In recent years data has surpassed circuit-switched voice and TDM leased lines as the predominant traffic on telecommunications networks. As result, carriers have increasingly deployed IP routers and Ethernet switches to handle the rapid growth in data traffic. Carriers face a quandary, however, because current IP/Ethernet networks cannot handle constant bit rate trafficsuch as voice, DS-1/E1 and DS-3/E3that continues to provide the bulk of telecom carrier revenues.
What's needed is a way to adapt Ethernet from a best effort protocol optimized for IP data services to a multi-service platform capable of carrying time sensitive traffic (DS-1/E1 N x 64 kbps, DS-1/E1 and DS-3/E3) with the same quality of service as more expensive TDM networks. There are several advantages to such a converged multi-service platform including the elimination of redundant cabling, hardware and administration required for separate voice and data networks; more efficient use of network resources; lower operating and capital costs; and increased revenue opportunities.
The biggest challenge to achieving this vision of a converged packet-based network is duplicating the quality of service (QoS), signaling, and constant bit rate (CBR) management features provided by legacy TDM voice infrastructures. The two primary methods for transporting time critical services over EthernetTDM over IP and Ethernet plus TDMtackle these QoS issues in very different ways.
- TDM over IP solutions (TDMoIP) use in-band circuit emulation in order to packetize, prioritize and manage TDM traffic to be transported over Ethernet links. Special techniques such as prioritization, tagging, traffic shaping, buffering, bandwidth reservation and management are required to ensure adequate QoS. TDMoIP is designed to work with installed PBXs and PSTN protocols.
- Ethernet plus TDM solutions use an out-of-band hardware based approach to transport TDM traffic over Ethernet links without using any Ethernet bandwidth. Ethernet plus TDM works with installed PBXs and PSTN protocols, but unlike TDMoIP it avoids the need for packetization, prioritization and management of the time critical packets to ensure QoS.
In this paper, we'll examine the differences between TDMoIP and Ethernet plus TDM technologies, focusing on the design challenges of each approach and how they have been addressed. We'll also look at the pros and cons of each approach and evaluate how well each one fulfills the goal of creating a full service Ethernet platform for carrying IP data and TDM traffic.
Emulating Circuits vs. VoIP
TDMoIP and Ethernet plus TDM were both developed to address the shortcomings of voice over IP (VoIP), particularly its lack of support for TDM signaling and protocols, as well as issues concerning QoS.
VoIP was developed as a means to transport voice as packets over IP networks, rather than over the traditional PSTN. VoIP solutions promise lower costs, more efficient use of bandwidth, improved scalability, and innovative features compared to existing circuit switched networks. In contrast to circuit emulation techniques like TDMoIP and Ethernet plus TDM, VoIP solutions do not seek to emulate TDM circuits, services, or signaling.
As shown in Figure 1, a typical VoIP system consists of IP phones and PBXs that packetize voice traffic, and gateways that route packets to other parts of the LAN or WAN. Central-office softswitches are used to replicate the functions of voice circuit switching equipment in a packet environment. VoIP systems use evolving protocols such as H.323 and the session initiation protocol (SIP) to convert voice signals into data packets, route data efficiently to their destination, reassemble the packets, and correct for errors.
Figure 1: Diagram of a typical VoIP system.
The primary design challenge for VoIP is to ensure that the increased delay and distortion inherent in sending voice over IP networks does not lead to a significant degradation in the QoS compared to circuit switched networks. IP, TCP, and UDP transport layers do not guarantee any level of quality of service other than best effort. Even the real-time transport protocol (RTP), which contains timing information, does not guarantee timing performance (Table 1).
There are several network protocols and techniques that can be used to improve quality of service to an IP network. Differentiated services (DiffServ) is an example of a protocol that relies on prioritization of traffic in the routing layer. The reservation protocol (RSVP) enables reservation of resources on the network. 802.1p protocols allow for classification of frames at the Ethernet layer and 802.1Q is used for configuring virtual LAN (VLAN) ports and traffic classes. However, VoIP must always assume the network has the reliability and sufficient bandwidth to enable transport of time-critical data.
Another significant issue with VoIP is signaling, and how to match the myriad call setup, functions and services of circuit switched voice networks over a predominantly data oriented infrastructure. The VoIP approach to this challenge is to develop a parallel infrastructure that seeks to replicate and replace each segment of the voice infrastructure rather than to interface and complement it.
