Getting Ready for the End of Circuit Switching

The demise of circuit-switched networks continues to progress unchecked as the primary advantages of circuit-switched networks – guaranteed delivery, stable latency and network resiliency, to name a few – have been captured by innovative advances in Ethernet technology. Ethernet offers significant cost advantages over circuit-switched technologies such as SONET, and the move from circuit-switched to packet-based traffic has enabled much more efficient utilization of network connections through shared network links.

With the standardization of metro Ethernet, it is possible to reliably connect enterprise networks over the WAN. These connections, however, have become quite complicated as enterprises increase the number applications they are running, the types of data supported (i.e., voice and data), and the range of advanced features they must support, including scalability, redundancy and high availability. Additionally, QoS issues, such as preserving the priority of traffic between remote sites over the WAN, require greater network visibility and control as well as priority mechanisms that separate traffic flows.

The Rise of Multidimensional Ethernet
In its most basic form, Ethernet would be unable to meet the exacting real-time and performance demands of the metro. However, Ethernet technology continues to evolve, matching the reliability and performance of circuit-switched networks while exceeding their capabilities with value-added functionality such as QoS. For example, VLAN and VPN technology have enabled network administrators to manage effectively the increased complexity of converging networks. Additionally, advanced traffic management capabilities made possible by Ethernet, such as the ability to provide unique bandwidth and priority guarantees for each customer sharing a pipe, have enabled service providers to offer exciting, innovative, and highly profitable network services.

While Ethernet clearly owns the enterprise, the large-scale metro network has posed many challenges for Ethernet, making it appear that circuit-switched networks would be able to maintain their market share. However, recent advances in Ethernet technology threaten to push circuit-switched technologies out of the metro as well.

Multidimensional Ethernet overcomes the final barriers to Ethernet deployment in the metro through five key technologies: Hierarchical QoS, Ethernet cross-connect, MAC-in-MAC, Ethernet Automatic Protection Switching (EAPS), and hardware-based instrumentation.

Hierarchical QoS
Supporting latency-sensitive applications such as VoIP over high-speed connections carrying tens of thousands of flows requires deep QoS support. Through Hierarchical QoS, hardware-based controllers manage bandwidth and priority on a per-subscriber basis, with multiple QoS levels per subscriber. Implementing bandwidth control in hardware allows the Hierarchical QoS to scale to 40 times more subscribers per system than previous deployments, with no impact to the switch performance.

Ethernet Cross-Connect
Efficient service provisioning requires the ability to make flexible yet simple connections to multiple content providers. Ethernet cross-connect capabilities enable cost-effective switching of individual subscribers to different and multiple content providers, by handling service connections at Layer 2. Service providers now can serve business and residential customers from the same network. This convergence of business and residential traffic offers tremendous benefits to network operators. The consolidation of infrastructure equipment using an Ethernet cross-connect architecture yields a deployment cost up to eight times lower than traditional Layer 3 approaches.

MAC-in-MAC
The robustness of Ethernet has progressed through many stages. Layer 3 MPLS improved performance by cleanly separating forwarding control functions from payload data, thereby leveraging the speed of Layer 2 switching into Layer 3. Unfortunately, this required that already burdened core routers would have to maintain a separate Virtual Route Forwarding table for each VPN and that private routes associated with each subscriber be shared with service providers, forcing subscribers to partially relinquish control and security of their network.

The introduction of Layer 2 MPLS provided inherently more flexibility than Layer 3 VPNs since Layer 2 switches can ignore the higher layer protocols of the immediate network, as well as eliminated the need for provider switches to maintain any routing information (i.e. private routes) for a subscriber’s network.

The proposed IEEE p802.1ah standard (also known as MAC-in-MAC) builds upon existing virtual metropolitan area network (vMAN) technology to provide MPLS like scalability in an all-Ethernet network. vMANs provide isolated tunneling through VLANs, enabling service providers to separate subscriber traffic while providing different levels of QoS. While vMANs do improve scalability, they are still insufficient for use in large metro networks and provider backbones because of the 12-bit (4096) limitation of the VLAN ID field.

MAC-in-MAC inserts a Provider MAC address along with a much larger service ID field in each frame that is transparent to customer data. The result is an Ethernet network that economically can scale to millions of service VLANs, more than 4,000 times the number supported by traditional VLAN and vMAN deployments. MAC-in-MAC also offers carrier-class scalability and reliability, reduced complexity and seamless interoperability with vMANs without the complexity of MPLS.

Ethernet Automatic Protection Switching
An open protocol defined in IETF RFC 3619, EAPS is a ring-based resiliency protocol that increases network availability and uptime as well as provides sub 50ms restoration failover response. EAPS is a software protocol that does not require special hardware and can run on any port of the Ethernet switch. This allows economical deployments with both modular and fixed configuration switches. EAPS also supports a variety of advanced capabilities typically found only in SONET systems. One example of this is dual-node interconnect where two rings are connected by two separate switches, allowing communication to continue in the event of an outage of one of the interconnect switches. Finally, EAPS allows for simple capacity upgrades using additional ports or even link aggregation groups for the ring connection.

Hardware-based Instrumentation
Switches long have had software-based monitoring capabilities such as RMON and S-Flow to help identify and categorize user traffic. But as the number of subscribers on a switch scales for metro applications, software-based systems fail to maintain switching performance under load. Zero-performance impact hardware counters dedicated to monitoring per-subscriber/per-service performance at line rate enable real-time embedded SLA monitoring for improved visibility into network and application performance. This increased visibility increases network efficiency, and new XML APIs allow simple integration with carrier-scale OSS/BSS systems.

Through multidimensional Ethernet, all of the pieces are in place for Ethernet to provide circuit-switched capabilities and reliability to metro Ethernet networks – deep QoS support, efficient service provisioning, carrier-level scalability, high availability, and accurate monitoring – tolling the final bell for circuit-switched networks.

 
Peter Lunk is senior product manager for service provider solutions at Extreme Networks Inc. He can be reached at plunk@extremenetworks.com.

Extreme Networks Inc. www.extremenetworks.com

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines