Optical/IP Networks

Ethernet to Outpace Sonet?

Over the past few years, Ethernet speeds have gradually caught up with Sonet. Now there's a possibility that 40-Gbit/s transmission technology might get deployed in Ethernet environments ahead of Sonet ones.

The reason is a simple one: cost. History has shown that telecom operators only start upgrading transmission networks when there’s a strong economic argument to do so. And there's reason to believe that 40-gig Ethernet gear will prove to be 35 to 40 percent less expensive than 40-gig Sonet gear, according to Menachem Abraham, president and CEO of Mintera Corp., a startup developing 40-gig subsystems.

The potential savings could be even larger, because telecom operators will be able to avoid using expensive Sonet interfaces in other equipment such as routers, Abraham adds.

The best way to understand this argument is to consider a typical arrangement with a router connected to a 40-gig transmission system, via a 10-gig connection. If this is 10-gig Ethernet, the transponders at either end of the connection (which convert optical signals into electrical ones and vice versa) are likely to cost 80 percent less than those of 10-gig Sonet connection, according to Abraham.

Moreover, the use of Ethernet avoids a lot of complicated synchronization and multiplexing of signals inside the 40-gig gear, says Abraham, who points out that this eliminates the need for a couple of expensive Sonet chips. All that’s needed instead is a relatively low-cost asynchronous multiplexer that combines flows from different routers without carrying them in separate channels.

To understand this more fully, take a look at Light Reading's latest report, 40-Gig Forecast, published today. A diagram on page 4 identifies the key elements in a thoroughbred 40-gig Sonet system. In an Ethernet environment, a much lower-cost receiver could be used, and the Sonet processing block and Sonet mux could be replaced with a single, simple G.709 async mux, as already noted.

Abraham says he can already see the day when routers will have 40-gig Ethernet interfaces that will connect directly with the 40-gig forward error correction chip in this diagram. “I believe that 40-gig Ethernet may get enough momentum that people in the IEEE [Institute of Electrical and Electronics Engineers Inc.] will want to ratify it as a standard,” he says. “The cost savings are tremendous.” — Peter Heywood, International Editor, Light Reading http://www.lightreading.com
padal 12/4/2012 | 8:29:28 PM
re: Ethernet to Outpace Sonet? and which are seeing any kind of pent up demand
ATMfrIP 12/4/2012 | 8:29:15 PM
re: Ethernet to Outpace Sonet? Hey I don't know if this post belongs here or on the research page but here goes...

A lot has been said about IP and how it is the future (and present...) With a whole bunch of existing switches being ATM or Frame Relay, plus a whole bunch of optical switches in the market. What is the performance of the network when pure IP traffic flows over them...

Typically IP packets are of varying size from a couple of bytes to a huge chunk. If you have it traversing the network designed to support some specific format/size etc... don't you think these switches are going to have a huge performance hit? For example, if you have 64 byte packets flowing over ATM, each IP packet is going to occupy 2 cells... and there you have a 50% hit in performance!! Of course, as the pkt size increases performance does improve...

Any input from the experts on this performance issue? Is the only feasible solution to yank out the existing h/w... and replace it with IP centric solution??
fiber_r_us 12/4/2012 | 8:29:13 PM
re: Ethernet to Outpace Sonet? Your ATM example is part of what most people refer to as the "cell tax". Indeed, this is one of the reasons people want to get rid of it. Most of the IP people claim there is little use for a complex layer 1/2 functions like FR, ATM, or SONET when you are running IP over it. Since most of the traffic already is (or soon will be) IP, IP becomes the driving force. Hence, there is a strong desire on the IP side to get rid of the unneccessary layers. Data centric carriers want to get rid of the layers not only due to efficiency, but also because it becomes much easier to operate. It all boils down to expenses that the carriers are experiencing due to all of the various network layers. Fundamentally, the carriers want a simple, cheap, easy-to-operate network infrastructure. This is the only way the carriers will survive in the long run.
DKP 12/4/2012 | 8:29:00 PM
re: Ethernet to Outpace Sonet?
> Until there is a way to have a "traffic
> contract" on a pure IP connection, ATM still
> doesn't have a real replacement.

Doesn't Yipes (for example) offer traffic contracts? There is no ATM in their network.

Belzebutt 12/4/2012 | 8:29:00 PM
re: Ethernet to Outpace Sonet? Until there is a way to have a "traffic contract" on a pure IP connection, ATM still doesn't have a real replacement.
doco 12/4/2012 | 8:28:59 PM
re: Ethernet to Outpace Sonet? > Until there is a way to have a "traffic contract" on a pure
> IP connection, ATM still doesn't have a real replacement.

There are IP based QOS mechanisms - most of which have learned from the mistakes of ATM and are much easier (and cheaper) to implement.

But - if you can throw bandwidth at the problem - that is often the cheapest way to ensure QOS.

ATM has lots of complicated (read expensive) ways of making sure that everyone gets exactly their fair share of the bandwidth in all situations. Real world has shown that this usually isn't necessary. If you provide good enough quality the applications take care of shortfalls by doing things like extra buffering.

For those who will scream that you can't provide "toll quality" voice or video service without quality guarentees - sit back and think about how crappy your typical digital cell phone call is. There are huge latencies, common drop outs, lots of static, etc. And people accept this. I would guess that customers will accept slightly less quality if they can get service for less cost. Take Wal-mart as another example of this.

ATM is on life support. If it wasn't for DSL having chosen ATM instead of Frame Relay, we would be planting flowers on it's grave this spring.
ATMfrIP 12/4/2012 | 8:28:57 PM
re: Ethernet to Outpace Sonet? Thanks to all for replying to my queries...

So what you guys are saying is that IP over anything else basically sucks... but we have to live with it because it is omnipresent...

And then there seems to be this difference in opinion regarding QoS over IP v/s ATM... isn't it a proven fact that ATM does do real QoS... the fact that bandwidth can compensate for IP's QoS shortcomings, isn't a great solution... 10-20 yrs down the road, what if that much bandwith is not enough... what then? provide more bandwith?? this doesn't look like a feasible thing??

what say?
harrouet 12/4/2012 | 8:28:57 PM
re: Ethernet to Outpace Sonet? 1- "traffic contract" is were carriers make money and can *differentiate* themselves. I agree that most of consumers and Internet users
don't need that.
2- Remember what SBC said with project pronto (ATM) : "we don't need one network optimized for data and one optimized for voice. We need a network that can take both". SBC is an investor in Atrica (Ethernet solutions for /carriers/), and know very well about GbE. Though they chose ATM.
3- Stop seeing Ethernet everywhere : it's only an access technology for now, and it's been sold by new entrants for one year at max.
4- See what serious carriers deploy : ATM on SONET. And they have good reasons for that, and they had hundreds of experts like you that helped make this choice. Yipes! (although I am not saying they are not serious) is financed by Juniper, maybe that's why they are all IP (?).

I like more a newly born standard called DTM. Until it's ready for deployment, think ATM, because IP is just DIY when you want to maintain a network.
Sign In