Chapter 20 Cyganski Book

Monica Stoica, Boston University

smonica@cs.bu.edu

Sending mail

• Suppose we want to send a piece of mail to our friend David Coldera's office at his place of employment. We address an envelope to him via a well known protocol for paper mail. The address describes the broad geographic region on the last line, and narrower and narrower destination descriptions as we work our way up the lines.

Mail• We then stick the mail in a mail pick-up tray in our

department's office which is labeled: OFF CAMPUS. • An interoffice mail handler will come by and toss this

tray into a sack with the ``Off Campus MailServices'' address for delivery to the office that handles mail leaving the university. This first delivery has nothing to do with the address we wrote on the envelope, but simply is determined by the internal code implied by the tray we used and the label on the sack it was placed into.

• Once it arrives at the Off Campus Mail Services office, the letter is removed from the sack and routed into the U.S. Mail sack versus one of the sacks used for exchange of mail between a few local universities. This choice is made on the basis of the destination address we wrote on the envelope.

Continuation• At the post office, our friend's name and division are

altogether ignored. The post office forwards the letter to other post offices, based first upon the city and state (and finer zip code information, in reality) and then finally on the company address.

• Once it arrives at the company, interoffice mail workers place the envelope in a box destined for a particular building and jot the room number on the envelope based upon the division location and the person's name. The detailed delivery information is obtained by looking up the name in a company mail directory.

• Finally, a mail delivery person places the letter into our friend's in-box based on the match of the room numbers (after all, they can't be expected to remember everyone's name in a large company).

delivery of an e-mail. • We will now send a piece of e-mail to our friend David

Coldera's PC at his place of employment. We will address the e-mail via a well known protocol. The address describes the broad category of recipients (.com) in the last segment of to specify, and narrower and narrower destination descriptions as we work our way toward the beginning of the line with our friend's computer name and account name last in this right-to-left reading of the address parts.

• Our computer checks the tail end of this address and determines that it is destined for an off campus location. (It would have to have ended in .bu.edu to remain on campus.) So, our computer sends it to a router (bbbisp.bu.edu), which handles all data packets going off campus.

Email• A   router is an electronic device that will handle data that

arrives and departs from the Internet Service Provider's (ISP's) network. Our computer sends the data there by wrapping it in an electronic sack with the router's LAN address. The sack is an Ethernet packet and the OUI for the router is the label on the sack. This first movement of the e-mail has nothing to do with the detailed address of the destination, and is simply based on the fact that the destination is off campus.

• Once it arrives at the router, the e-mail is removed from the original Ethernet packet and routed onto one of several high-speed digital telephone T1 cables. In this case it is routed to one that connects this router with our ISP's network. The decision is again simply based on the fact that we have no direct connection to the final destination.

Email• a router uses a directory scheme known as the Domain

Name System to look up the host name in the address (in this casezoggy.clark.com), which provides a more detailed geographic code known as an IP address (think of it as a zip code) for the routing of the data.   

• Once it arrives at the company's router (generic data.bbb.net), the data is placed in another Ethernet packet destined for the particular PC that appeared in the address (zoggy). That Ethernet address is obtained by looking up the computer's name in a local directory (it is called an Address Resolution Protocol, or ARP, table). 

• Finally, having reached the computer, zoggy, the e-mail is removed from the Ethernet packet and placed into our friend's e-mail arrival box.

Addressing Schemes• One of the facts exposed by the above treatment is that the

exchange of information on the Internet is directed by three different sets of addresses! This may seem unnecessary at first glance, but there are actually good reasons for using each (some were indicated in our mail analogy above).

• OUI Addresses: These 48-bit addresses uniquely identify every Ethernet (and Token-Ring LAN card) ever made. These are permanently attached to each card at the time of manufacture.

• IP Addresses: These 32-bit addresses identify every attachment of a machine to the Internet (this has been carefully worded to be exactly correct as a single computer may be attached in more than one way to the Internet). Groups of these addresses, called sub nets, are geographically co-located.

.• DNS Host names: These alphanumeric addresses parallel

the IP Addresses in identifying individual network connections and sub nets, but are further distinguished by a domain hierarchy based on some or all of the following: country code, service code, network name, and/or organization name. Let's examine a concrete example. There is a computer on our campus that is known as cs.bu.edu . This name is actually the DNS host name for the computer. It is easy for people to remember this name because it breaks down into natural units:• cs:This is the primary computer in the Computer Science (CS)

Department.

• bu:The CS department of which we are speaking is in an institution known as BU.

