Core MPLS Switch Tenor Networks Inc. more or less invented this category. The core service switch is meant to act as an upgrade to large-core ATM switches (think Cascade GX550). These switches provide deterministic IP QOS, MPLS adaptation at the edge of the core network, and wire-speed switching of MPLS aggregated flows through the core of the network. In this camp companies like Tenor, Point Reyes Networks Inc., Vivace Networks, and Maple Optical Systems are leveraging MPLS to create switches that transform service provider transport networks into IP services networks. With these platforms, every service is transformed into IP/MPLS, including voice, making these switches quite similar to core ATM switches of old, just much more scaleable and IP-efficient. The challenge: These are damn hard to build; carriers aren’t convinced deterministic IP QOS is achievable; and many worry that MPLS will become as complex and difficult to scale as ATM. In other words, these companies are in a rather anxious moment in history: They’ll either pay off huge or fizzle. There isn’t much room in between.

Optical On-Ramp Kind of a hokey term, but this camp is distinct from the MPLS switch crowd in its use of multiple switch fabrics to accommodate a variety of services in their native formats. The 6500 from Tellabs Inc. (Nasdaq: TLAB; Frankfurt: BTLA) is one example; or the recent 1630 SX from Alcatel SA (NYSE: ALA: Paris: CGEP:PA). Add to this the forthcoming switch from Équipe Communications Corp., which is the next-gen core switch that Ascend meant to build but could not (everyone responsible for this program left for startups, including Équipe). These products are extensions of the “bandwidth managers” of recent years offered by Alcatel and Lucent Technologies Inc. (NYSE: LU). They terminate private lines and perform grooming and crossconnect of TDM (time-division multiplexing) circuits. They typically have an ATM fabric to manage data traffic, and now, in this iteration, have a means of migrating these services to MPLS. They reside at the same location as the core MPLS switch — they just don’t require that everything be converted to MPLS. This makes them more palatable to the large entrenched carriers that aren’t able to make wholesale changes to their network cores as quickly as some of the newer entrants or greenfield operators.

Service Edge Switch Instead of switch fabrics over 100 Gbit/s like the core switches above, these tend to have fabrics in the 5- to 20-Gbit/s range, and focus their attention on micro-flows (below DS3) rather than the macro-flows handled by core switches. Because these systems reside much closer to the end user, they are often responsible for true service creation, performing subscriber management functions. In this camp the variety of features is wide, making it difficult to position vendors neatly. Some include a transport and Layer 1 protection function, others leave that to the embedded transport network. Village Networks Inc. sort of fits, as do Atoga Systems and Network Equipment Technologies Inc. Companies such as Gotham Networks, Ellacoya Networks Inc., and Redback Networks Inc. (Nasdaq: RBAK) and Quarry Technologies Inc. are building edge service switches that allow operators to quickly provision and customize services for their retail customers. These switches have real appeal to service providers stuck in traditional models of provisioning limited-margin circuits. Why sell a T1 to a company when you could sell them a customizable “service portal”? Sounds good on paper right now, but getting service providers to adopt new business models is always a pain, so vendors have to be patient — right at a time when carriers are starting to act a bit more conservatively in the face of a slowing economy.

The IP Services Switch This camp includes CoSine Communications Inc.; Lucent, through its SpringTide acquisition; Nortel Networks Corp. (NYSE/Toronto: NT), through its Shasta subsidiary; and now Celox Networks. Like the core MPLS switch, these focus exclusively on IP and sit alongside the transport network. Their bread and butter is the VPN, and hopes are tied to a rapidly emerging VPN market in 2001 and 2002. So far, the outlook is good for these players. Their goal is clear and straightforward (they will never be called God boxes) and VPNs are catching on among large and medium-sized businesses.

It’s important to state here what these switches are not: edge routers. The edge router crowd continues to evolve the functionality of the router with MPLS and high-touch packet processing. This will continue to blur the boundary between an edge router and an edge switch, though services will always be a focus at the network edge. Cisco Systems Inc.’s (Nasdaq: CSCO) upcoming GSR 10000; and equipment from Laurel Networks; Amber Networks Inc.; Unisphere Solutions Inc. (Nasdaq: UNSP), via Redstone; Juniper Networks Inc. (Nasdaq: JNPR); Riverstone Networks; and Redback Networks Inc. (Nasdaq: RBAK), via Siara, represent players here. Where edge routers fit into the landscape of networks in 2001 is an ongoing debate, but they will often sit alongside service switches in complementary fashion, not in competition.

