Traffic Engineering and Dumb Aggregation Pipes
February 4, 2009
Let’s talk about traffic engineering … I won’t start with Poisson tables–just some fundamental thoughts.
Any type of communications application, whether data, video, voice, or graphics, always originates and terminates in the local metropolitan and related access networks. You will not find origination or termination of such applications on an oceanic or long haul network. Those are dumb aggregation pipes that add little value.
Trying to keep things simple, optical fiber is lit with a laser. The distance a laser travels has much to do with the type of network, type of fiber, quality of fiber, and the number of service nodes. Lasers are expensive and regenerating their wave lengths after a certain distance is also an expensive proposition. It is much easier and cheaper to light an oceanic or long haul network, as they are strictly aggregation pipes of bits that get their signal regenerated every 60 to 100 miles. For oceanic or long-haul, the fiber is spliced over great distances–miles at a time–to create it’s physical footprint.
Splicing is an important concept to understand–it is taking two fiber strands, on two different cable sheaths, and uniting them as one. Every time you splice fiber (or cut copper as a matter of fact) you introduce a loss of carrying capacity to the signal (electrical or optical), as any splice will have a margin of error. Even the slightest margin of error adversely affects the signal amplitude. More splices in a given cable requires more regeneration over short distances, which is costly.
In industry terms, we call this a “loss budget”. When fiber is installed, the loss budget defines how much bandwidth you can reliably support over a pair of fiber strands as you splice or add nodes. Lots of buildings and homes means a more sensitive loss budget, as well as a need for additional available fiber inventory within the sheath as bandwidth demand grows.
What affects the loss budget of fiber? Location, uninterrupted signal distance and fiber count!
In comparison, for the thousands of buildings and homes in any market, the metropolitan fiber backbone and the access network are subject to hundreds–if not thousands–of splice points or nodes to interconnect users. Unlike an oceanic or long-haul network, it’s loss budget sensitivity is much higher. The physical costs associated with installing a metropolitan network are 10x higher than an oceanic or long-haul network. The add/drops of signals and multiple nodes in a metropolitan access network are complex and costly when compared to oceanic and long-haul networks.
From a traffic engineering perspective, for every mile of aggregated traffic in an oceanic or long-haul network, 10 miles of local metropolitan fiber backbone is required to feed it. For every mile of local metropolitan fiber optic backbone, you need 100 miles of access network to feed it.
The signal engineering in metropolitan settings is monumental and complex. In simple terms, long-haul/oceanic networks require less fiber than local networks to operate–they are simpler and cheaper to build and run. This is why an OC-X or GigaBit Ethernet circuit are cheaper on long-haul routes than metropolitan routes, which are complex and costly. OC-X or GigaBit circuits offer more redundancy and diversity than just two fiber strands between New York City and the beltway serving as a dumb aggregation pipe.
On Wednesday–coaching for Harry Homemaker, the Enterprise CIO, and Financiers…
Before We Piss Away Money–An Explanation of Networks
February 2, 2009
Did you know there is a factor of 10x more non-fiber sheath owners–non-owners of “infrastructure”–than there are of actual fiber sheath owners?
Did you know that the key to fiber route value is placement–unique routes, versus over-built, redundant, joint-built fiber routes?
Let me explain. If you have 5 carriers on the same fiber route, each with their own fiber sheath, what you have is a 40-year supply and demand imbalance and an inability to make a profit. Often this imbalance creates “distressed assets.” If you are from the beltway, and thinking about handing out our cash, read this paragraph a few times until you understand it. History has a way of repeating itself.
Just like UNE-Loops of copper, an IRU is not network ownership, it is a rented privilege. If your carrier has an IRU-based business model of a few fiber strands, you may need to ask some questions about future supply availability and/or overall capacity. Those who own the cable sheath control the supply and have unlimited capacity ability.
The costs to deploy fiber optic networks increase every year as labor is 60% of the cost!
What cost a $1 to build in 2000, now costs $2.00. Yes, distressed assets–they are distressed for a reason. The route…redundant cables…joint-built routes…fire-sold IRU’s to competitors, etc. These reasons gave credence to a massive fiber glut (circa 2001 to 2006), according to the real smart guys on Wall Street.
Since 2001, the Ma Bells of America have collectively spent over $25 billion on the deployment of new fiber optic networks. With such a supply glut as recent as 2001-2006, one would think there would be no need.
What is going on?
Owners of fiber optic cable infrastructure that allow IRU’s to be sold are actually providing open networks, which promotes real facility or network-based competition. However, our regulators don’t discriminate as they should between open and closed networks. The government should be bending over backwards for open-access network providers to promote substantive competition. More on this subject in a future blog.
Anyhow, there are different types of networks: oceanic, long haul, metropolitan and access. Let me describe each one in simple terms because I want to make sure we don’t piss away money needlessly:
Oceanic: Dropped from a barge to the ocean floor, this is a cable sheath wrapped inside of a steal sheath that connects continents or countries. Its capabilities are strictly transit and aggregation of point-to-point traffic. As many bits you can aggregate is what gets shoved down this pipe.
Long Haul or Regional Long Haul: This is the terrestrial version of Oceanic, but without the need for a barge or steal sheathing around the fiber cable. Long haul networks can be buried underground or strung from poles. This type of network terminates in what is called a carrier hotel in a given city or metropolitan statistical area (MSA). Long Haul capabilities are strictly transit and aggregation of point-to point traffic. As many bits you can aggregate is what gets shoved down this pipe. A national long haul network can run coast-to-coast, while a regional long haul network can cover a state or a region, such as the Northeast.
Metropolitan Network: This is the backbone network inside a city, suburb, town or village that enables connectivity between the access network (last-mile) and the long haul networks. These networks are not just bit stuffing aggregation networks.
Access Network: This is the last-mile or first-mile network that connects into the “edge” or metropolitan network backbone, which in turn connects to local data centers or the carrier hotel. Wireless last-mile broadband fantasies aside, the typical flavors of the access networks are copper, or–in less than 10% of situations–fiber optics. Copper has physical limits for moving bandwidth just like wireless silicon where fiber optics does not.
So what is my point?
Just follow the demand…do not build supply-side infrastructure with government tax dollars without the demand.
Next lesson…Traffic Engineering.
Feeds
