______________________________________

Please sign: I pledge that I have neither given nor received help on this assignment.

______________________________________

Please answer the following questions:

• 1. Which of the following would generally be the greatest disadvantage of a network layer that natively maintains a notion of end-to-end connection flows? (1 point)
  (a) Its routers will tend to have very large forwarding tables
  (b) The end system device will be "dumb"
  (c) Routers will maintain per-flow state

• 2. Which of the following issues motivated CIDR? (1 point)
  (a) Inefficient utilization of the IP address space
  (b) The inability to support IP multicast
  (c) The need to simplify forwarding table lookup operations

• 3. In which of the following cases it is more likely that an IP packet will be fragmented en-route? (1 point)
  (a) The source node is on PPP, while the destination is on an FDDI network.
  (b) The source node is on an FDDI network, while the destination is on PPP. (because the maximum packet size for FDDI is larger than that of PPP, which causes the gateway router between the two networks to fragment IP packets)
  (c) The source is on an 802.11 network, while the destination is on an 802.3 network.

• 4. Your Seattle-based startup company owns the entire IP address space that belongs to the class-B IP network 131.106.0.0. The company has 1000 machines (with one IP address per machine) and does not anticipate any growth in its computing infrastructure. Due to hard financial times, the company decides to sell its IP address space to nearby Microsoft, except for one subnet that should be just large enough to include the company's computers.
  (a) What should the length of the subnet field (in bits) of IP addresses of hosts in your company after the Microsoft arrangement takes place? (1 point)
  Answer is 6 bits. Observe that you need a subnet that is only large enough for 1000 machines. Therefore, the host field has to have 10 bits (2^10 =1024). In a class B network the first 16 bits are reserved for the network address (and to identify the class). Only 16 bits are left to identify the subnet and host. If of these, 10 bits are needed to identify the host, the subnet field must be 16-10=6 bits long.

  (b) Your local ISP configures its edge router such that all packets to network 131.106.0.0 are forwarded to Microsoft (via interface 0) except those packets that are destined for your subnet. Packets destined to your subnet are forwarded to your company's router (via interface 1). You arrange with the ISP (and with Microsoft) to give your subnet the numeric ID: 1 (which would be the content of the subnet field). Write two forwarding table entries (for the ISP's router) to configure the router such that it forwards traffic properly to your subnet and to Microsoft on network 131.106.0.0. (2 points)
  

  Subnet Number       Subnet Mask          Interface
  131.106.0.0         255.255.0.0          0
  131.106.4.0         255.255.252.0        1
  

  Observe that in the above configuration, traffic that goes to your subnet will match both entries. CIDR will choose the longer match which is the second entry (please read CIDR). In contrast, traffic that matches the first entry but not the second will go to Microsoft, which is precisely what you want.

  (c) In addition to the two entries you just wrote, the ISP realizes that the same router has a third entry, which routes data to Microsoft Research. It turns out, Microsoft Research also branches off interface 0. It owns the entire class-B network at address 131.107.0.0 (e.g., try telnet www.research.microsoft.com). Can you reduce the aforementioned three entries into two without losing any information in the forwarding table? If so, write the two forwarding table entries needed. If not, explain why this is impossible. (1 point) Yes, it is possible to abbreviate the routing table to the following two entries:

  Subnet Number       Subnet Mask          Interface
  131.106.0.0         255.254.0.0          0
  131.106.4.0         255.255.252.0        1
  

  Observe how the second byte of the subnet mask in the first table entry is 254, not 255. This means that only the first 15 bits of the IP address are to be matched. Hence, both 131.106 and 131.107 will match the first entry and be routed to Microsoft, unless the IP also matches the second entry (a longer prefix) in which case it will be routed to you.

• 5. Compare ATM switches and IP routers in terms of (i) efficiency of lookup in the forwarding table, (ii) robustness with respect to failure, (iii) frequency of table updates, and (iv) the length of the packet header field based on which forwarding decisions are made. (2 points)

  
  Metric           IP                 ATM
  Lookup           Less efficient     More efficient
  Robustness       Higher             Lower
  Table updates    Less frequent      More frequent
  Field            Longer             Shorter 
  
  

• 6. ATM routers use a different VCI to refer to the same connection at each node along the connection's path. Why? (1 point)

  (1 point)

  To avoid having to assign globally unique VCIs. Assigning a globally unique id to each connection on the Internet will be extremely difficult. In contrast, each router can easily assign a locally unique id.