Seven,

Read my post again..you are preaching to the choir here :)

-Amit

So, if the carriers and the vendors decided to completely screw their customers...which is what you actually wrote.  There are 2 reasons for legacy:

1 - Customers

2 - New technologies don't do everything that the old ones did and the old ones still have customers.

Really think that Phone Companies want to have Pay Phones circuits or 4 wire leased line or dry alarm contact circuits?  And yet new ones are sold and installed (in very small volume) every year.

So, you best get over this idea of eliminating legacy.  Can't happen.  Frame Relay works as well now as it did in 1990.  If that's all somebody needs, they feel no obligation to change.

seven

Seven, If the operators and vendors had agreement to remove all legacy, then what PictureThis said is great. Now, that doesn't happen in real world. To maintain legacy, it is insane amount of overheads and multi-level encapsulations resulting in inefficiency. The solution to this is not introducing another encapsulation introducing OTN, but trying to eliminate layers.

If MPLS-TP can do it, or we could have the desired "carrier" objectives captured in MPLS, that might be a better way out.

".operators should harmonize and streamline their service offerings in order to accommodate the most cost-efficient architectural models AS WELL AS move towards enabling OTN technologies."

This is potentially the funniest thing I have ever read on these boards.  Seriously...operators changing what they want to do because their vendors might have to spend extra money to do it.  What planet have you been living on for the last 10 years.

seven

Completely agree with PICturethis. The only value add OTN could have brought is multi-vendor DWDM line side interop and so, just a digital wrapper. Like TMN, this is a far fetched dream as of today.

IMO, there seems to be little justification to develop a completely new switching granularity. This burdens with supporting both SONET/SDH and OTN and doesn't add any economic value.

You mean Intune? No, I've always thought of them as being completely separate from OTN; they've got their own way of separating and grooming traffic. Plus, they're more for metro rings than for the core, aren't they?

And are these guys doing something in the OTN / ODUflex / G.HAO domain:

http://www.lightreading.com/document.asp?doc_id=208181

Any commercial success?

There are tons of valid technical points listed below.  However, I see absolutely no valid economic arguments.  If its not profitable for operators, it's not practical.  THIS is why large scale OTN deployments connecting several disjointed domains-regions-technologies have not been successful or practical to this point.

Instead of developing OTN in order to meet the needs of every data rate for every customer in every network scenario imaginable...operators should harmonize and streamline their service offerings in order to accommodate the most cost-efficient architectural models AS WELL AS move towards enabling OTN technologies.

Only with this two-phase parallel approach can operators (especially large tier 1s) hope to compete with smaller, more agile and modern market competitors with aggressive prices points and higher margins.

While I enjoy a technology rant as much as anyone...its not just about the bites and bits, rather the dollars and cents (or euros if thats your thing;)

Any actual implementations (GA products / in-service networks) for "ODUflex with hitless adjustment"?

And is this different than SDH VCAT-LCAS (other than capacities)?

My replies are inline:

-Amit

1. the ability to scale is one big advantage. OTN is buit for wavelength-level capacaties, and at 100G there is no sonet/sdh standard at all. it is OTN.

[Amit] True. True dream of OTN was line side multi-vendor interop but that hasn't been successful. And I'm wondering why do we lack 100G SONET/SDH standard? More than anything technical, my feeling is that this was not considered as there was so little market traction on STM-256.

2. the ODU0 granularity was built to match 1 GigE traffic, making OTN a fit for 1G, 10G, 100G, and beyond. 1G is a big deal because alot of aggregation networks are being based on 1Gig pipes.

[Amit]Yes. Isn't that inefficient though? The SONET/SDH pipes rarely need to carry full wire speed 1xGE. It is more often a fraction (atleast in the SDH market). Also full GE can be carried in VC4-8v and switched at that granularity as well.

3. ODUflex allows the size of the OTN pipe to be resized to closely match the capacity of whatever data rate is being used. And, with ODUflex with hitless adjustment, the re-sizing will be able to be done dynamically (on the fly) - making OTN look alot like Ethernet. 

[Amit] ODU flex is nothing but SONET/SDH VCAT + LCAS which enable same capability - hitless addition/removal of VC/VT in a VC Group. For bigger granularity, I'm wondering if there could have been a combination of contiguous & virtual concatenation. For example, to carry 4G worth bandwidth, a VC416c x 2v seems much simpler if there was standards direction towards it. Wouldn't this have been much better than evolving a completely different OTN switching? 

4. Synchronous vs. Asynchronous: this is a fundatmental difference between sonet and OTN, even though they are both TDM. I'm not sure that this directly contributes to efficiency in packets but it does appear to have an impact on cost.

[Amit]Note that many of OTN switching equipments also are supporting SONET/SDH switching. So there can't be any cost advantage for omitting supporting synchronization here. On the other hand, it can only lead to more complex ASICs and engg effort thereby increasing cost rather than reducing it. Is the real cost advantage a result of Moore's law or really Synch vs Async?