At the March 7 Lunch ‘n Learn, OIT’s Senior Manager of Networking, Peter Olenick presented Campus Networking and the Internet: How They Work.
“Most people really don’t care how networking works… they simply care that it does work,” summarized Olenick.
During his hour session, Olenick reviewed the technologies involved in campus and internet networking with some focus upon how that technology is actually used on campus and why this complex environment sometimes does not operate as users expect it to.
This image of a network pipe illustrates some of the complexity. There are electronics, logical components, there are network services, and then there are the actual applications that run on top… together, they make up what most users think of as their network connection. All aspects must work in order to provide useful service.
Olenick explained that the current trend in network technology for the past several years has been to talk about a seven-layer model. His talk focused not upon software applications or network sessions or the physical wire or how the light transmits, but rather upon layers two and three, the soft protocols to which we have all agreed in order to guarantee the efficient flow of data.
The network involves packets of data. Data is stored in wrappers, sometimes with wrappers stored within wrappers. As information flows, some wrappers will be stripped away and others added, all to make sure that information is delivered appropriately from point to point.
Ethernet is the pervasive technology used to provide network service at the local level, in other words within the campus. Packets are placed in envelopes on the campus’s ethernet network. There are three important pieces in the envelope: Where is the data going, where did it come from, and what type of data is it.
Ethernet or Media Access Control [MAC] addresses are unique for each PC or Ethernet device in the world. The first three characters are assigned by IEEE; the manufacturer uniquely assigns the remainder.
Olenick discussed how Ethernet networks are constructed. There is the Ethernet cable, there are hosts with an Ethernet interface card, and there is the Ethernet protocol that allows only one device to speak at a time. A host sends out a packet and each device on the network receives the packet. As the bits are assembled, one bit at a time, each host looks to see if the destination address matches its own MAC address. If there is no match, the host simply stops listening. The host with the matching address will continue to read the packets off the Ethernet, and assemble them so that they can be used by the applications for which they are intended. As part of the delivery, the network checks for various kinds of errors.
Along the delivery path, the network is constructed hierarchically with campus routers, that serve portions of the campus or which can forward messages off campus to the internet. The Princeton campus network uses a fiber optic backbone to connect approximately 100 academic/administrative buildings, 50 dormitories, eating clubs, and apartments, and 26,000 registered devices. The campus has two Internet service providers and has Internet2 connectivity for research that requires high speed networking.
The University’s core is set up within the Computer Center at 87 Prospect Ave.
The core campus network radiates out to 11 hub sites in various campus buildings. Within most buildings, there are wall boxes within closets providing twisted pair or fiber optic connections to the desktop. In no case is the connection to the desktop more than 100 meters to a desktop. The architecture facilitates local traffic. It’s possible, for example, to communicate within departments without burdening the campus’s core network.
Olenick also presented graphs that illustrate the flow of data by day, month, and year. Not surprisingly, traffic varies substantially depending upon the time of day and whether students are on campus.