X-Raying the Internet Backbone

        By Aaron Price, UltraNet Webmaster ( aprice@ultranet.com)

        This article is reprinted, with permission and minor editing, from Odyssey, an on-line publication of UltraNet Communications, Inc. The article can be found at http://www.ultranet.com/odyssey/archive/110896/feature.html
        Aaron Price is a former ACS Documentation Services Assistant

        Mutually Assured Redundancy

        The Internet was originally designed by the US military in an effort to speed up weapons research. So the originators designed their first network with nuclear war in mind, something on a lot of people's mind in the early 1960's. Because of this they developed a networking system (ARPANET) that would still function with huge holes blown out of it. This theory of redundancy is at the heart of the modern Internet.

        The Internet is made up of a bunch of large networks. They are interconnected at many places so that if one goes down, information can be routed through a second connection. The father of the modern Internet backbone was run by the National Science Foundation and called NSFNET. Universities, government agencies, and other access providers would connect to multiple points on this network. This was their way to ensure redundancy.

        However, in 1993 the NSF opted out of the Internet backbone business. The idea of one large net with lots of "onramps" was replaced by a bunch of private "onramps" that would voluntarily connect to each other. These new "onramps" are called Network Access Points (NAPS).

        NAPS are run by commercial companies like Sprint, MCI, and PacBell. The big three "official" NAPS are in San Francisco, Chicago, and Pennsauken, New Jersey. All Internet connections are theoretically connected to these three points.

        However, using a concept called "hot potato routing" networks are gaining more control over where their data travels on the Internet. This allows routers to find the quickest way to a destination, even if it means bypassing one of the original big three NAPS.

        Tracerouting Across The World

        A good way for a surfer to see the backbone in action is by using a traceroute program. These programs send a packet over the Internet and follow it. Think of it like tagging an whale in the ocean then tracking it as it plays, eats, and travels.

        Traceroute programs have been written for almost all operating platforms including DOS, Windows 3.1, Macintosh Open Transport (not MacTCP), UNIX, VAX, etc. In [UltraNet's] PC toolkit we have included a program called HopCheck. Windows 95 comes with a traceroute client built-in called tracert.

        If you ever have trouble connecting to a site run HopCheck or a traceroute program on it. This will tell you where you are being blocked so you can know what is keeping you out. Here is an example of a typical traceroute I made to home.netscape.com.

        The first column is the hop number, the second is the site name, the third is their IP address, and the final three columns are the times each packet took to get a response.

        traceroute to www29.netscape.com (198.95.251.72), 30 hops max, 40 byte packets

        1   infra-w-4.mbo.ma.ultra.net       (199.232.56.71)     2 ms    3 ms    2 ms 2   agis-ultra.boston.agis.net       (206.185.153.37)   66 ms    5 ms    5 ms 3   a0.717.washington2.agis.net      (206.185.153.234)  27 ms   44 ms   21 ms4   mae-east.agis.net                (192.41.177.145)   22 ms  199 ms   36 ms5   cpe3-fddi-0.washington.mci.net   (192.41.177.180)   36 ms   26 ms   30 ms 6   core1-hssi3-0.Washington.mci.net (204.70.1.221)     28 ms   30 ms   26 ms

        In this traceroute, the packet got "stuck" at MCI's Washington DC hub. Here is an example of a successful traceroute:

        traceroute to www29.netscape.com (198.95.251.72), 30 hops max, 40 byte packets

        1   infra-w-4.mbo.ma.ultra.net           (199.232.56.71)      2 ms    1 ms    2 ms2   agis-ultra.boston.agis.net           (206.185.153.37)     4 ms    4 ms    4 ms3   a0.717.washington2.agis.net          (206.185.153.234)   18 ms   30 ms   26 ms4   mae-east.agis.net                    (192.41.177.145)    26 ms   21 ms   23 ms5   cpe3-fddi-0.washington.mci.net       (192.41.177.180)    97 ms   54 ms   25 ms6   core1-hssi3-0.Washington.mci.net     (204.70.1.221)      26 ms   26 ms   25 ms7   borderx2-fddi-1.SanFrancisco.mci.net (204.70.158.68)    117 ms  229 ms  263 ms8   borderx2-fddi-1.SanFrancisco.mci.net (204.70.158.68)    254 ms  219 ms  225 ms9   netscape-ds3.SanFrancisco.mci.net    (204.70.158.122)   124 ms  123 ms  120 ms10  www29.netscape.com                   (198.95.251.72)    114 ms  114 ms  122 ms

        Future Wizardry 1

        When performing multiple traceroutes you will find that sometimes the path changes. This is a function of the aforementioned technology of "hot potato routing". UltraNet uses state-of-the-art technology to find the quickest path between you and your destination. Sometimes it is quicker to go through Dallas than Chicago when Chicago is congested. UltraNet's system will determine this and send you along the appropriate route.

        One brand new form of routing technology that applies this idea to a local scale is called Asynchronous Transfer Mode (ATM). UltraNet uses ATM technology to control UltraNet's main T3. This allows us to get you to your destination on more than one path simultaneously. Basically it means better redundancy and quicker access time.

        Metropolitan Fiber Systems (MFS) has created an alternative to the NAP system. Their method wires up an entire city via a complicated and in-depth array of fiber-optic connections. These Metropolitan Area Ethernet (MAE) connections give tremendous redundancy and offer inexpensive connections to corporations operating in that city. MFS offers this service in San Jose (MAE-WEST), Los Angeles (MAE-LA), Dallas (MAE-DALLAS), Chicago (MAE-CHICAGO), and Washington DC (MAE-EAST).

        Right now there are two basic ways to deal with congested networks. The first is to effectively use new technology and streamline network architecture. This includes ATM technology, better routing equipment, and increasing physical bandwidth capabilities. The second method is simple, build more NAPS to create more infrastructure and redundancy. So far NAP companies have been up to the challenge and most of them have announced plans to significantly upgrade their technology over the next year.2

        1 UltraNet is a commercial backbone Internet Service Provider (ISP), whereas UNT is a non-backbone ISP, providing Internet services only to our campus community. We have some features of service providers, but are in many ways just customers ourselves. UNT does not use ATM technology at this time. We do attach to MAE-DALLAS via two load-balanced T1 lines.

        2 Internet II is another effort to improve the Internet. Click here for more information.


        Previous Article <====> Next Article

        If you have any problems or questions about this server, contact us assoon us as soon as possible. You can send mail to the following address:www@unt.edu