| New levels of development for the modern generation of enterprise networking By Amnon Ptashek |
| Enterprise networking is facing a new generation of speed, architecture, and applications. New market trends such as requirements for high Ethernet speeds – 1 Gigabit to the desktop and 10 Gigabits to the backbone/servers – as well as the need for higher security, high availability, and even mobility through WLAN, have drawn a new and unfamiliar need for redesign of networking switches both in the hardware and software levels. The following article discusses the criteria driving the new enterprise-networking infrastructure. Impact from industry standards In the last three years, the leading industry standard organizations, mainly IEEE and IETF, have presented new standards for the silicon industry focusing on merchant switching ASICs and components to address those new architectures. The changes touched a wide spectrum of architectures from the physical layer, like cable testing and Power over Ethernet to the MAC layer, with link aggregations, improvement in the classical spanning tree, and robust Layer 3 with redundancy capabilities utilizing VRRP and ECMP to the switching gear architecture. Thus, these changes presented a new generation of stackable architecture for this new wave of enterprise switches. Hardware and software requirements Getting deep into the new hardware architecture necessitated by switching ASICs are highly complicated Layer 2 and Layer 3 forwarding, filtering, and switching machines. These switching ASICs operate on Gigabit and 10-Gigabit speeds while they have to operate in conjunction with the trend of convergence around a centric IP network. That means enabling voice, video, and data for utilizing different IP packet fields and techniques that include IEEE class of services and IETF Quality-of-Services (QoSs), such as DiffServ. Those traffic constraints require implementing buffering techniques, many priority queues, and interfacing with counters for implementing traffic engineering, i.e., policing, shaping, and rate limitation. Traffic engineering is only one example. Multicasting is another. New applications that enable multicasting of multimedia traffic over the Internet require protocols such as Protocol Independent Multicast (PIM) to build multicast trees between the source and the receiving stations. Switching ASICs implement those tree topologies, one to many, while maintaining traffic constraints for QoS. However, the new generation of switching ASICs is just a partial component of the entire picture. The control software is the key component, which can optimize architectures and enable new applications. ASICs with APIs The new switching ASICs are coming to market with intelligent Application Programmable Interfaces (APIs). These intelligent APIs enable switch and router vendors to implement complicated systems such as high-availability chassis and redundant stackable switches with a short development cycle that enable quick time to market. Without intelligent APIs, it would be almost impossible to implement the entire requirement of the IP protocol and the system solution. New generation switching ASICs This new generation of Gigabit-scale switching ASICs with their unique multi-layer capabilities and intelligent control software are already available in the market. Companies such as Marvell are offering enterprise switching packet processors that meet the newly required capabilities. . . . . . |