Service providers today, and the equipment providers supplying these service providers, face a daunting task: preparing for the next generation of mixed services. Those include voice/VoIP, IP-VPN, Ethernet LAN extensions, Internet access, Fiber channel, and Escon. Services providers want to accomplish this while continuing to deliver todays traditional revenue-generating voice and T1/T3 (PDH) data services.
In fact, most public and private communication networks are facing a fundamental shift away from voice traffic toward a preponderance of data traffic, including voice-over-IP and Ethernet on their transport links. This shift is putting increasing pressure on the traditional circuit-based transport network SONET and SDH, its European equivalent. These time-proven and widely installed networks for voice need to become more efficient for non-voice traffic. This is not an insignificant challenge, because SONET was designed from the beginning to support time-division-multiplexed (TDM) traffic and not the bursty traffic flows typical of data communications. Consequently, traditional SONET/SDH transport networks tend to be less bandwidth-inefficient when confronted with data applications. With new software and semiconductor technologies, however, data-over-SONET/SDH can provide an efficient way to transport data services over today's SONET/SDH transport network.
Ethernet plays a key role, because it is understood and supported by virtually all enterprises. The controversy over how to send Ethernet traffic in metropolitan area networks is quickly settling on SONET as the answer. The issue remains how to support new IT & broadband data services and make SONET better and less expensive without requiring a forklift upgrade, while maintaining support for legacy networks. Ethernet over SONET (EoS) and Gigabit Ethernet over SONET (GeoS) are shaping up as key drivers and new revenue sources in the communications market.
To address the technical challenges presented by multiservice networks, the engineering community has to provide T1 voice, a PBX connection alongside an Ethernet connection to the Internet. The community has devised two different approaches. The newest, IEEE non-standard approach, dubbed "pseudo-wire", combines T1 TDM and Ethernet services at the packet level and outputs a packet stream to the network, whether contained within PDH or SONET payload. The newest ANSI & ITU standardized technology takes a different approach, and puts the Ethernet packets into a TDM payload, combined with traditional T1 data services. This combined TDM stream is then connected to the wide area network, whether within PDH or SONET. One aspect that makes this second approach so interesting now is it is based on General Framing Protocol (GFP), a new ITU standard that specifies the mechanics for switching and aggregating information in such mixed service networks. The standard has recently been expanded to cover PDH trunking requirements, enabling a mix of private line and switched data services over traditional copper or optical access network.
Now, new framer ICs bring two important innovations to the SONET market that distinguish them from their predecessors: support for virtual concatenation (VC) and for the GFP standard along with complete Layer 2 Ethernet interworking, aggregation, and switching. These new ICs are fully compatible with other framers as well as with capabilities in companion switching, mapping, and Ethernet/IP/MPLS chips, enabling OEM designs, to support multi-rate, multi-protocol, and multi-channel solutions over PDH and SONET. These are key capabilities for service provider customers who need to provide their customers with a wide range of networking options.
Generic Framing Protocol
Defined under the ITU G.7041 specification, GFP is a protocol-agnostic frame delineation and encapsulation mechanism for transporting Ethernet or any arbitrary data service packets. GFP is a framing procedure for octet-aligned, variable-length payloads for mapping into SONET/SDH synchronous payload envelopes. GFP packets map into SONET/SDH VT/VC12 or STS/STM frames. GFP also provides encapsulation for higher layer protocols such as Ethernet and PPP. GFP comes in two flavors: frame-mapped and transparent.
In the case of Gigabit Ethernet and Internet Protocol, the frame-mapped method is the preferred option for full layer 2 interworking. In essence, an Ethernet access network can be combined with private line T1 services and each service can be aggregated and switched as VT1.5 (1.5Mb/s) frames. Combined with Ethernet/IP Layer 2 aggregation and switching, a service provider can efficiently aggregate, manage, restore, and transport any Ethernet and IP (T1/T3 IP, POS, etc.) service within and through existing PDH & Sonet/SDH network(s).