TDMoIP: The Emulation Approach
TDMoIP is a means of emulating TDM circuits over next-generation IP infrastructure. This technology was developed to address the shortcomings of VoIP, particularly its lack of support for TDM signaling and protocols, as well as issues concerning QoS. Circuit emulation provides virtual circuits within an IP network to support existing TDM equipment and services, without changes to functionality or signaling of existing telephony equipment.
As shown in Figure 2, TDMoIP systems offer standard T1/E1 interfaces that connect to existing PBXs and CO switches and support PSTN protocols and signaling standards. In TDMoIP the synchronous bit stream is packetized and prioritized. QoS options are configurable with VLAN used for layer 2 priority (802.1p and Q), type of service for IP level priority, and known UDP Ports for layer 4 priority.
Figure 2: Diagram of a typical TDMoIP system.
There are efforts underway within the Internet Engineering Task Force (IETF) to standardize techniques for carrying TDM traffic across IP networks in the PWE3 Working Group. The PWE3 Working Group is focused on developing the standards for the encapsulation and emulation of "pseudo wires" across IP and multiprotocol label switching (MPLS) packet switched networks. Internet drafts include TDM service specification for pseudo-wire emulation edge-to-edge, TDMoIP, and TDM circuit emulation over a packet-switched network, but no single proposal has yet garnered consensus. Alternatively, the Metro Ethernet Forum Circuit Emulation Services (CES) working group is in the process of defining circuit emulation over Ethernet, potentially working in conjunction with ITU-T Study Group 15. Issues being considered in supporting time sensitive traffic over a packet network include framing, signaling, clocking, jitter and wander, messaging, alarms, service performance, frame structure and application requirements.
The key design challenges for TDMoIP include:
- How to effectively emulate TDM circuits in a packet data network
- How to provide seamless integration with existing TDM protocols and signaling
- How to ensure that QoS is not degraded by packet delay and distortion
The first challenge, circuit emulation, is provided by converting TDM traffic into packets that are forwarded over the IP network and reconstituted at the destination (by removing IP headers, concatenating the packets, and regenerating the clock). This process is analogous to VoIP conversion of voice signals into packets, except that TDM systems have the added requirement of regenerating the traffic into its TDM format (e.g. T1/E1) including the associated signaling and protocols.
The second challenge for TDMoIP is to provide seamless integration with existing TDM systems. To accomplish this task TDMoIP provides physical layer extension over IP, which makes the system transparent to protocols and signaling and enables support of legacy services and interfaces. This is accomplished by maintaining the integrity of framed or unframed T1/E1s and by providing support for cross network clock regeneration. Naturally, standard T1/E1 interfaces are provided that are compatible with ports on existing PBXs and circuit switches
The third challenge is providing QoS that is comparable to that currently provided by the legacy TDM infrastructure. QoS is the ability of a network to assure that it can meet traffic and service requirements. QoS performance criteria include network downtime, network error performance, lost transmissions, fault detection, and recovery. A key element of QoS is the ability to provide service contracts for time sensitive traffic. Although this goal is at odds with the basic principles of Ethernet, which has no mechanism for clock synchronization or recovery, QoS techniques exist outside of the Ethernet standard to overcome this limitation. These techniques include prioritization, tagging, traffic shaping, buffering, bandwidth reservation and management.
TDMoIP itself does not provide mechanisms to ensure time delivery guarantees and relies on lower-layer services. In addition to using Layer 2/3 priority tags, switching and routers within the TDMoIP stream need to be configured in respect to those priorities. TDMoIP works only in a relatively benevolent environment, with sufficient bandwidth, network reliability and QoS aware switches and routers (Table 2).
Ethernet plus TDM: Tapping into Ethernet
Ethernet plus TDM, like TDMoIP, is a means of emulating TDM circuits over point-to-point Ethernet links. In contrast to TDMoIP, however, Ethernet plus TDM does not accomplish this task by converting TDM traffic into packets, which must be prioritized and managed. Instead, Ethernet plus TDM taps into overhead in the Ethernet header to create a constant-bit-rate (CBR) TDM link that transparently operates over the existing Ethernet pipe without jeopardizing dc balance on the network (Figure 3).
