This is from my ip2k.com wordpress that I forgot the password to and deleted.  For a while, I was #1 on google for ‘connection aggregation’ but not anymore.

This started off as an idea in the back of my mind and eventually turned into a few pages of text, then into a whitepaper used for a final project in my networking class (recieved A+ too)

first page is removed because you don’t need to know my name:

ccap.pdf

The Community Connection Aggregation Project

IntroductionCCAP is a logical CAN (community area network) topology intended for increasing internet connection speeds among communities by a factor of n at little or no additional cost.

Overview

To provide an internet connection with a theoretical speed limited only by the number of participants for little or no additional cost to the end user by utilizing a central “hub” to aggregate connections incoming from diverse consumer-level connections into a ultrafast connection to be redistributed among all participants. Since internet traffic is rarely sustained streaming, this would easily facilitate a 100mbit connection for the same monthly price as a 6mb connection using mostly inexpensive common consumer level networking hardware wherever possible.

Fig 1.1

In fig 1.1 you can get a very basic understanding of how CCAP works. By aggregating or combining the connections, it is possible for 3 users with 5mb/1mb to 1 user at a time. Likewise, if house A is streaming a video using 2mb/.5mb, there is still 13mb/2.5mb left in the pool for anyone to use. If user B tries to load a webpage, it will still be over twice as fast as it would not utilizing CCAP.

Theory

Since internet traffic is rarely a sustained high-speed stream (with exceptions such as large P2P file transfers, gaming, streaming movies, etc), we can combine multiple users’ bandwidth “headroom” (free bandwidth above and beyond what they are using) and allow everyone involved to utilize it.

Incidentally, this also creates a somewhat fault-tolerant network, assuming in the rare likelihood that each node would be on a different physical network segment. Nevertheless, it does provide a certain level of fault tolerance against hardware and “pole-to-house” cable failure, as this would result merely in a decrease in bandwidth proportionate logarithmically to n connections.

Hardware

Ideally one would already posses all the required hardware as it is fairly common nowadays with commodity network cards and computers. The network cards do not have to be homogeneous although such a situation may facilitate easier setup. For example, I have a system with 4 NIC cards, all PCI and all the same model 3com. This allows the use of enumerated XLn interface names, all using the same drivers.

SETUP:

xl0: From House A

xl1: From House B

xl2: From House C

xl3: To Switch back to houses

Since we only need one outgoing connection to a connectivity device, only one card is used as output. If this were a faster setup (say, 10xFIoS 20mb/20mb connections), we would use 100baseTX cards for input and a 1000BaseTX card for output, as theoretical maximum speed would be 200mb/200mb – overhead.

Wireless Option

There also exists in theory the option to use wireless to facilitate easier linking of all required hardware. The most obvious advantage to this is the omission of outdoor wiring, in most cases completely. Wireless access points can be configured to broadcast the signal to wireless NICs in the connection aggregation device, which could output via wired NIC feeding access point to which all nodes would then have access; that would look something like this:

fig 1.2

Obviously this is the most desirable configuration, but it does require a bit more expensive hardware and speeds are limited at the moment to common 802.11 Draft-N which tops out at 108mb. More commonly used 802.11G standard specifies 54mb which could easily be exceeded using this topology, especially with upcoming advances in speed such as FiOS and Uverse, even fast cable modem (8mb). Although you may aggregate enough connections to make nessecary the use of a Draft-N return AP, incoming APs only need to be as fast as the individual connection, in most cases 802.11b would suffice but G would be the preferred method for a more modern connection. This also means that xl0-2 may be “b” or “g” cards instead of the more expensive “draft-n” variety. In fact, xl3 would preferably be a wired interface feeding the WAN port of a draft-n access point.

In such a system if range is a problem as it likely would be with a larger community it is important to note that external antennas and/or amplifiers may be nessecery to attain maximum speeds. Also worth noting is the deployment of such a system in a densely populated area as WiFi channels may already be in use and kill your signal strength or even override your signal. 802.11b/g allows only channels 1-11 to be used in america so there is not a ton of choice if you wish to set up a large network.

Software

The use of freely available software is emphasized within CCAP because of it’s community driven, open source, experience enhancing, low-to-no-cost spirit and DIY ethic. An example of a suitable software package that could be utilized in this manner is pfSense, a freely available m0n0wall-based distribution built upon FreeBSD with added functionality such as the use of plugins and more importantly, connection aggregation, load balancing, seamless failover, and advanced routing.

Note: This concept has not been tested by myself or anyone that I know of, if you try this out and it works / doesn’t work, please let me know in the comments how it went. Thanks.