Virtual concatenation and the Generic Framing Procedure boost the ability of legacy networks to handle Ethernet-based traffic efficiently.
Mimi Dannhardt
PMC-Sierra, Inc.
Although there are various predictions regarding future trends, there is no question that today's communications landscape is dominated by two networking technologies: Ethernet in the local area network (LAN) and SONET or SDH in the public telecommunication companies' (PTTs') wide-area networks (WAN). Connecting remote sites via a single high-speed LAN for employee access to corporate servers, remote storage sites, and Web hosts represents a critical reason for Ethernet LANs to be extended over metropolitan, nationwide, or even international distances.
A number of technologies have been employed to transport LAN traffic over the PTT network -- frame relay, ATM, packet over SONET/SDH, multilink-PPP, and others -- each requiring inter-working the native Ethernet traffic to the transport protocol prior to transmission. In some service models, customers are required to inter-work their network traffic prior to handing it off to the public network; in others the carrier takes complete responsibility for the function. Both approaches call for specialized equipment for inter-working, and both create network management issues.
The inter-working function generally must terminate the Ethernet and map the underlying Internet Protocol (IP) traffic into a new Layer 2 (L2) or, alternatively, encapsulate the Ethernet within another L2 technology.
One of the challenges of Ethernet over SONET/SDH is the differing rates between the two technologies. Ethernet rates are typically 10 Mbits/sec, 100 Mbits/sec, or 1 Gbit/sec, always increasing in factors of 10. On the other hand, SONET/SDH rates are optimized for voice traffic and do not match the optimal rates for transporting the Ethernet data stream. These rate mismatches, illustrated in Table 1, make carrying a single Ethernet connection over a SONET pipe bandwidth inefficient.
Virtual concatenation
Virtual concatenation, based on International Telecommunication Union recommendation ITU-T G.707 (2000), is a technique that allows SONET/SDH channels to be multiplexed in arbitrary arrangements. This permits custom SONET/SDH pipes to be created that are any multiple of the basic rates.
Using virtual concatenation, the SONET/SDH transport pipes may be "right-sized" for Ethernet transport. For SONET, virtual concatenation rates are designated by STS-m-nv for high-order concatenation, where the nv indicates a multiple n of the STS-m base rate. Similarly, low-order virtual concatenation is designated by VT-m-nv (VT stands for "virtual tributary"). For SDH, the rates are designated by VC-m-nv. In effect, the SONET pipe size may be any multiple of 50 Mbits/sec for high-order virtual concatenation (STS-1 or VC-3), or 1.6 Mbits/sec (VT-1.5)/2.176 Mbits/sec (VC-12) for low-order virtual concatenation.
All the intelligence to handle virtual concatenation is located at the endpoints of the connections, so each SONET/SDH channel may be routed independently through the network. Equipment in the center of the network need not be aware of the virtual concatenation. This allows for deployment over existing SONET/SDH networks.
The pipe may be sized for the average bandwidth for a single connection or to provide a statistically multiplexed transport pipe. In virtual concatenation, data is striped over the multiple channels in the virtual concatenation group (VCG). Control packets, which contain the information required for reassembling the original data stream, are inserted in some of the currently unused SONET/SDH overhead bytes. This information contains the sequence order of the channels and a frame number, which is used as a time stamp. The receiving end-point is then responsible for reassembling the original byte stream. This includes compensating for differential delay that may have occurred by different routings or paths that the channels took through the network.
Dynamic bandwidth allocation
A specification for dynamically changing the bandwidth used for a virtual concatenated channel, the Link Capacity Adjustment Scheme (LCAS) is defined by ITU-T recommendation G.7042. Using this technique signaling messages are exchanged within the SONET/SDH overhead to change the number of tributaries being used by a VCG. The number of tributaries may be either reduced or increased and, in the absence of network errors, the bandwidth change may be applied without loss of data.
Under LCAS, bandwidth can be adjusted based on time-of-day demands and seasonal fluctuations. For example, businesses can subscribe to higher-bandwidth connections for backup or other applications when the demand for bandwidth, and hence the cost, is lower. LCAS can also provide "tuning" of the allocated bandwidth. If the initial bandwidth allocation is only for the average amount of traffic rather than for the peak bandwidth, and the average bandwidth usage changes over time, the allocation can be modified to reflect this change. This tunability can be used to provide (and charge for) only as much bandwidth as the customer requires.
LCAS is also useful for fault tolerance and protection, since the protocol has the ability to remove failed links from the VCG. As the data stream is octet-striped across the tributaries in the VCG, without such a mechanism if one of the tributaries has errors, the entire data stream has errors for the duration of the error within the tributary. The LCAS protocol provides a mechanism to detect the tributary in error and automatically remove it from the group. The VCG ends up operating at a reduced bandwidth, but the VCG continues to carry error-free data.
