Master of Science in Internet Engineering

Internet Technologies I

Assignments

Assignment 1: Initial reading assignment

Please read chapters 1 (sections 1.1-1.3), 2 (sections 2.1-2.4), 3 through 5, and 7 through 10 (don't get bogged down with sections 7.7.4-7.7.6 or 7.8; and you can skip 10.4-10.5 and 10.22) of Comer. We dealt with some of this material in the first class session, and will cover the rest in the second (and possibly part of the third).

Assignment 2

• Reading assignment: please read chapters 12 and 13 of Comer on UDP and TCP. We'll be spending most of next class session on some remaining issues relating to IP, but then we'll be turning to the "transport layer" protocols, and we'll be looking closely at TCP in particular.
• Written assignment, due October 27th. This problem relates to IP addressing and routing.
  
  Assume that
  • an ISP has delegated to your organization the class C IP address blocks 216.114.150.x and 216.114.151.x, and that your task is to configure your organization's networks to connect to the Internet through the ISP using these addresses.
  • the topology of your organization's networks is as in the diagram above: you have three routers, and five internal networks: additionally, the numbers of computers on your various networks is such that you'd like to effect the following distribution of addresses over the networks:
    • half of the 216.114.150.x block should be allocated to network 1
    • one quarter of the 216.114.150.x block should be allocated to network 2, one quarter to network 5
    • half of the 216.114.151.x block should be allocated to network 3, half to network 4.
  • you connect to the ISP via a point-to-point line such as a T1, and that the ISP has told you to configure your end of that line (i.e. the external interface of the router at the outer edge of your network) with the IP address 210.210.210.210 and mask of 255.255.255.252, and also that the IP address of the interface at their end of the point-to-point line is 210.210.210.209.
  For the purposes of this exercise, all you need to do is "configure the routers" R, R2, and R3, i.e.,
  • assign IP addresses and masks to their interfaces, consistent with the required division of the IP address space
  • specify any routing table entries needed to reach a destination machine on any of the networks. Please--for pedagogical purposes--specify explicit routes to each of the internal networks (i.e., don't simply rely on the "default route" mechanism).
  • it's not necessary to say anything about the configuration of other computers in the system--we'll just assume that networks 1-5 contain a bunch of computers and that their configuration just happens magically.

  To help you with this problem, I've expanded my session notes to include some worked-out packet-forwarding examples.

  Optionally

  • Describe how the routing tables of R, R2, and R3 would differ (if at all) if you do take full advantage of the notion of the "default route"
  • What's the most economical way the ISP can represent your networks in its routing table?
  • What would be the most economical way for the ISP to represent your networks in its routing table if it had given you the address blocks 216.114.151.x and 216.114.152.x (instead of 216.114.150.x and 216.114.151.x)? Can you make a general observation of the form "if my ISP gives me an address block of size x, the first address in that block will have the form y"?

Assignment 3

• The following list is a mix of analytical and implementation-oriented homefun possibilities. Please do at least one of them or, if there's another issue having to do with one of the topics we've covered this term that particularly interests you, propose an alternative. Due date is Wednesday, December 12th.
• Let's say the administrators of the domain msie.marlboro.edu want to subdelegate wild.msie.marlboro.edu and IP addresses in the range 216.114.151.13-216.114.151.27. Your task is to provide name service for the subdelegated zones. Do the following: (a) install version 9.1.3 of the DNS server bind (see http://www.isc.org) on your Linux box; (b) configure it to handle the domain and addresses in question (the Linux DNS HOWTO will be useful to you: http://www.ibiblio.org/pub/Linux/docs/HOWTO/DNS-HOWTO); (c) tell me what changes need to be made to the nameserver handling msie.marlboro.edu and 151.114.216.in-addr.arpa to make this work. On the question of delegating a block of addresses smaller than the size of a class C, see RFC 2317 (http://www.ietf.org/rfc/rfc2317.txt) and my notes on DNS from the last class session.
• Use the packet sniffer /usr/sbin/tcpdump on your server to generate a packet trace of a TCP connection (or a set of related TCP connections, such as a typical web-browsing event might involve). Prepare an annotated version of the tcpdump output in which you characterize the connection in terms of the various elements of TCP we've discussed: connection management, end-to-end flow control, retransmission, congestion control, and so on (of course, depending on the particulars of the connections you generate, there may not be interesting things to say under all these headings. Moreover, if your connections involve repetitive patterns--e.g. an FTP connection that transmits a 10 mb file in a sequence of virtually identical 1 kb chunks, don't feel obliged to annotate each and every packet). "man tcpdump" will tell you all you've ever wanted to know about tcpdump, but to get you started, you might try running tcpdump in one window and generating a TCP connection in another as follows: in one window, type "/usr/sbin/tcpdump -w outfile host www.ntu.edu.au", where "outfile" is the name of the file to which tcpdump will direct its output, and "host www.ntu.edu.au" restricts the packet capture to just those packets moving between your machine and a specific target; and then in a second window, telnet to port 80 on the target machine, and after the connection's been established (in the following, what you're supposed to type is in bold; after the "GET" line, make sure to hit return twice).

  [root@black-rock-city /root]# telnet www.ntu.edu.au 80
  Trying 138.80.128.23...
  Connected to www.ntu.edu.au.
  Escape character is '^]'.
  GET /index.html HTTP/1.0
    

  This should generate an HTTP response from the server, and the connection should close out. Now you can go back to the first window, do a control-c to shut the packet capture down, and if you then do "tcpdump -r outfile", or probably better, "tcpdump -r outfile | more", you'll see the raw packet trace
• Configure your Linux box to use IPv6. See http://www.bieringer.de/linux/IPv6/IPv6-HOWTO/IPv6-HOWTO.html. This is a very open-ended proposition, and before you attempt it you should both feel pretty comfortable mucking around with Linux and be prepared for glorious failure. When you're ready to talk to an IPv6 peer, I'll provide one.
• Although TCP adapts to perceived network congestion, transport protocols aren't obliged to do this--UDP, notoriously, doesn't. Further, while TCP adapts to perceived network congestion on a per-connection basis, it doesn't do any kind of congestion control that involves coordination among multiple concurrent connections. For both of these reasons, the congestion avoidance scheme in place in the Internet currently is widely thought to be problematic. Balakrishnan and Seshan have proposed something called the "Connection Manager" to remedy the situation--a module that would control the sending rates of all the data flows emanating from a particular device. Starting at http://nms.lcs.mit.edu/projects/cm/ and with the documents pointed at from that page, write a brief (~ five page) paper characterizing the problem that the Connection Manager's trying to solve and the way it tries to solve it.