It will be interesting to watch which carriers take the plunge next year. Wholesale and retail carriers each have good reasons to give these a try, making service switches increasingly attractive to investors. RBOCs (regional Bell operating companies) need these to get out of the circuits business; carriers’ carriers can benefit from the “virtualization” of their fixed assets; and CLECs (competitive local exchange carriers) need anything that increases revenue streams in a tightening capital market. Fingers are crossed that MPLS delivers on its promise and doesn’t end up spoken of in the same dismissive tones that ATM often is today.

— Scott Clavenna, director of research, Light Reading http://www.lightreading.comNow that we’ve seen the highs and lows of 2000, here's a New Year's resolution for carriers in 2001: Start turning bandwidth into dollars. Or, more specifically, packets into dollars.

Optical networks have given national carriers tremendous capacity, but not yet a way to make big bucks from that capacity. Optical switching adds the ability to more flexibly manage lightpaths across a backbone and consolidate core network equipment, but it remains a method of switching circuits, from STS1s up to OC192s. This focus on the transport network — and thus the range of services supported by an optically switched network — remains limited to three or four “grades” of capacity leases. This allows carriers such as Williams Communications Group (NYSE: WCG) to improve the economics of wavelength services, but it does little to specifically address a burgeoning demand for enhanced IP service creation, and it does nothing to help carriers make money from such IP services.

Enter the “service-aware switch.” This isn’t an industry category just yet, but it will be by the onset of 2001. It’s a decidedly broad term but, broadly speaking, the service-aware switch is a hardware and software platform built to take service providers out of the commodity IP transport business into enhanced IP services.

These switches have substantial packet processing capabilities and can therefore perform a “deep read” of packets as they enter the switch. They perform the necessary classification, queuing, shaping, and policing of flows to give operators the opportunity to build enhanced IP services around these features. These switches often collect statistics on this processing and make it available to billing systems or for SLA (service-level agreement) verification and trend analysis.

With switches like these, carriers can, ideally, simplify their existing data networks by pushing service creation to the edge of the network — the metro POPs (points of presence) — and providing a common adaptation layer, typically MPLS, or multiprotocol label switching. (Adding MPLS to IP is being proffered as a way to replace pricey ATM as a switching and traffic management technology, and a host of vendors are lining up behind it -- startups and stalwarts alike.)

So what sort of enhanced services can these switches deliver? Anyone studying ATM over the past ten years will recognize them: multimedia, videoconferencing, circuit emulation (voice), secure VPNs (virtual private networks), bandwidth on demand, MAN/WAN internetworking, transparent LAN — all the applications ATM was supposed to support, but without the nasty “cell tax” or limited scaleability beyond OC12.

The service-aware switch market is already crowded and in need of segmentation. Like most other equipment categories, a rough division into core and edge systems is useful. In the case of the service-aware switch, each of those can be further segmented to pure IP and hybrids. Here are the four categories of service-aware switches, as announced thus far:

Core MPLS Switch Tenor Networks Inc. more or less invented this category. The core service switch is meant to act as an upgrade to large-core ATM switches (think Cascade GX550). These switches provide deterministic IP QOS, MPLS adaptation at the edge of the core network, and wire-speed switching of MPLS aggregated flows through the core of the network. In this camp companies like Tenor, Point Reyes Networks Inc., Vivace Networks, and Maple Optical Systems are leveraging MPLS to create switches that transform service provider transport networks into IP services networks. With these platforms, every service is transformed into IP/MPLS, including voice, making these switches quite similar to core ATM switches of old, just much more scaleable and IP-efficient. The challenge: These are damn hard to build; carriers aren’t convinced deterministic IP QOS is achievable; and many worry that MPLS will become as complex and difficult to scale as ATM. In other words, these companies are in a rather anxious moment in history: They’ll either pay off huge or fizzle. There isn’t much room in between.