• edu:BU is an educational institution.

.• The DNS system was invented not to route information, but

to make it easy for people to remember computer addresses. Why not just remember the IP addresses that the Internet routers themselves will use? The IP address that one obtains on looking up cs.bu.edu in the DNS directory is let’s say: 130.215.16.20. Will you remember this address tomorrow?

• Why do routers use the IP address instead of directly using the DNS host name?This address is broken into parts that can be used to route the data over a long trip without being buried in detail. In the case of our example, the first part of the address, 130.215, is reserved for exclusive use by WPI. Thus, based on this part of the address, e-mail can be routed to WPI from anywhere in the world. Once it gets here,the next segment (16) can be used to route the data to a specific building, while the last (20) is routed to the specific machine.

• Finally, when the data gets to the router that will place a packet on the Ethernet LAN for the last leg of the trip, the Ethernet address 08-00-2B-BC-C2-BC can be used to alert the specific machine. 

• Although it is complicated, the protocols used to build the Internet actually arise from common-sense solutions to problems of information engineering that have been encountered before in other forms of human communication.

Tracing a Route

• There is a means by which we can have every routing device on the path of a datagram report its name to us. Using this we can discover the path that our Internet traffic will take from one site to another. In the next table we show the record of such a discovery process, from a campus to the computer named rover.acorn.net, which is in the Akron-Summit County Public Library library in Akron, Ohio.

Routers• Two interesting things can be gleaned from the table . • The first is the number of routers (16) that need to be

traversed on this path. This provides a good picture of the complexity that underlies that Internet despite the impression one obtains of the Internet being a single entity.

• The second piece of interesting information is related to the three numbers at the end of each of the 16 router identification lines. Each router along the path was actually probed three times.

• Hence, with 16 routers 48 round-trip probes were completed, and the times were recorded. The three numbers report the round-trip time (in thousandth of a second) from our machine initiating the probe until the response returned.

Routers• For example, the three probes of router number 12 yielded

round-trip times of 160.183, 189.177, and 301.07 milliseconds. Note the wide variation from one probe to the next!

• In fact, while generally the round-trip response for farther routers took longer than for closer ones, since the probing tests are all separate sometimes all three tests of a closer router (like number 12) happened to take longer than the next three probes of a farther router (number 13).

• This underscores the fact that the network cables and routing equipment in the Internet are all shared! If the level of communications traffic happened to increase momentarily while we were probing router 12, the round trip time would exceed that of number 13 even though this test passed through 12.

The backbone of the Net• The various ISPs are like the letter carriers who pick up

and deliver the mail to your door. • The ISPs will handle movement of data directly that

passes from one customer to another of the same ISP. • However, data destined for another ISP in the same region

is passed between the ISPs through the services provided by a regional Internet access provider.

• Data headed outside the region will be passed by the regional provider to a national provider, which in turn may purchase access to a backbone cable from a backbone provider to send large collections of data from coast to coast (or similarly large distances).

Details• At a greater level of detail, the Internet loses the

appearance of a simple and well organized hierarchy like that presented above.

• Here’s an example of a small piece of the Internet in which some of the transmission technologies have been revealed.

• On the next image you will see, on the left, three LANs (two based on two types of Ethernet and another on a Token-Ring network) that connect to routers at their locations in order to connect to an ISP. The ISP will connect them to each other and to the remainder of the Internet.

Now, how do these three customers connect to their ISP?

• The ISP does not hire a back hoe and dig a trench to lay new wire for this connection.

• Instead, the ISP leases connections from the local telephone company. Hence,the telephone company provides T1 and T3 lines (described in an earlier chapter of this book) to connect the data from these companies to the telephone exchange switches.

• Thus, the telephone company's role is to set up virtual circuits from the companies' routers to the ISP's Internet Protocol (IP) router (via lines labeled A, B, and C in the diagram). The telephone company's equipment is actually unaware of the content of the data being moved; its job is simply to be a conduit.

ISP• Now, at the ISP, a router examines the Internet

addresses in the data packets and makes decisions as to which way to direct the packets. Some are directed (downward in the diagram) via another telephone T1 line directly to other customers of the ISP. The ISP in this case has actually linked two nearby customers via another telephone-company-provided circuit (V.35) so as to not have to run two long high-bandwidth cables to both.

• Some data, however, are directed (upward in the diagram) via high-bandwidth connections (in this case an optical-fiber-based Wide Area network, or WAN) to a regional access provider where it may end up being directed to another ISP or onward toward a national provider.