Verizon Mulls the Mesh

A chip startup is helping Verizon Communications Inc. (NYSE: VZ) revive the dream of building a mesh-based core that could replace ordinary Sonet rings.

The work surfaced in September, when researchers from Verizon and chip startup Parama Networks Inc. presented a paper at NFOEC. More recently, Parama's been making noise on its own: The company announced its technology earlier this month, and officials say a $7 million Series B funding round is on the way (see Parama Intros 'ADM on a Chip').

Verizon executives weren't available to comment, but Dean Casey, the company's former director of optical networking, says Verizon has eyed a mesh core for a couple of years.

Equipment vendors were keen on the idea, but none could build a mesh that met the 50-millisecond restoration provided by Sonet. "We stopped pursuing it for a while when we didn't get anybody to deliver us a product," says Casey, who left Verizon late last year to become an independent consultant.

Carriers have pondered meshed networks for years, hoping to overcome the inefficiencies of rings. In particular, bidirectional line-switched rings hold half their bandwidth in reserve in case of failures -- an inefficiency a mesh could overcome.

The mesh hasn't proven feasible yet, however. Besides the 50ms question, a mesh is more complicated to manage, a problem that potentially eclipses any cost savings. Rings also respond predictably after a failure, a quality that carriers cherish.

Part of the difficulty was that Verizon and others envisioned a mesh without Sonet, sending services natively across DWDM and offering protection at the wavelength level. Parama proposes to keep Sonet around but alter BLSR so that rings don't have to reserve protection bandwidth. It's similar to the philosophy preached in Resilient Packet Ring Technology (RPR), where more of a ring's bandwidth is freed up for use.

Parama's approach caught Verizon's eye. "It was all Sonet, which made them feel good," says Hemant Bheda, Parama CEO.

The companies' NFOEC paper proposes using rings in a different form, allowing them to overlap or share spans -- capabilities for which Parama has patents pending. The rings wouldn't reserve bandwidth for protection; instead, the network would respond to failures by creating new routes, cutting across multiple rings as necessary. Because paths can overlap and share spans, the new route could be built on top of active traffic.

Since there aren't any explicit protection paths, carriers could offer classes of protection -- full Sonet-like guarantees for a higher price, or best-effort recovery for a discount.

Verizon might have a chance to test-drive this kind of mesh soon. Parama got its first chips back from manufacturing nearly three months ago and says they're ready to sample.

Parama's goals for the chips go beyond the mesh network. The company aims to squash an add/drop multiplexer (ADM) into one line card, one with software-programmable ports that can be set for OC3, OC12, OC48, or Gigabit Ethernet. This would allow service providers to keep one generic line card for all those interfaces, reducing the cost of ADM systems (see Parama Intros 'ADM on a Chip').

— Craig Matsumoto, Senior Editor, Light Reading
parag11 12/5/2012 | 2:34:05 AM
re: Verizon Mulls the Mesh
Ciena has been doing sonet mesh for ages.

truelight 12/5/2012 | 2:34:03 AM
re: Verizon Mulls the Mesh Not with 50ms protection switching !

tavaar 12/5/2012 | 2:33:59 AM
re: Verizon Mulls the Mesh First, I've heard people define a mesh as almost anything under the sun. So one must be careful when using the term. An effective Mesh protection scheme is really a group of rings sharing protection across each other and espectially where they share physical links. But you always must have a physically diverse protection path as a cable cut is 10 times more likely than equipment failure. A Ring is inherently diversely routed. Each link within a Mesh must be as well.

Second, Just like common SONET/SDH ring protection schemes you need to differentiate between Path and Line based protection.

Third, Mesh protection is great assuming you are OK with full restoration following a SINGLE point of failure. A Mesh is inherently larger than a ring so therefore the risk goes up for multiple failures as you take advantage of spreading the protection bandwidth.

The fourth thing to consider for Mesh protection is that you must always have enough available protection bandwidth available in the Mesh to protect each path in your worst case link inside the Mesh. This is where network planning becomes a nightmare. But I still think this is very doable in a well defined restricted Mesh topology. I believe Brightlink Networks was working on something along these lines when they went under. Their internal HyperTorus Mesh switching fabric followed a very structured interconnection design as well.

So how does RPR compare to Mesh? From the "more efficient use of bandwidth" argument, I see it in a closer comparison to unprotected BLSR bandwidth (E-NUT or Enhanced Non-preemptable Unprotected Traffic GR-1230) which gets taken out first when a failure occurs on the working protected paths.

And how does integrating ADM functionality into a single chip (oops the transport overhead is in a separate chip) help solving the Mesh equation. This sounds like the E-NUT enhancement to BLSR.
particle_man 12/5/2012 | 2:33:39 AM
re: Verizon Mulls the Mesh Good comments.

I agree that mesh can mean many things. You can actually do a pretty interesting continuum from unprotected through 4F BLSR with various floavors of "mesh" in the middle where you trade off resiliency with cost.

The reason mesh schemes are cheaper is that they give you less protection. You can progress from there to a philisophical or business argument on how much protection is required for a given application. I think the whole thing dissolves into actuarial tables in the extreme.

I think the real benefit from a carrier point of view is to be able to offer different protection capabilities (service levels)at different price or cost points. This is a lot of work to implement though.
unet 12/5/2012 | 2:33:12 AM
re: Verizon Mulls the Mesh How does brightlink come in to this equation ?
any switch (independent of how it is realized) will fit this application.
Sign In