Report Stirs Ethernet vs Sonet Debate

Analysts agree that Ethernet has a long road ahead before it overtakes Sonet in metro area networks (MANs). But just what flavor of Ethernet will prevail, how long it will take before it becomes widely deployed, and where, is still being hotly debated.

Marian Stasney, senior analyst with The Yankee Group, published a report yesterday saying that, while service providers are certainly considering Ethernet as a low-cost option for delivering high-speed access in metro networks, actual deployments are still thin on the ground (Yankees See Rosy Gig-E Future).

“Eventually, Ethernet will take over in metro,” says Stasney. “But [that's] at least seven to 10 years down the road.”

Stasney cites several barriers to widespread deployment of Ethernet in the near term. Although Ethernet equipment is much cheaper to purchase and maintain than Sonet gear -- thus saving on initial capital expenditures and operational costs -- the fiber infrastructure needed to support Ethernet services in the metro is still far from ubiquitous.

Further, metro Ethernet applications like voice over IP and storage networking are still immature, making carriers less willing to adopt a pure Ethernet approach in the near future.

”Service providers trying voice over IP say that quality varies substantially between vendors’ gear,” says Stasney. “I’m not sure end users are comfortable with using pure Ethernet for storage yet, either. Even greenfield Ethernet players are being asked to provide services other than Ethernet to handle these services.”

There is also a tremendous amount of Sonet already installed in most carrier networks. Some estimates show that there are at least 100,000 Sonet rings operating in the U.S. and possibly twice that number of SDH rings operating outside the U.S., according to Michael Howard, co-founder and principal analyst with Infonetics Research Inc.

Both Stasney and Howard agree that cutbacks in capital spending mean service providers won't lay massive quantities of fiber, or rip out older Sonet infrastructure, to deploy a technology that may not be able to generate revenue right away.

No hard numbers from the report have been released to the public yet, but Stasney says she is basing her research on conversations with 30 to 40 Ethernet, Sonet, resilient packet ring (RPR), and wave division multiplexing (WDM) equipment vendors, as well as data that was collected for another Yankee Group report on service providers, which hasn’t yet been published.

Other analysts agree with her basic premise. Chris Nicoll, vice president at Current Analysis, and Mark Lutkowitz, VP of optical networking research at Communications Industry Researchers Inc., both say they agree that Sonet will continue to dominate the metro for at least the next three to five years.

“This is probably the first time that I can say that I agree with Yankee,” says Lutkowitz. “The rate at which Sonet gear becomes obsolete is very slow. That stuff will be around for another 30 years. There’s no question that Sonet is here to stay.”

But Lutkowitz seems a bit leery of the specifics of The Yankee Group report: “Projecting the market seven to 10 years is ridiculous."

Fellow analysts may also quibble with Stasney on another point. She is bullish on the prospect of pure Ethernet eventually becoming the technology du jour in the metro, while others say that some combination of Ethernet over Sonet, or even the emerging packet technologies like RPR, will rule.

“It’s hard to say which technology will win out,” says Stasney. “RPR will likely be extremely important, but with Ethernet over Sonet you lose benefits of both technologies.” Analysts who disagree point to specific examples to back up their case.

Ocular Networks Inc. and Native Networks Ltd., which both offer Ethernet over Sonet, are getting serious attention from service providers,” says Nicoll from Current Analysis. “But companies like Atrica Inc. and Extreme Networks Inc., pure Ethernet guys, are finding legacy carriers a harder nut to crack. There aren’t that many Cogent Communications Inc. or Yipes Communications Inc.-type providers out there for them.”

- Marguerite Reardon, Senior Editor, Light Reading

<<   <   Page 2 / 3   >   >>
lightmaster 12/4/2012 | 7:55:34 PM
re: Report Stirs Ethernet vs Sonet Debate 50 ms restoration for packet rings is the easy part with respect to voice. The ability to carry synchronous traffic is the harder part. Every voice call has to be serviced sychronously (in the case of current TDM, every 125 microseconds), which means that you can't just have the normal free-for-all in an ethernet network.

All of the standards are addressing the transport part, but there are equal issues in the switching elements for voice traffic. SONET elements (ADMs, Cross Connects) have delay in the hundreds of microsecond range, while ethernet switches are in the microseconds. If this isn't fixed, echo cancellation becomes a huge issue. And then there's jitter and jitter buffering issues...

The standards are working on all of this, but it will take time. Standards also have to be developed for circuit emulation for T1s so you can develop equipment that carrys PBX traffic for the MILLIONs of customers who don't want to buy new VoIP equipment to suit the service provider.

Once all that is done, you have the carrier trials to proove in the new technology (about a year), then a rollout over time. The bottom line is that a 5 to 10 year adoption curve for packet rings in large carrier networks, i.e ones that carry voice and make money (not Yipes) would actually be a pretty good success story, equivalent to the success of SONET.

