Circuit-switched networks: Will they last forever? Or are they ready to give way to a new ecology of communications networks based on datagrams and packets? These are the questions that nobody seems to be able to answer these days.
Current telecom transport networks, built with pieces ranging from add/drop multiplexers (ADMs) to digital crossconnects and optical switching systems, are almost entirely comprised of Sonet or SDH systems. They are well standardized (though not terribly interoperable), as reliable as Swiss watches, and familiar to telecom engineers worldwide.
Their shortcomings, however, have gotten a great deal of attention in the past four years, as entrepreneurs have looked to engineer a New World Order based on Internet and Internet Protocol (IP) technology. Sonet is, after all, ill-adapted to bursty data traffic, while the Internet grooves on Ethernet-based interface rates of 10 Mbit/s, 100 Mbit/s, and 1 Gbit/s, adapting to shifting traffic patterns and logical topologies. Sonet is about nailing up pipes, leaving them for eternity, and charging a decent fee to make certain they never go down for more than an imperceptible 50 milliseconds. This has served a network of voice callers well, but IP found itself either uncomfortably constrained by narrow pipes, or equipped with overprovisioned connections it could never possibly fill. (Think of a 51-Mbit/s STS1 provisioned for a 10-Mbit/s Ethernet interface. No matter how hard you drive it, 41 Mbit/s will always go unused, on both working and protect paths.)
Many startups tried to evolve the trusty Sonet ADM by integrating packet and cell switch fabrics, circuit switch fabrics, and even DWDM uplinks. But most if not all of this new equipment employed proprietary bandwidth management, mapping and framing techniques, so the large incumbent service providers did not bite. At one time, an ample supply of greenfield carriers embraced these developments, but there aren’t many of those folk left. So, it became clear that if Sonet and SDH were to survive, they needed to do it in incumbent networks – and that means they needed standards.
Along came Nortel Networks Corp. (NYSE/Toronto: NT) and Lucent Technologies Inc. (NYSE: LU), with a proposition to save Sonet for future generations: Generic Framing Procedure. In short, GFP is a traffic adaptation protocol, designed to support variable and fixed-packet transport over a general-purpose, high-speed communications channel. In other words, it’s flexible, and it can deal in both worlds – whether it’s bursty IP data traffic or the more steady stream of voice or storage interconnect.
GFP has a few key strengths. For Ethernet services, compared to packet-over-Sonet (POS), Frame Relay, or other framing procedures that employ HDLC (High-Level Data Link Control), GFP is a much more elegant mapper, with a very low, deterministic overhead and low processing requirements, making chips much cheaper to create. Additionally, GFP preserves the relevant MAC layer information for Ethernet, from destination address through Frame Check Sequence (FCS), supporting a true Ethernet private line that matches the operations requirements of a traditional TDM private line. For storage services, GFP can operate in what’s called Transparent Mode, mapping Fibre Channel, Escon, Ficon, or even digital video into fixed GFP frames. Thus, on the same system, GFP can support a mix of traditional services and emerging ones, managed as transport connections just as Sonet/SDH circuits are today. Operators love that kind of thing.
GFP represents a key element of next-gen Sonet/SDH. It’s well standardized in American National Standards Institute (ANSI) T1X1.5 and the International Telecommunication Union, Standardization Sector (ITU-T) and could be an agreeable method for incumbent providers to extend the useful life of Sonet/SDH networks while opening the door for multiprotocol transport over OTN (optical transport network, once called digital wrapper).
But it gets even better! Bringing the whole next-gen Sonet/SDH family together are virtual concatenation and LCAS, the Link Capacity Adjustment Scheme. Virtual concatenation supports much finer granularity circuit provisioning (groups of STS1 circuits or even VT1.5 circuits), without requiring the service provider to upgrade existing Sonet/SDH gear. This allows for “right-sized” pipes for packet services. A customer could take what it has in place already, such as a 10/100-Ethernet port on an enterprise switch, and allocate Sonet bandwidth as needed, boosting up the bandwidth in time. This effectively creates an Ethernet private line that scales cost-effectively, without the need for new customer equipment.