GFP allows flexible and efficient adaptation of multiple protocols for transmission over SONET & PDH networks. GFP has extremely low overhead requirements and robust frame delineation qualities. As a result, GFP makes it easier for service providers to perform switching tasks, avoiding overhead-intensive flag-based adaptation schemes, such as packet over SONET used for IP router interconnection over wide area networks. This simplifies component complexity, lowers chip count and required real estate, while reducing equipment size and manufacturing costs. With an increasing number of transport protocols being used over SONET, such as point-to-point protocol (PPP) and high-level data link control (HDLC), service providers need an efficient method to support multiple protocols with simplified hardware and software designs. By doing so, equipment and network efficiency can be enhanced, and ultimately costs can be reduced.
Virtual Concatenation
Concatenation is a method used to bind several individual signals together to create a super-rate logical channel and thus more bandwidth than is available in one of the constituent signals. Contiguous concatenation, an existing but inflexible technology, requires that all intermediate network nodes support large contiguous concatenation functions. Many installed network elements in SONET/SDH networks cannot support non-blocking contiguous concatenation and would be prohibitively costly to upgrade or replace. Virtual concatenation provides network elements at both ends of the signal path the ability to send/receive arbitrary individual SONET payloads/virtual circuits in a concatenated group. So rather than have to upgrade all the network elements in the network to support larger payloads, VC confines upgrade costs to the ends of the path, simplifies management into a single group, enabling fully scalable, completely non-blocking, data services. Virtually concatenated links are therefore intelligible between the source and sink network nodes and can interwork independent of the network elements in between.
Virtual concatenation is one of the big reasons that Ethernet over Sonet (EoS) has become a force in the metropolitan networking sector. Developed by the International Telecommunication Union, the virtual concatenation spec allows carriers to provision any SONET (SDH) pipes on an VT (VC12), STS-1 or STS-3 (STM1) basis. Therefore, designers can bind together the appropriate number of STS-1s to support a Gigabit Ethernet stream while provisioning a different set of VTs to handle a Fast Ethernet stream. As a result, VC enables the equipment (and ultimately the service provider) to provide multiple right-sized channels for an increasing number of data applications over SONET. Combined with Layer 2 Ethernet aggregation, switching, and security, the service provider can offer fully transparent VLAN and VPN services and achieve >95 percent bandwidth efficiency on the access transport networks.
For example, transporting a Gigabit Ethernet signal over SONET (GEoS) using older-technology transport structures would result in a 42 percent bandwidth efficiency when using an STS- 48c (2.5 Gbits/s) contiguously concatenated channel commonly found on older Sonet/SDH networks. With the older concatenation approach, the network provider could slice up his capacity in an OC-48 link into units of STS-1 (51 Mbit/s), STS-3 (155 Ambit/s) or STS-12 (622 Mbit/s). However, it was difficult to mix and match the various pipes because there would be no guarantee that the same contiguous pipes would be available at both ends and throughout the network. The new approach solves this problem using VC that allows different-sized pipes. With VC, the GEoS network provider can use 95 percent of the transport channel for data applications and the remaining six STS-1 channels for voice applications. In some multi-protocol applications, VC may reduce wasted capacity by as much as 75 per cent. Ultimately, VC enables mixing and sizing in any way the service provider desires and the IT customer demands.
In addition, VC enables new service levels, opening the possibility for enhanced revenues. With VC it is possible to assign any amount of protection capacity to a signal in a virtually concatenated group that requires some level of capacity insurance. This means a service provider can guarantee bandwidth availability for a particular application, and charge a premium for such service.
Service over SONET
Todays LAN extension services typically are transported as private line service and terminated as frame relay, ATM-UNI, or IP. Using Data over SONET/SDH features, service providers can offer LAN Extension services via direct Ethernet connection across the user-to-network boundary. Using next generation SONET features such as virtual concatenation, the bandwidth of the transport channel can be "right-sized" for the data service. Additionally, the LAN extension services can be statistically multiplexed over a shared SONET/SDH channel for bandwidth efficiencies. These features allow for lower cost services by providing less expensive handoff interfaces and for bandwidth efficiencies through the finer transport granularity of GFP & virtual concatenation and statistical multiplexing.
