Testing Transport Scenarios For Ethernet At 10 Gigabits/Sec

Oct. 8, 2008
10GbE may be viewed as an integral part of the transport environment, and therefore there are several possibilities on how to transport and test Ethernet traffic at this rate. By Guylain Barlow, Innocor Ltd. 10 Gigabit

10GbE may be viewed as an integral part of the transport environment, and therefore there are several possibilities on how to transport and test Ethernet traffic at this rate.

By Guylain Barlow, Innocor Ltd.

10 Gigabit Ethernet (10GbE) represents a ten fold data rate increase over Gigabit Ethernet (GbE), and the same holds true for GbE over Fast Ethernet. As a natural evolution, the classic test concepts borrowed from lower rate Ethernet apply to 10 Gigabit Ethernet. Here, “classic test concepts” refers to Ethernet tests such as the simulation of network data traffic using traffic stream generation or performance tests such as RFC 2544. However, at data rates in the vicinity of 10 Gbps, traffic is generally carried in the network backbone or as backhaul to a central office.

As a result, 10GbE may be viewed as an integral part of the transport environment. The best practical example comes from the IEEE 802.3ae 10GbE standard in the form of 10GBASE-R and 10GBASE-W. The latter, also referred to as 10GbE WAN, proposes the use of OC-192/STM-64 SONET/SDH to transport an Ethernet payload including the Physical Coding Sublayer (PCS). The former, 10GBASE-R, also called 10GbE LAN, represents the natural ten-fold data rate evolution from GbE. A key reason behind the adoption of 10GbE WAN is the omnipresence of SONET/SDH transport networks. The scenarios provided below outline several possibilities on how to transport and test Ethernet traffic at 10 Gbps.

Scenario 1: 10GbE LAN Backbone

By default, a 10GbE service implies the use of 10GbE LAN interfaces. 10GbE LAN may be used in a backbone configuration to interconnect multiple sites. A network diagram example is shown in Figure 1. In terms of maximum transmission distance, the 802.3ae standard provides clear guidelines. More specifically, what is considered an engineered link can reach up to 40 km at a wavelength of 1,550 nm. In such a case, testing link and interface hardware integrity is important. As in any protocol, transmission integrity errors are best detected at the lower layers, which in this case is at the 64B/66B PCS. In most cases, integrity errors will be reflected at the upper layers and reported as Media Access Control (MAC) Frame Check Sequence (FCS) errors. However, this does not always hold true, and the PCS sublayer provides more reliability and insight on the condition.

For example, a PCS coding error that occurs while transmitting Inter-Frame Gaps (IFG) cannot be detected at the MAC layer. The analysis of PCS errors will report more information, such as the block error ratio that reveals the number of errored 66-bit blocks over the total number of blocks. In addition, the PCS can detail whether the errors affect the 2-bit synchronization (sync) headers, and/or the remaining 64 bits in each block. Sync header errors can also be classified into isolated errors or more severe bursts causing a HI BER condition (16 invalid sync headers in 125 µsec) or a loss of lock (link down). Such test indicators may well be used in conjunction with the classic Ethernet tests at the upper layers. However, when using 10GbE LAN, the PCS sublayer gains in importance as it provides accurate indicators of transmission integrity.

Scenario 2: Using the SONET/SDH Infrastructure

As an example for this scenario, a service provider offers a 10GbE service to its customers. Let’s suppose the 10GbE service is required between distant sites and the service provider operates a SONET/SDH core network. In this case, the service may be delivered on 10GbE LAN interfaces and carried across the SONET/SDH backbone. This implies the use of a device that supports both 10GbE LAN and 10GbE WAN interfaces. An example of this configuration is provided in Figure 2. Since the maximum data rate of 10GbE LAN is greater than that of 10GbE WAN, this requires the use of flow control.

The 10GbE LAN line rate is 10.3125 Gbps and the 10GbE WAN line rate is 9.95328 Gbps. When accounting for the 64B/66B PCS encoding, the maximum 10GbE LAN data rate is 10.0 Gbps. For 10GbE WAN, after considering the SONET/SDH overhead and the PCS encoding, the maximum data rate is 9.294 Gbps. Flow control will regulate traffic transfer and help minimize data loss. The two available techniques consist of pause frames, which can be supported down to the lower rate Ethernet interfaces, or ifsStretch, which is specific to 10GbE. An interesting combination test case for this scenario is to run one-way tests between 10GbE LAN and 10GbE WAN ports. A useful tool for this test case is compatible auto-synchronized Ethernet payload on both interfaces such as a Pseudo-Random Bit Sequence (PRBS) that can be checked bit-by-bit at the receiving end.