Everyone wants the world to change overnight...it doesn't.

mrs5c 12/4/2012 | 7:55:26 PM
re: Report Stirs Ethernet vs Sonet Debate Does anyone know the typical $/MB pricing that cogent, telseon, etc. are charging? Typical RBOC tariff rates for Sonet local loop services are between $10-20/MB depending on the speed plus whatever $ they charge for Inet access. This typically puts the rate upwards of $300-$400/MB. Are all the metro CAPs charging Cogent type rates $10/MB for local + internet access? I could see the trend for Ethernet CAP moving faster if the price points are this low.
tintin 12/4/2012 | 7:55:25 PM
re: Report Stirs Ethernet vs Sonet Debate FiberFan,

"Ethernet vs. SONET. From all of my reading the answer to this question seems to be RPR."

RPR/DPT (aka IEEE 802.17) may make sense for a ring topology where most nodes are peers and traffic patterns are not predominantly star-like.
But it doesn't make too much sense for hub-and-spoke topologies (which is typical for Metro Access Rings), because it raises the cost of the spokes significantly (and unnecessarily).
The primary reason for this is the overwhelming cost of OEO components needed for RPR interfaces.

jmd 12/4/2012 | 7:55:23 PM
re: Report Stirs Ethernet vs Sonet Debate I've yet to see the business case completely laid out that justifies the equipment purchases required by the metro providers.
mma 12/4/2012 | 7:55:17 PM
re: Report Stirs Ethernet vs Sonet Debate "50ms restoration is key to support legacy TDM/voice networks. This kept calls (hundreds to thousands of them) from dropping and causing the Class 4/5 switch to melt down. There are probably a couple of other reasons, but maintaining calls (or Tandem Links) was the main reason."

My recollection is that hit timing on analog switch interfaces is 200 ms and that on digital switch interfaces the signaling should be frozen after an out of frame for two seconds until a carrier group alarm is declared to disconnect the connections. Therefore, I don't think that the 50 ms requirement derives from switched network trunking.

More likely it is due to either signaling oddities with operator services, coin services or voiceband special services, but it might be due to the effects of noise bursts or dropouts on analog modem behaviour. Do the services and signaling types which drove the requirement still exist, at least in sufficient number to justify the 50 ms requirement?
flanker 12/4/2012 | 7:55:12 PM
re: Report Stirs Ethernet vs Sonet Debate ...The ability to carry synchronous traffic is the harder part. Every voice call has to be serviced sychronously (in the case of current TDM, every 125 microseconds), which means that you can't just have the normal free-for-all in an ethernet network...

lightmaster 12/4/2012 | 7:54:59 PM
re: Report Stirs Ethernet vs Sonet Debate The debate about the need for 50ms has been held many times before we got to packet rings. We had it in ATM and we had it in the DWDM realm. The bottom line is that many carriers differentiated themselves when they introduced SONET using 50ms resortation as a benchmark of reliability, and sold it to their business users as an obsolute neccesity. Anyone who backs off on it gives their competitors sales team ammunition to label the other providers network as less reliable.

It may be a dumb reason for holding onto a requirement, but that doesn't mean it's not real. Sorry to break the news, but engineers don't run the world.

This point, however, is moot. 50ms resortation is the easy part and will be adopted into the RPR standard. The ability to carry synchronous voice traffic is much more difficult, and much more important to the adoption of RPR as a feasible solution for larger carriers who get most of their revenue from voice. This is going to take YEARS to standardize and more years to sell.
jshuler 12/4/2012 | 7:54:57 PM
re: Report Stirs Ethernet vs Sonet Debate You must not be a customer. Or if you are, you aren't talking to the right vendors...
jshuler 12/4/2012 | 7:54:57 PM
re: Report Stirs Ethernet vs Sonet Debate RPR has significant advantages even in hub-and-spoke networks, due to its use of both sides of the ring for both protected and non-protected traffic, and its ability to stat mux traffic into the entire bandwidth. Some RPR implementations (Luminous) can even provide the TDM circuits within that stat-muxed bandwidth, which means no pre-allocation of circuit vs. data bandwidth (STS-1s or VTs) and no wasted, partially full pipes. And yes, those circuits are stratum timed. In addition, data bandwidth can be oversubscribed, vastly multiplying the revenue per unit cost on the ring. So your arguments are red herrings.
optomyth 12/4/2012 | 7:54:57 PM
re: Report Stirs Ethernet vs Sonet Debate Any predictions or time frame on the potential for an all optical core competing with the new sales of both SONET and Ethernet pushing them to the edge?

Of course tremendous technology achievements must be made and the inevitable standards addressing protection switching among other things must be developed, but new optical layer 1 products may displace SONET and Ethernet at the core, hopefully, in the not too distant future.

- Opto
<<   <   Page 2 / 3   >   >>
Sign In