Todays Internet Access services typically are handled by DS0/DS1 private line service. Using Data over SONET/SDH features, service providers can offer Internet access services via direct Ethernet connection across the User-to-network boundary. Additionally, the LAN extension services can be statistically multiplexed over a single shared SONET/SDH channel for additional bandwidth efficiencies. The added benefit for this type of service is the ability to provide a single handoff to the IP backbone network. With N Internet access clients on the edge of the network (ex. 20 clients each with 100 Mb ports), there can be a single, traffic aggregated, higher speed interface (GbE) to a backbone router. This approach provides service providers with a lower port count and therefore lower overall solution cost. With the new Data over SONET/SDH features, such as GFP linear encapsulations, the traditional SONET features of fault isolation, performance monitoring, and fast restoration are maintained.
Efficient and cost effective Data over SONET/SDH solutions will affect the access network and the metro network. Access networks need to provide 1) direct mappings of data protocols to SONET/SDH at the edge of the network and 2) efficiently sized transport pipes that match the end customer data bandwidth requests. Via GFP, the network service provider can provide direct packet-to-SONET encapsulation without adding additional protocol layers (as mapping through ATM would do) and without adding variable bandwidth inflation. New products in this access space, such as low cost multi-service aggregation devices in the form of pizza boxes, already provide these capabilities.
Metro networks need to aggregate Data over SONET/SDH flows from various access points into more efficient transport pipes. This is accomplished by using next generation SONET features, such as GFP linear mappings and virtual concatenation. The linear mapping provides a unique customer tag for each data service. The service provider controls the tag, allowing them to guarantee privacy of customer data even over shared SONET transport channels. This mechanism is for point-to-point data services, so network maintenance, performance monitoring and restoration are handled by legacy SONET/SDH mechanisms.
For point-to-multipoint data flows, layer 2 protocols such as RPR can be used to provide additional packet aggregation in metro ring topologies. Also, with RPR the maintenance, performance monitoring and protection can be handled via the RPR protocol itself. Therefore, RPR over SONET/SDH uses the SONET/SDH framing capability, but efficiently uses other traditional SONET/SDH features. Protocols such as RPR will have a positive efficiency benefit and will provide further services. However, the protocols will take longer to gain traction due to more complicated management, billing, and operations issues. Also, depending on the amount and type of traffic and number of access points located in the access network, ggregation of data service in the future will occur closer to the network edge rather than the metro network core.
Data over SONET/SDH services will evolve from basic services to more advanced services where the opportunity for greater network efficiencies and revenue generation will be realized. Today Data over SONET/SDH services consist mainly of point-to-point services without any statistical multiplexing. This is quickly evolving to flexible topology (point-to-multi-point, ring, linear, mesh) connections with statistically multiplexed data services over a SONET channel(s), which yields greater network efficiencies, connectivity, and customer services.
Silicon advances market
Advances in semiconductor technology are giving chip designers higher processing performance and more effective real estate to add functionality. For example, GFP and VC are now supported in next-generation SONET/SDH framer ICs that also include a wealth of valuable traffic management capabilities. The expanded capabilities and modularity of these new ICs changes the OEM design constraints and promises service providers new levels of flexibility. These new ICs enable OEMs to develop generic trunking architectures for delivering/transporting mixed services over any arbitrary trunk type, copper or fiber. With GFP, packets heading toward the Internet, to another Ethernet connection, to the corporate LAN, or to the TDM voice network are aggregated within the SONET architecture, enabling new services to co-exist with old services on the existing network.
Mappers and framers are complex ICs required in each node of todays global telecommunications systems. These devices enable the logical access to transmission lines as well as the transport of client signals on these lines. Framers synchronize to the transmission frames and allow access to the payload and overhead channels. Mappers insert and extract the client signals into and from the payload containers and perform the required rate adaptations, for example PDH mappers map/demap PDH signals into/from a SONET/SDH payload container. Framers and mappers must, therefore, support various line rates as well as client signal rates and types and numerous communication protocols to support a multitude of networking situations.
