affect all virtual routers. As the number of virtual routers increases, the routing horsepower per virtual router also decreases. This results in design tradeoffs, such as processing and route table capacity.
Security risks (e.g., denial of service attacks) to one router potentially become security risks to all routers in the system, robbing precious resources from other routers and possibly preventing them from having the CPU cycles to service routing updates. Virtual router architectures also do not provide the simultaneous support of router provisioning or fully independent network management for individual router customers. Even if a carrier wanted to provide its customer with complete management control, it couldn't, because of the inability to hand over control of shared resources through a multitiered management system. Multitiered management that can provide one set of autonomous, isolated controls for the owner and another for the user hasn't been available until now.