LCAS, which does not have as many supporters, is designed to provide operators with greater flexibility in provisioning virtual concatenation groups (VCGs). This helps adjust bandwidth in service and provide protection options. Someday, LCAS could be used to dynamically adjust bandwidth based on signaling from attached devices, such as switches or routers, but that may take a while (lest it suffer the fate of ATM switched virtual circuits). In the meantime, LCAS will likely be used for scenarios when a customer requests more bandwidth on an Ethernet private line.
At the OFC conference last year, a representative from SBC Communications Inc. (NYSE: SBC) said he hoped to manage their edge network as a single logical element, via a common management system. This multiservice edge would support just about any network protocol imaginable, including things such as TDM, Ethernet, storage traffic, and digital video. All of this would be transported over their Sonet-based network via well accepted standards. GFP, along with virtual concatenation and LCAS, is making this vision possible.
Today, most vendors are working on adding GFP interfaces to their Sonet ADMs, while a few router vendors are adding GFP to their POS interface cards to support virtual concatenation of POS links directly from the router. You’ll even see some Ethernet switch vendors implementing GFP interfaces as an alternative to POS.
Looking at vendor developments today, there are two main classes of GFP product:
1) Multiprotocol Metro Switch: An OC48 or OC192 Sonet system with ports for TDM, Ethernet, and in some cases storage or RPR using GFP to provide a standardized method of mapping and framing. As this model evolves, look for new visions of a “GFP Switch,” a sort of black box with GFP on the outside, circuit and packet switching on the inside, providing a true multiservice edge platform.
2) Multiprotocol CPE/CLE: This comes in the form of a pizza box or an even smaller, cheaper wall unit that sits in a wiring closet and provides direct access from a subscriber to a carrier's transport network, supporting a mix of Ethernet and TDM on the client side and an OCx port on the network side. Fiber is still required to the building, but these are the key customers an out-of-region ILEC intends to poach. The multiprotocol CPE is under development today, but not much is shipping. Look for lots of announcements later this year.
So are the customers really biting? So far, the answer seems emphatically yes, and this applies to worldwide markets. In North America, the incumbent local exchange carriers (ILECs) are planning to attack their competitors out of region using GFP to take services from subscribers (T1s, Ethernet, and even storage, in some cases) and backhaul them over leased circuits to their metro POPs. The invading LEC can buy a private line from the incumbent that connects directly from the customer to the invader’s POP, so no network equipment need be collocated in the incumbent’s central office, avoiding the associated access charges and crossconnect charges. The ILECs may find themselves freed of CLECs, but now they must face each other, and it will certainly get ugly. GFP gives them an interesting tool, if they choose to go that route.
This backhaul model is interesting – it creates the need for a very scaleable aggregation platform for GFP-based transport services. It may come as small surprise that Lucent, Nortel and Tellabs Inc. (Nasdaq: TLAB; Frankfurt: BTLA) have devoted significant resources to the Lambda Unite, OPTera Connect, and 5500 NGX, respectively; this must certainly fuel the hopes of startups in this space, such as Mahi Networks Inc. and Polaris Networks.
But the open question remains: Is all this talk of next-generation Sonet/SDH just forestalling the inevitable, throwing good money after obsolescence? If your plan really is to head out of region and snatch customers away from rivals, why not look to a next-gen pure-packet solution? Putting multiple services over a shared packet infrastructure certainly appears to be more efficient than working with Sonet, even if GFP and virtual concatenation give you greater control over circuit provisioning. But let’s remember that these are ILECs we’re talking about – and that includes ILEC sales people, support personnel, and back-office managers. They’re used to dealing with rock-solid transport systems. The new money is still in basic connectivity services such as Ethernet private line, storage transport, and virtual RPR rings among multiple corporate locations. The ILECs know how to do transport, and that’s where I would put my bet. GFP is a good card to play in this market.
In 2003 you’ll start seeing GFP show up in marketing material, product brochures, and grand visions of a unified transport network architecture. GFP makes for good positioning, though someone needs to come up with a better name: Future Sonet? The Next Next Sonet? Not Your Father’s Sonet? Brave New Sonet? Well... It obviously won’t be me.
— Scott Clavenna, Director of Research, Light Reading