Scenario 3: Using the OTN Infrastructure

Scenario 2 describes when the backbone uses ITU-T G.709 OTU2 interfaces. OTU2 interfaces were designed to transport OC-192/STM-64 client signals, although other client types may be used. Since G.709-based interfaces use Forward Error Correction (FEC), the distance between regenerators may increase when compared to standard SONET/SDH interfaces. With OTU2 interfaces a 10GbE WAN signal can be the client signal across the backbone. Consequently, a service that originates as 10GbE LAN may be carried over an OTU2 transport core. This implies the presence of a device equipped with 10GbE LAN and OTU2 interfaces that provides flow control capabilities.

Figure 3 provides a sample diagram. Similarly, a test device can help ensure Ethernet compatibility across both interface types. One way to achieve this is by generating and testing one-way traffic across both interface types. A test often used is to generate traffic from the 10GbE LAN interface with a loopback on the OTU2 network interface side. Conversely, tests can verify the conformance and performance of the transport device by generating traffic from the OTU2 port and looping back on the 10GbE LAN access side. This requires the generation of 10GbE WAN client payload into OTU2.

Scenario 4: Generic Transport with GFP

As networks evolve, a strategic endeavor is to carry multiple services and protocols over a common infrastructure. A protocol developed to help achieve this goal is Generic Framing Procedures (GFP). GFP was conceived for use over different types of infrastructures including SONET/SDH and the Optical Transport Network (OTN) OTU2 interface. With GFP-F (Frame-mapped GFP), these protocols include Ethernet MAC, Point to Point Protocol (PPP), and Resilient Packet Ring (RPR). As a result, GFP-F provides one more option to transport Ethernet traffic at 10 Gbps over an OC-192/STM-64 core.

In this case, a network device equipped with 10GbE LAN interfaces maps Ethernet MAC traffic onto GFP-F. Unlike scenario 2 that uses 10GbE WAN in the core, mapping to GFP-F eliminates the 64B/66B PCS sublayer. However, the Ethernet MAC frame information is fully preserved using GFP-F. The key benefit of this scenario is to concentrate various services including Ethernet within the same core network. Figure 4 provides a pictorial representation of three different ways to map 10GbE onto GFP using combinations of OC-192/STM-16 and OTU2. The concept of testing one-way traffic between a 10GbE LAN port and a port running GFP also applies. The same can be said for separate tests with a loopback on either the network interface side running GFP or the 10GbE LAN access interface.

Scenario 5: Point-to-Point with Forward Error Correction (FEC)

When considering Scenario 1, there exists a technique to extend the maximum transmission distance beyond the 802.ea standard. Quite simply, the overhead and FEC used in OTU2 can be reapplied directly to a 10GbE LAN client. This results in a line rate signal of 11.049 Gbps or 11.095 Gbps, hence different from the OTU2 line rate of 10.709 Gbps. The ITU-T G.709 recommendation for OTU2 documents the use of fixed stuff columns specifically for OC-192/STM-64 clients. With these fixed stuff bytes and the exact same overhead and FEC, wrapping a 10GbE LAN signal results in a line rate of 11.095 Gbps. Without these fixed stuff bytes, which means that the corresponding area of the payload is filled with data, the resulting line rate is 11.049 Gbps.

At the present time, there is no ratified specification that specifies these two line rates. However, the digital wrapper overhead fields and standard FEC are identical to those in OTU2. Since a large number of transport equipment manufacturers can already provide network interfaces that run at these two rates, they can be viewed as de facto standards. This scenario has the benefit of requiring no Ethernet flow control and extending the transmission distance for point-to-point 10GbE LAN applications. Figure 5 illustrates the use of the 11.049 Gbps and 11.095 Gbps rates.


10GbE is the evolution of Ethernet at a higher data rate. Since it is at the top of the Ethernet hierarchy from a data rate perspective, it is used in network backbones. The adaptability of Ethernet and global network evolution entails that there are multiple ways to carry 10 Gbps Ethernet traffic across a network. This provides new opportunities for vendors and service providers, and also implies that test techniques are required to help address these recent developments. The five 10 Gbps Ethernet transport scenarios described above are becoming a reality. Service provider networks will use these scenarios based on the specific advantages of each case. As 10GbE becomes entrenched and networks evolve, it will be interesting to see how the next highest Ethernet rate will impact the network core.

Guylain Barlow is the product manager for the test and measurements division at Innocor. He can be reached at [email protected].


Product URL: Click here for more information


To join the conversation, and become an exclusive member of Electronic Design, create an account today!