Optical On-Ramp Kind of a hokey term, but this camp is distinct from the MPLS switch crowd in its use of multiple switch fabrics to accommodate a variety of services in their native formats. The 6500 from Tellabs Inc. (Nasdaq: TLAB; Frankfurt: BTLA) is one example; or the recent 1630 SX from Alcatel SA (NYSE: ALA: Paris: CGEP:PA). Add to this the forthcoming switch from Équipe Communications Corp., which is the next-gen core switch that Ascend meant to build but could not (everyone responsible for this program left for startups, including Équipe). These products are extensions of the “bandwidth managers” of recent years offered by Alcatel and Lucent Technologies Inc. (NYSE: LU). They terminate private lines and perform grooming and crossconnect of TDM (time-division multiplexing) circuits. They typically have an ATM fabric to manage data traffic, and now, in this iteration, have a means of migrating these services to MPLS. They reside at the same location as the core MPLS switch — they just don’t require that everything be converted to MPLS. This makes them more palatable to the large entrenched carriers that aren’t able to make wholesale changes to their network cores as quickly as some of the newer entrants or greenfield operators.

Service Edge Switch Instead of switch fabrics over 100 Gbit/s like the core switches above, these tend to have fabrics in the 5- to 20-Gbit/s range, and focus their attention on micro-flows (below DS3) rather than the macro-flows handled by core switches. Because these systems reside much closer to the end user, they are often responsible for true service creation, performing subscriber management functions. In this camp the variety of features is wide, making it difficult to position vendors neatly. Some include a transport and Layer 1 protection function, others leave that to the embedded transport network. Village Networks Inc. sort of fits, as do Atoga Systems and Network Equipment Technologies Inc. Companies such as Gotham Networks, Ellacoya Networks Inc., and Redback Networks Inc. (Nasdaq: RBAK) and Quarry Technologies Inc. are building edge service switches that allow operators to quickly provision and customize services for their retail customers. These switches have real appeal to service providers stuck in traditional models of provisioning limited-margin circuits. Why sell a T1 to a company when you could sell them a customizable “service portal”? Sounds good on paper right now, but getting service providers to adopt new business models is always a pain, so vendors have to be patient — right at a time when carriers are starting to act a bit more conservatively in the face of a slowing economy.

The IP Services Switch This camp includes CoSine Communications Inc.; Lucent, through its SpringTide acquisition; Nortel Networks Corp. (NYSE/Toronto: NT), through its Shasta subsidiary; and now Celox Networks. Like the core MPLS switch, these focus exclusively on IP and sit alongside the transport network. Their bread and butter is the VPN, and hopes are tied to a rapidly emerging VPN market in 2001 and 2002. So far, the outlook is good for these players. Their goal is clear and straightforward (they will never be called God boxes) and VPNs are catching on among large and medium-sized businesses.

It’s important to state here what these switches are not: edge routers. The edge router crowd continues to evolve the functionality of the router with MPLS and high-touch packet processing. This will continue to blur the boundary between an edge router and an edge switch, though services will always be a focus at the network edge. Cisco Systems Inc.’s (Nasdaq: CSCO) upcoming GSR 10000; and equipment from Laurel Networks; Amber Networks Inc.; Unisphere Solutions Inc. (Nasdaq: UNSP), via Redstone; Juniper Networks Inc. (Nasdaq: JNPR); Riverstone Networks; and Redback Networks Inc. (Nasdaq: RBAK), via Siara, represent players here. Where edge routers fit into the landscape of networks in 2001 is an ongoing debate, but they will often sit alongside service switches in complementary fashion, not in competition.

It will be interesting to watch which carriers take the plunge next year. Wholesale and retail carriers each have good reasons to give these a try, making service switches increasingly attractive to investors. RBOCs (regional Bell operating companies) need these to get out of the circuits business; carriers’ carriers can benefit from the “virtualization” of their fixed assets; and CLECs (competitive local exchange carriers) need anything that increases revenue streams in a tightening capital market. Fingers are crossed that MPLS delivers on its promise and doesn’t end up spoken of in the same dismissive tones that ATM often is today.

— Scott Clavenna, director of research, Light Reading http://www.lightreading.com