You're right, the major cost isn't the bandwidth (at least the in the U.S.), but the current technologies (cable modem, DSL, and wireless) are
asymmetric, and high upstreams kill the performance of the first and
Your points about the marketing and usage value of higher asymmetric is right on. Not only are the higher numbers attractive, they generally reflect a residential subscriber's usage pattern (there are some on this listserv who have pointed out that those with very high symmetrical speeds, 100 Mbps, for example, do have higher upstream, but I think that's because they are more attractive P2P nodes) and so residential broadband networks have been designed for asymmetric service. One of the reasons that business broadband is more expensive is that they not only use their 'pipe' more heavily than a typical user provisioned with the same speeds (i.e. bandwidth costs are more), they also prefer a symmetrical connection for their e-mail server and web traffic which requires different (lower volume and more expensive) equipment and/or they consume more of that shared upstream link. BPON/GPON is also asymmetric, as you point out, but because the marketed highest-end speeds are a fraction of the standards' capabilities, the asymmetry and potential oversubscription are easily overlooked. This works to Verizon FiOS' advantage while marketing its symmetrical plans. I personally prefer Active Ethernet-based fiber solutions for the reasons you allude to -- they more closely match enterprise network architectures (that's why we see Cisco in this space (i.e. Amsterdam's fiber network) and so networks of this type can leverage that equipment, volumes, and pricing): symmetrical in speed and switched. The challenge with the Active Ethernet architecture is that most often active electronics need to be placed in the field, while many PON solutions can use passive optical splitters. Frank -----Original Message----- From: Sean Donelan [mailto:sean@donelan.com] Sent: Monday, January 21, 2008 4:47 PM To: Frank Bulk Cc: nanog@merit.edu Subject: RE: An Attempt at Economically Rational Pricing: Time Warner Trial On Mon, 21 Jan 2008, Frank Bulk wrote: thoroughly third. There are symmetric versions for all of those. But ever since the dialup days (e.g. 56Kbps modems had slower reverse direction) consumers have shown a preference for a bigger number on the box, even if it meant giving up bandwidth in the one direction. For example, how many people want SDSL at 1.5Mbps symmetric versus ADSL at 6Mbps/768Kbps. The advertisment with the bigger number wins the consumer. I expect the same thing would happen with 100Mbps symmetric versus 400Mbps/75Mbps asymmetric. Consumers would choose 400Mbps over 100Mbps.
Long-term, fiber avoids the upstream performance issues.
Asymmetric fiber technologiges exists too, and like other technologies gives you much more bandwidth than symmetric fiber (in one direction). The problem for wireless and cable (and probably PON) is using shared access bandwidth. Sharing the access bandwidth lets you advertise much bigger numbers than using dedicated access bandwidth; as long as everyone doesn't use it. The advantage of dedicated access technologies like active fiber (or old fashion T-1, T-3) is your neighbor's bad antics don't affect your bandwidth. Remember the good old days of thicknet Ethernet and what happened when a single transceiver went crazy, the 10Mbps ethernet coax slowed to a crawl for everything connected to it. The token ring folks may have been technically correct, but they lost that battle. There was a reason why IT people replaced shared thicknet/thinnet coax Ethernet with dedicated 10Base-T pairs and switches replaced hubs.