Next-generation communications system architectures present a number of difficult new challenges, especially when implementing internal intercard and interchassis interconnects for handling multiprotocol traffic. In the past, designers attempted to address these
challenges by using either in-house or merchant-proprietary approaches. Yet they have been hampered by the higher cost, longer development cycles, and lack of flexibility of the solutions. It's clear that next-generation systems will be built on a foundation of standards-based interconnects. The leading standards-based communications interconnect choices are Ethernet and Advanced Switching Interconnect (ASI).
Ethernet offers an excellent protocol for long-haul, multihop out-of-the-box communication links with robustness, recoverability, and routing intelligence built into the core transmission technology. But, Ethernet has shortcomings for implementing multiprotocol backplane links within a system. Encapsulating other protocols offers great flexibility for transmission of multiprotocol traffic over IP-based networks, but this can result in an untenable overhead burden for implementing short multiprotocol backplane connections inside a system.
If a system requires layer 3 switching or higher, then add-on approaches such as using a TCP/IP offload engine (TOE) to reduce the higher-level overhead are typically required. This approach tends to just "move the problem" rather than solve it, requiring extra bandwidth and more processing hardware, consuming more space and power, and ultimately increasing system cost.
In comparison, ASI offers an off-the-shelf alternative standard solution that is simpler to implement. Because the key elements for multiprotocol backplane links (quality-of-service, source-based routing, redundant links, etc.) are all implemented in hardware, there is no need to add on a TCP/IP management layer or to set up MACs with complex address routing at each end point.
ASI is based on PCI Express, allowing designers to leverage familiar physical and link layers while taking advantage of ASI functionality such as true peer-to-peer communication, multihost support, protocol encapsulation, robust quality-of-service (QoS), high availability, and redundant connections. Future edge routers can now pass TDM, cell, and packet traffic over a single physical connection.
ASI's path-based routing method differs from the table-based routing of Ethernet. The sending line card defines the transmission path. All subsequent traffic forwarding is handled without the need for extra memory to support routing tables or incurring the processing overhead for handling routing table search-and-lookup functions. This leads directly to lower-cost switch components. In addition, source-based routing provides a unique "signature" that can be used as an inherent security mechanism to validate transactions, prevent spoofing, or limit access to a unique set of senders.
Key considerations when making the choice between Ethernet and ASI in a communications system are the number of protocols being transported, the bandwidth requirements, and the degree of intelligence on each line card. ASI has advantages when:
However, if many discrete chassis are interconnected, Ethernet is the proper choice. Both technologies will coexist in the future, with ASI offering a standards-based alternative to Ethernet that is leaner on hardware and simpler to implement while providing more robust traffic-flow capabilities.
The ASI Special Interest Group is a nonprofit collaborative trade organization tasked with developing and supporting a switched interconnect and data fabric interface specification for communications, storage, and embedded equipment. More information can be found at www.asi-sig.org.