Hence, one of the key components for SONET network interfaces is the SONET framer, which interfaces directly to the optical transmission equipment. Framing designates or marks channels within a bit-stream, providing the basic timeslot structure for telecom. Framing creates an underlying pattern using a regularly spaced marker to provide byte-level synchronization and divide the bit-stream (which is always transmitting) into separate channels and overhead (which enables status information to be sent without affecting payload). SONET framing is based on the STS-1 bit-stream, which transmits at 51.84 Mbits/s or 810 bytes every 125 microsec. using On-Off keying. In receive mode, SONET framers descramble the line signal, synchronize to the transmission frame, allow access to the payload and overhead channels, and perform demultiplexing of the lower rate tributary signals contained in the received frame. In transmit mode, they perform these same functions inversely.
Other SONET/SDH framer ICs available today address squarely the multi-protocol, multi-rate issues with gigabit Ethernet (GbE) over SONET (GEoS). These devices combine gigabit Ethernet standard products together with add-drop multiplexer (ADM), VC, and other functions on a single die, again reducing an OEMs time-to-market while improving on a systems cost, power and size. Traditional Ethernet is a simpler and less expensive solution for moving data traffic over available bandwidth; however, it is only optimized for point-to-point or mesh topologies and not SONET/SDH network ring topologies. hus, Ethernet cannot take advantage of inherent ring topology functions to implement protection mechanisms.
New GEoS solutions, however, can transport GbE frames over existing SONET/SDH rings networks by taking advantage of virtual concatenation to map and un-map arbitrary groups of STS-1/VC-4 signals, building higher capacity payloads whereby a stream of data is divided among a set of STS-1/VC-4 carriers as it is transmitted, then realigned and multiplexed as it is received. This capability enables network OEMs to connect to SONET networks more quickly and easily, reducing overall system power and cost for Ethernet over SONET (EoS) and packet-over-SONET (PoS) applications.
The network implications
The SONET infrastructure is likely to remain the dominant transport medium in optical networks for some time to come. The useful life of SONET will be extended as SONET evolves to become more efficient and flexible in supporting new transport services. OEMs designing CPE for the SONET market should seek out component suppliers that can offer complete solutions, as well as technology that can extend an OEMs investment in hardware and software design. The new market economics demand that designers consider at least four aspects of a vendor's offering:
Expertise in integrating both optoelectronics and silicon: Vendors that have products and support teams in both areas are best able to help customers reduce the time and expense of developing new communications equipment while boosting speed, bandwidth efficiency and time-to-market. Next generation network equipment will need to support supercharged SONET/SDH transport, integrated optical networking, dense wave division multiplexing (DWDM) and multi-service interfaces on demand, including Ethernet and time division multiplexing (TDM), to deliver significant economic benefits to service providers.
Semiconductors that meet stringent SONET timing and jitter requirements: In the high speed synchronous networking world, timing is crucial, much more so than in traditional data communications applications. Designing and building and verifying functional elements such as STS-1 processors embedded within a framer IC that meet SONET timing requirements are tasks that demand experience in both silicon design and SONET implementation.
Pluggable components: Optical components that meet the telecom requirements for reliability and durability should also be pluggable. This pluggable feature gives equipment vendors and service providers the capability to switch devices as needed to accommodate different needs for speed and distance support. Even more important, pluggability means that OEMs can build boards partially populated and add components later when new services need to be turned on.
High functionality ICs. To remain competitive, OEMs need to be working with silicon at the cutting edge in functionality and performance. IC suppliers, in turn, need to be on top of advances in silicon design and manufacturing technologies as well as up-to-date on communications standards and evolving market requirements. Advances in silicon technology enable IC engineers to pack more features and functions on-chip, while reducing device costs and power requirements. Knowing what features and functions to include involves a deeper understanding of the market requirements and the customers' design goals.
Chris Hamilton is Director of Access Systems at Agere Systems (Allentown, PA).