|
Networks & Infrastructure
Move over SONET, here comes Ethernet transport
Benefits are clear, but wide-scale adoption will be evolutionary
by Michael Kennedy
An Ethernet-based metro transport vehicle is emerging now to take the place of SONET/SDH that has been dominant for more than a decade. The driving force for this transition is the preponderance of Ethernet use within the customer premise as well as the increasing use of IP/MPLS at the network core. Replacement of SONET/SDH in the metro area creates an end-to-end packet optical network that reduces complexity and cost while improving network resilience and performance.
The Ethernet transport concept also assumes some Layer 2 and Layer 3 functions that are provided by IP/MPLS today and were delivered by a combination of ATM switches and Layer 2 routers in the past. The PBB (Provider Backbone Bridges) and PBB-TE (Traffic Engineering) standards initiatives are part of that effort.
Ethernet is attractive because it is a simpler technology than SONET/SDH, ATM and IP/MPLS and it is very widely deployed. Very high manufacturing scale also makes Ethernet hardware very cost effective compared to IP/MPLS that has relatively low production volumes.
Architectural changes threaten the status quo among telecom systems vendors. In this case established vendors of carrier-class switch/routers are facing direct challenges by vendors of Ethernet switches used in the data center and potentially by optical transport vendors that have the opportunity to incorporate more Ethernet switching functionality into their DWDM transponders. A niche opportunity also exists for developers of the centralized control plane software and firmware used in Ethernet transport solutions. Soapstone Networks and Gridpoint Systems are the best known of these companies.
SONET/SDH transport has maintained its dominant market position for a long time because it is very good technology. It employs a deterministic design and when deployed in a ring configuration it provides service restoration in less than 50 ms. SONET/SDH network management is centralized. This provides more robust and efficient network performance and reliability, and takes cost out of each network element. Rigorous OAM (Operations, Administration, and Maintenance) is in place—this is essential for large telecommunications networks. Finally, SONET/SDH OAM does not require highly sophisticated (and paid) technicians—operations can be scaled up to a very large workforce.
Ethernet transport uses 1 GE and 10 GE interfaces instead of SONET/SDH interfaces and is designed to retain SONET/SDH’s strengths while addressing its weaknesses. Ethernet interfaces are much less expensive than SONET/SDH interfaces. Bandwidth efficiency is improved through Ethernet’s use of statistical multiplexing. Ethernet supports much higher speed services than those associated with SONET-based DS-1 and Frame Relay services. Ethernet supports VoIP which is more cost effective and bandwidth efficient than voice circuit switching technology.
The second element of Ethernet transport involves creating a solution for managing the flow of packets through the metro network. While IP/MPLS provides the required functionality it is very expensive and lacks the deterministic behavior of existing SONET/SDH and TDM multiplexer solutions. Most existing metro Ethernet solutions employ simpler and cost effective switched Ethernet VLAN solutions in order reduce the cost of the metro network solution. However, this approach depends on Spanning Tree protocol for creating a loop-free topology across the metro network. This approach involves some probabilistic behavior and does not optimize traffic utilization across the network. In addition, both IP/MPLS and Ethernet VLAN embody a distributed approach to packet routing/switching decision making where every network element is involved in the process. This increases cost and complexity.
Ethernet transport uses a centralized management system (Control Plane) that takes the place of distributed management systems such as Spanning Tree and IP/MPLS. This permits use of the same low cost Ethernet switches used in the VLAN solution while providing network control functionality that is deterministic and has functionality much like IP/MPLS. The centralized management approach in addition provides a level of control that is not easily attained via a distributed model. For example, it supports Connection Admission Control (CAC) which is used to protect Video on Demand (VoD) services from catastrophic failure caused by a dropped session. The long holding times of VoD sessions prevent their adequate control by traditional switching and routing techniques. As another example, centralized control enables network operators to immediately invoke contingency plans for network recovery that have been developed in advance rather than rely upon individual network elements to search for a viable configuration through a series of iterative attempts.
An Ethernet transport vehicle intended as a replacement of existing SONET/SDH metro area transport is emerging now. It builds on the process of transforming Ethernet from a tool for local area networking to a carrier-class transport tool. Though technical and market conditions favor Ethernet transport’s success this will be a slow process. The all encompassing nature of centralized management leaves no room for error or mistakes. Years of testing in low risk situations will be required before adoption on a wide-scale is achieved. The incumbent vendors may well adapt this approach as it reaches maturity. Though Ethernet transport is fully applicable to the network core change there will be even slower.
Michael Kennedy is co-founder and Managing Partner of Network Strategy Partners, LLC (NSP) www.nspllc.com —management consultants to the networking industry. He can be reached at mkennedy@nspllc.com.
|