Storm & 360 NWs use SN16K switches. 360 will increase from 16 to 120 SN16K switches by '01 end
Emerging Technology: SONET Rings Into Mesh -- The Future Of Next-Gen Networks, Part II
It's the Services, Stupid
So what are these new, mesh-enabled services? Glad you asked. After looking around for major mesh players, we could only find two who really tout their mesh architecture with live (though sometimes beta) traffic: 360networks and Storm Telecommunications. Both use Sycamore's SN 16000 switches as the main source of optical switching intelligence. Astral Point has won three customers: Time Warner Telecom, Lighthouse Communications, and Advanced Telcom Group-all American integrated communication providers using the ON 5000 for some degree of mesh functionality. (Further details weren't available at press time.) Let's look at Storm first.
Storm is a London-based international wholesale and retail carrier with an optically switched network of roughly 9,500 kilometers currently in operation, with 20 major international gateways across Europe, New York, and Washington, D.C. Optical services include:
A Synchronous Digital Hierarchy (SDH), a SONET look-alike service with 50ms restoration (the network recognizes it as a traditional SDH/SONET link).
Flexible Pipes that allow customers to increase and decrease their bandwidth on command in increments of 1Mbit/sec, from STS1 to STM16. Currently, this takes two hours and is done through a customer service call center; plans to automate the system are scheduled for first quarter 2001.
A Flexible Network that allows customers to set up bandwidth links around the network on demand within a predefined geography-similar to turning bandwidth on specific paths on and off.
International Gigabit Ethernet connections. Storm reports advantages in its architecture that include lower cost on like-for-like networks and greater visibility and control for customers who can access real-time availability, severely errored seconds, alarm status, and network map information. Other advantages are improved infrastructure utilization, simpler management (done in software), and lower latency and jitter performance. Storm has announced the BBC and AboveNet, an Internet connectivity and colocation provider, as customers.
360networks has caused a bit more of a stir, with more press coverage, conference presentations, and high-profile partnerships. The company currently has 16 Sycamore switches deployed in North America and Europe with beta customers, and it plans to grow to 120 switches worldwide by the end of 2001. The carrier plans to differentiate itself on the ability to provision and manage restoration paths, either dynamically or ahead of time, supported by dynamic provisioning capability from the core to the edge. This approach differs from most other carriers, which exploit mesh in their cores but still rely on SONET at the edge. 360networks will focus on speeds ranging from OC-3 to OC-192 and Gigabit Ethernet, with three classes of protection service:
Unprotected, with no restoration, for data networks where the customer wants to use higher-layer (layer 2 and/or layer 3) restoration schemes, such as routing protocols like OSPF or Intermediate System to Intermediate System (ISIS). Target customers are ISPs running non-real-time interactive data.
Mesh-protected service, with SONET ring-like availability and subsecond restoration, within a continent. This is aimed at customers running non-real-time data (most Internet traffic) who want layer-1 protection, since most router configurations don't notice link failures if the link restores in under one second.
One to one protected service, again with SONET ring-like availability and 50ms restoration, targeted at customers running real-time interactive traffic, such as voice or two-way video.
Gill maintains that dynamic provisioning delays can be brought down to two weeks. It might take only a few minutes of pointing and clicking to configure the circuit, but extra time is still needed to complete the service-provisioning process-taking into account potential physical assignments for the customer, such as optical patch panel position, space, and power, along with setting up service billing and customer records in the operational support system.
So What's the Problem?
On the surface, going to a mesh seems like the only choice for providers, but many carriers are resistant, at least for now. Customers seem not to care. According to Storm, as long as its service acts transparently like good ol' SONET, it's happy.
“Mesh is ready for prime time, but most service providers probably aren't,” says Nicoll. “Incumbent service providers have a ring or SONET legacy that is incredibly hard to move away from. Everything has to change: network equipment, provisioning systems, billing, operations, and training. That's a lot to bite off. Some providers are biting the bullet, but eventually, most will [switch].”
Carrier skepticism is rooted in both tradition and technology, says Clavenna. “Evolving a ring network to a mesh requires a complete change of mindset, and many carriers haven't yet reached the comfort level required to alter the fundamentals of their network architecture. Meshes are also difficult to manage and may be prone to outages from software glitches. Rings have the clear benefit of simplicity, which lets carriers sleep at night.” For this reason, look for greenfield carriers to go mesh first. Because large Interexchange Carriers (IXCs) run so many different business services from their networks, most of which use SONET transport, these carriers will likely be the last to go.
Interoperability is also a concern, though Storm plays this down somewhat. Running management in software means you can tweak the system to accommodate other vendors. But for now, Storm is only using Sycamore gear for optical intelligence. Standards would help: While carriers don't necessarily have to wait for final ratification, such as in the case of Multiprotocol Label Switching (MPLS), some engineers hope MPLS, which labels different wavelengths for managing and steering traffic through optical networks, will ease interoperability problems.
In the end, carriers will more likely opt for a migration path that takes them from a predominantly SONET network to a hybrid ring/mesh infrastructure. Sycamore and Astral Point are vocal proponents of the approach, and given the need to leverage the installed base of SONET rings, this seems to make sense. “Entire network topologies cannot be upgraded overnight,” says Wang. “So you will see many networks that combine both ring and mesh topologies, until they can migrate to an all-mesh topology. Some providers talk about moving to a mesh first at the core of the network, and leaving the access portion as SONET rings.”
That migration will likely take the form of using mesh switching to link independent legacy ring networks together. This leaves the physical plant intact (no forklifts!) and allows carriers to incrementally upgrade their switching as desired. Or, they can reverse the process and configure virtual SONET/SDH rings on top of the physical mesh architecture, thus providing SONET-like restoration along with mesh-style provisioning advantages, while still saving on money and bandwidth by avoiding the SONET ring-style 1-to-1 protection scheme.
The Next Move
It's hard to get a grip on just how fast mesh mania will sweep providers en masse, if ever. Clearly, most carriers are in a wait-and-see mode, and they'll be watching early deployers and adopters carefully. Most analysts believe in a mesh's benefits but remain unsure they're enough to force incumbent providers' hands. Because meshes don't provide any profitable, incremental services pitched at the end user, at least one provider interviewed for this article felt mesh wasn't worth the effort. If this is correct, mesh networks might become the exception, not the rule.
In the meantime, it's worth looking at the greenfield carriers, especially in the Gigabit Ethernet space. Competition should provide some good deals in terms of price-per-bit and tiered services. Start-ups will be extra-conscious of competing on flexible offerings, so the end user can customize usage and billing packages more easily with the greenfield carriers than with more established players.
The key will be the operational support system. Without first-class customer management capabilities-including order entry, service activation, and billing-the provisioning and tiered protection services that are mesh's greatest strength become useless. If you're not comfortable going with an unproven provider, look for telcos who are comfortable with a combined ring/mesh approach. But don't worry- right now, you won't find too many of these hybrid options to choose from.
networkmagazine.com |
