Wizard and TD have hit on a couple of the major themes in the bandwidth space.
1. Despite the overwhelming success of todays' Gbp routers (the 12000 from CSCO in particular), they are falling behind the ever-increasing bandwidth processing requirements demanded by service providers. The 12000 provides only 60 Gbps of switch capacity and I believe the Juniper M40 only has 40 Gbps (with strictly OC-48 interfaces). As the DWDM pipes provide incredible raw bandwidth, it will indeed be a challenge for the existing players and the startups to keep pace. The big backbone providers will be needing Tbps of throughput and hundreds (not a dozen or so) of OC-48 or 192 connections. If Nexabit's claims are true, its scale is unbelievable (up to 6400 Gbps of capacity)
So what happens is that as DWDM scales, the existing routers cannot handle the bandwidth. So additional layers of SONET equipment (usually Add/Drop Muxes) must be deployed to scale the bandwidth down. This costs a lot of money in terms of hardware and management costs. Companies like Sycamore are integrating these SONET features into a box which starts to bring the logic that Wizard was talking about. In their 6000 and 8000 products, DWDM is incorporated but is a relatively minor player - the real value add is the intelligence.
2. From a metro standpoint, a couple of somewhat related metrics are very important to understand. Traditionally, the cost of telcos core ran roughly 80% for fiber, oeo regenerators, and amplifiers and only 20% in the CO for ADMs and DCCs. This spurred DWDM development as the major cost, regenerators, were now replaced by EDFA's. This dramatically changed the cost structure. To amplify 16 channels, the cost went from $800,000 (a 3R regenerator for each channel) to $50,000 for one EDFA! That is huge! Since the trunk costs are now coming down, the node/CO costs (ADMs, DWDM, DCCs) are now increasing in relation to the whole cost. Thus the ratio is changing to 80% in the CO and 20% in the trunks. How is this relevant to the metro? Well the same cost structure exists in the metro as is now evolving in the core. Since the trunks are rather short, much less regenerators are used, and more emphasis is placed on CO intelligence. The ability to switch, provision, restore, maintain, monitor etc. is THEE main priority of the metro area. So even though more bandwidth is needed, the trunks (and thus mega bandwidth) are not the problem. One can argue that the CO is more important than overall bandwidth capacity since (and here is the current problem for metro DWDM) it does no good to have all that bandwidth if it costs you millions in SONET gear to add intelligence. The economics just aren't there.
Next gen systems will incorporate this SONET intelligence (indeed Supercomm provided many startups that claim they are already there) into the DWDM terminals (or visa versa). Once these solutions provide more than just fiber gain, and start to provide service capabilities, then metro will take off. I'd say we're a good two years away.
Bulldozer |