Figure 3: Diagram of a typical Ethernet plus TDM system.
The Ethernet plus TDM approach works by multiplexing TDM circuits, a management circuit, and a reference clock (for synchronous traffic) into a framer, and then combining the aggregated time-sensitive constant bit rate traffic on the Ethernet data stream in the physical coding sublayer (PCS). The Ethernet interfaces are standard physical layer (PHY) and media access control (MAC) interfaces such as MII, TBI or GMII, and transceivers operate at standard 100 or 1000 Mbit/s rates. The network segment behaves as an Ethernet link, and the Ethernet data passes unaware of the additional constant bit rate traffic.
The CBR channel can be used to transport a variety of time critical circuits, such as:
- DS-1/E1 N x 64 kbit/s
- Operations, Administration and Management (OAM)
- Synchronous Clock
Because the Ethernet plus TDM scheme operates as a dedicated circuit, the load on the Ethernet link does not affect the quality of service of the TDM path. The TDM circuits have guaranteed toll-quality low latency and jitter, as they do not contend for the data pipe. Since the TDM circuits behave as a hard-wired circuit, Ethernet plus TDM, unlike TDMoIP, does not require VLAN tagging, priority labeling, or buffering of data to ensure QoS. Avoiding the complexity of traffic management reduces the cost of equipment.
In addition to providing TDM transport, the Ethernet plus TDM approach can be used to provide a private management channel for remote Ethernet CPEs. Using the out-of-band CBR channel in the form of a UART circuit, a network operator can perform traditional telephony administration and management such as loopback, alarm indication, signal detection, dying gasp, redundancy controls and read/write of remote registers. The management signaling, which is sent out-of-band, is not mixed with users traffic. Because the management channel is on a private dedicated circuit, the speed of signaling is constrained only by the physical media, and occurs in the order of microseconds. It is thus suitable for time-critical TDM management functions such as physical layer status monitoring (Table 3).
Winning out in the Long Run
While VoIP has grabbed attention as an effective way to deliver voice services over IP links, time has proven that the QoS issues associated with VoIP make it a less attractive option for carriers and equipment designers. TDMoIP and Ethernet plus TDM provide a viable option for bridging the gap between the IP and voice worlds. But, before implementing these techniques, designers must closely evaluate the pros and cons of both of these technologies.
- Circuit Extension over IP: The Evolutionary Approach to Transporting Voice and Legacy Data over IP Networks, by Yaakov Stein, Chief Scientist, RAD Data Communications, Ltd.
and Eitan Schwartz, Vice President and General Manager, TDMoIP Technologies, RAD
Data Communications, Inc.
- Ethernet Moves Into the Metro, Nan Chen, Metro Ethernet Forum, OE Magazine, August 2002
- TDM Circuit Emulation Service over Packet Switched Network (CESoPSN), Network Working Group Internet Draft, A. Vainshtein, I. Sasson, and A. Sadovski, Axerra Networks; E. Metz, KPNQwest; T. Frost, Zarlink Semiconductor; and P. Pate, Overture Networks.
- TDMoIP Concepts and Trends, Eitan Schwartz, VP and General Manager, TDMoIP Technologies, RAD Data Communications, Ltd.
- TDM over IP: PWE3 Working Group Internet Draft, Yaakov Stein, Mordechai Anavi, and Ronen Shaashoua, RAD Data Communications
- TDM Service Specification for Pseudo-Wire Emulation Edge-to-Edge (PWE3), Prayson Pate, Overture Networks, Ron Cohen, Lycium Networks, and David Zelig, Corrigent Systems
- Technology 101: TDMoIP, RAD Data Communications
- Transforming Enterprise Communications: An IDC White Paper Sponsored by Nortel Networks, Pim Bilderbeek
- Voice Performance Over Packet Networks, Danny De Vleeschauwer and Jan Janssen, Alcatel Network Strategy Group
About the Author
Gerry Pesavento is the president and CEO of Teknovus. Previously, Gerry was a founder of Alloptic and an executive at DiCon Fiberoptics. He was also the founder of the Ethernet in the First Mile Alliance and sub-Track Chair of the IEEE 802.3ah Ethernet First Mile Task Force. Gerry can be reached at firstname.lastname@example.org.