Generic Framing Procedure
GFP, based on ITU recommendation G.7041, is a protocol for mapping packet data into an octet-synchronous transport such as SONET/SDH. Unlike High Level Data Link Control (HDLC) -based protocols, GFP does not use any special characters for frame delineation. Instead, it has adapted the cell delineation protocol used by ATM to encapsulate variable length packets. A fixed amount of overhead is required by the GFP encapsulation that is independent of the contents of the packets. In contrast to HDLC, whose overhead is data dependent, the fixed amount of overhead per packet allows deterministic matching of bandwidth between the Ethernet stream and the virtually concatenated SONET/SDH stream.
The GFP overhead can consist of up to three headers: a core header containing the packet length and a cyclic redundancy check (CRC) that is used for packet delineation; a type header identifying the payload type; and an optional extension header. Frame delineation is performed on the core header, which contains the two-byte packet length and a CRC. The receiver would hunt for a correct CRC and then use the received packet length to predict the location of the start of the next packet.
Within GFP, two different mapping modes are defined: frame based and transparent.
Frame-based GFP is used for connections where efficiency and flexibility are key. To support frame delineation within GFP, the frame length must be known and prepended to the head of the packet. In many protocols, this forces a store-and-forward encapsulation architecture to buffer the entire frame and determine its length, which may add undesirable latency. Frame-based GFP is good for sub-rate and statistically multiplexed services, as the entire overhead associated with the line coding and interpacket gap (IPG) is discarded and not transported.
In transparent GFP, all code words from the physical interface are transmitted, making it appropriate for applications that are sensitive to latency or for unknown physical layers. Currently, only physical layers that use 8B/10B encoding are supported. To increase efficiency, the 8B/10B line code is transcoded into a 64B/65B block code and then the block codes are encapsulated into fixed-size GFP packets. As all physical layer bits are transmitted, the transparent GFP requires more bandwidth than frame-based GFP but has lower latency. Transparent GFP is primarily targeted at storage area networks (SANs) where latency is very important and the delays associated with frame-based GFP cannot be tolerated.
Supported service models
Private leased-line services, typically provided via ATM, frame relay, or multi-link frame relay, are widely used to interconnect business locations.
An Ethernet-based leased-line service could be carried through the SONET/SDH network using GFP encapsulation and virtual concatenation. Ethernet private lines may be provisioned at various service rates from 50 Mbits/sec to 1 Gbit/sec utilizing STS-1 concatenation and from 1.6 Mbits/sec to 100 Mbits/sec utilizing VT1.5 concatenation. Ethernet private lines deployed over SONET/SDH offer the reliability and broad service area coverage associated with the carrier infrastructure. As a private line, data-rate guarantees and security are key offerings, as well as upgradeable bandwidth utilizing the LCAS protocol to adjust the bandwidth supplied.
Virtual leased lines or virtual private networks (VPNs) are services in which many customers share the same transport bandwidth. This leads to more efficient use of the transport bandwidth via statistical multiplexing, and thus lowers costs. Since the transport bandwidth is shared, this service is generally a more economical offering than a private leased line but does not necessarily offer the quality of service (QoS) provided by the private leased line. Instead, service parameters are controlled with service-level agreements (SLAs).
Add/drop multiplexers as the customer interface
The PTT networks are based on SONET/SDH rings. Rings are connected to provide complete connectivity around a metropolitan area and from city to city (Figure 1). The four basic building blocks, or types of equipment, used to provide this connectivity are shown in Figure 2.
Of these four building blocks, three have customer-facing interfaces: add/drop multiplexers (ADMs), terminal multiplexers, and multi-service provisioning platforms (MSPPs).
Some forms of MSPPs are ADMs with data interfaces and some are routers with SONET/SDH interfaces and switching. ADMs have traditionally been used to provide PDH (T1/E1/T3, etc.) and SONET/SDH drops to connect to specific customers, and provide a good place for Ethernet over SONET/SDH customer interfaces.
With most applications at either end of a metropolitan- or wide-area network ring being IP- and Ethernet-based, some experts have predicted the replacement of SONET/SDH by gigabit-speed native Ethernet (according to the telecom market research firm RHK, Inc., 2.4 million metro Ethernet ports are expected to be deployed by 2006). Fortunately, the ITU-approved standards for virtual concatenation and GFP over SONET/SDH make it possible for long-haul carriers to offer Ethernet services over their TDM networks and should substantially extend the life of this mature transport technology. What's more, SONET/SDH offers features that native Ethernet doesn't, such as bandwidth guarantees and redundancy in the event of a cable break.
Mimi Dannhardt, is a technical advisor within the Product Research Group of PMC-Sierra, Inc. (Burnaby, BC, Canada).