Metro Waves Plus Ethernet = Lower Cost
It goes without saying that every participant in any value chain wants lower cost inputs while maintaining higher price outputs. That essentially means that at the very beginning of any value chain, there typically exist products and services some of which are, or can be likened to, commodities. Electrons are an input at every stage and would fit most notions of what a “commodity” is.
So are semiconductors, housings, power supplies, resistors, screens, keyboards, fans, heat sinks, connectors, cables and ties. And though many will blanch at the thought, stable, profitable and important businesses are built on commodities.
Still, it also is important to remember that even commodities are differentiated and segmented. Buyers have choices at every stage of every value chain, for every input, no matter how “commoditized” an input is.
In fact, to the extent that Moore’s Law has a business corollary, it is that over time, every technological product gets smarter at the same price point. One might argue that Level 3 Communications has such thinking in mind in arguing the cost of content delivery services ultimately will be equivalent to the cost of simple IP transit.
There also are instances where technology is simplified to gain a cost advantage. Moving bits at layer two (Ethernet) or layer one (waves over optical fiber) instead of at layer three (managed IP) is one such example.
Equinix, for example, uses metro dark fiber and multiple 10 Gigabit-per-second wavelengths to virtualize data center capacity across a metro area. IBXLink allows enterprise customers to place servers in separate data centers but have it appear they are locally connected.
Diversity might be one customer driver, but there are provider drivers as well. Say Equinix reaches the capacity limit of a certain router, on a certain rack, at a certain location, and the next unit of capacity cannot be easily added to a rack or a cage. In that case, a viable option is to remotely install the next router at a different location and then logically connect the additional capacity to the existing customer network.
And of course, the solution only works when optical transport costs are minimized, says Lane Patterson, Equinix chief technologist.
One way to do that is to create a wave division multiplexing optical transport network by putting long haul optical transponders right on a router (using Infinera, for example) and using passive WDM filters on the premises and optical amplifiers in the outside plant to create a dedicated wave network.
Patterson says the savings can amount to as much as 60 percent compared to the cost of buying separate transport equipment. Of course there is a trade off. The approach isn’t as manageable as some other options and requires all failure protection to be done in the router.
And though it might seem heretical, another option is to dispense with quality of service mechanisms when 90 percent of the traffic is comprised of best effort bits. The engineering choice in that case is to use best-effort layer two Ethernet switches instead of higher-performance routers with prioritization mechanisms, saving as much as 90 percent on at least some parts of the capital spending front.
A third option is to combine some elements of WDM enabled by plug-in optics and switches such as the Force 10 to capture Ethernet switch cost savings. IP
