Time-division multiplexing (TDM) over packet technology isn’t a novel concept. It’s already known by a number of different names, such as CESoIP, CESoP, CESoMPLS, TDMoIP, circuit emulation over packet switched networks (CESoPSN) and structured agnostic TDM over packet switched networks (SATOP).
In fact, the technology has developed to the point where supporting legacy traffic over passive optical networks (PONs) is able to be accomplished with relative ease.
At its most basic, circuit emulation services over packet, or CESoP, technology tunnels TDM services across a managed packet switched network (PSN). In the specific case of a PON network, TDM traffic is tunneled from the optical network terminal (ONT) to the optical line termination (OLT). TDM traffic is packetised, put into a packet payload, and subsequently transported across the PON.
The payload is then received at the OLT and converted to TDM, aggregated, and connected to the Public Switched Telephone Network (PSTN).
To successfully deploy CESoP technology in PON equipment, four key features must always be available:
Carriers and equipment manufacturers need products that meet standards to ensure compliance and interoperability. Standards for CESoP service have been issued for various networks, such as Multi-Protocol Label Switching (MPLS) and Ethernet. The EthernetPON (EPON) standards don’t deal with CESoP. This is best handled by adopting existing standards.
Five standards govern CESoP services in PSNs. Three are relevant to the PON standards—the two pseudowire standards and the Metro Ethernet Forum (MEF) standard. The pseudowire standards consist of two draft standards from the IETF, the unstructured SATOP service and the structured CESoPSN service. The MEF issued a standard for carrying TDM traffic over an Ethernet connection, namely the Implementation Agreement for the Emulation of PDH Circuits over Metro Ethernet Networks, or MEF 8. This standard supplements the IETF draft standards.
Unlike circuit-switched networks, PSNs don’t have a timing structure. Timing is critical to ensuring that the TDM system works properly, and to the standards. Any PON system will have to handle the timing of the TDM trunk. In a TDM system, there are two ways to pass timing across the network, with each depending on the network’s available services:
- First, timing can be transported using the differential method. In this case, there’s a reference frequency or signal at both ends of the trunk. This isn’t appropriate for EPON networks, unless a reference like GPS exists.
- The second method is the adaptive mode. In this mode, a reference is available at one end (typically the central office) of the network. Time information is conveyed across the PON network and then recovered in the receiving block.
The method used to recover the clock is very important. The simplest form of recovery is the buffer method. This approach was adequate for ATM networks, but it doesn’t work in a PSN. That’s due to the larger amount of packet delay variation (PDV).
Another form of clock recovery is the averaging method, in which arrival time of the packets is averaged over a period of time. While better than the buffer method, this approach can’t be employed for some network loads. A better method is to use statistical operations on the recovered clock in order to ensure that the clock is accurately recovered.
SUPPORT FOR SERVICES
Whichever TDM service is chosen, it has to handle structured, unstructured, and fractional TDM modes. These are standard T1/E1 services that must be supported. Unstructured service, also known as unchannelised service, accepts TDM traffic and packetises the whole frame without knowledge of the framing information. The frame bit is also transferred across the PON. In this way, unchannelised service can be thought of as bit-by-bit transfer.
Structured service, also known as channelised service, keys on the channel information or the digital service 0s (DS0) level information. The service is able to switch and groom the traffic.
To be most functional in a network, any TDM service should be able to support point-to-point (leased line service), point-tomultipoint (star configuration) and multipoint-to-multipoint (mesh configuration). Ideally, a PON network will support all three services, meaning both existing and emerging services can be easily supported.
An example of a new service is pseudo Tie Lines. Previously, companies used Tie Lines to connect PBX switches in a mesh network. This service fell out of favour because a T1/E1 circuit had to be leased for each connection. With the advent of the PSN, a CESoP function that supports multipoint-to-multipoint needs only one connection to the network. However, it can generate multiple packets, each with a different destination address. Thus, Tie Line service once again becomes viable.
The IEEE governs EPON standards. To bring the standards to fruition, the IEEE worked with the “Ethernet First Mile Alliance, ”which is an industry organisation. The standards that govern EPON are 802.3ah-2004.
CIRCUIT EMULATION OVER EPON
First deployments of EPON systems have carried triple-play service from the service provider to a home or to an apartment building. As deployments grow, there will be an increasing need to handle legacy traffic as well as timing distribution. By using TDM over packet techniques, service providers can handle legacy traffic.
Timing is critical for applications in PSNs, including EPON. The nature of the optical connection and the media access controller (MAC) for the OLT and ONT are such that both the PDV and the distribution of the packets or the standard deviation will be low. This is an excellent criterion for clock recovery.
Transporting a TDM signal through a PON network requires a good clock recovery mechanism. Because there’s no clock distribution inherent in the EPON network, the clocks must be carried and timed adaptively in the network. Adaptive timing is defined as the ability to recover the original clock information from the transmitted CESoP stream. Using statistical techniques on the recovered clock gives the best performance, since it minimizes the error induced by the network. The induced error can be packet loss, PDV, latency, or number of hops. Note that the CESoP stream doesn’t have to terminate at the OLT or the ONT (Fig. 1).
Figure 2 illustrates a typical network configuration and features for an EPON network. The network is a PON with only two ONTs shown. The CESoP stream is sent from the OLT to the ONT. Though TDM traffic can be terminated in the OLT and ONT, it’s not always the case. An integrated access device (IAD) is connected to the OLT through an Ethernet link. This link carries TDM traffic from an adjacent building. On the right-hand side, the OLT can terminate the TDM traffic as shown by its connection to the PSTN, or it can be terminated in a CES gateway located elsewhere in the network.
As carriers look forward to triple-play rollouts, they must also consider the continued need to support T1/E1 traffic. The key challenge is to devise a convergence strategy that will enable a cost-effective PSN to seamlessly carry valued, revenue- generating legacy services.
Standards-based CESoP products available today allow the transport of TDM traffic over the EPON network. By using this bridging technology, carriers can evolve toward a PSN infrastructure without endangering existing revenue streams.