Fundamental Questions to Answer About Computer Networking, Jan 2009 Prof. Ying-Dar Lin, ydlin@cs.nctu.edu.tw Chapter 1: Introduction 1. How does Internet scale to billions of hosts? (Describe what structure and levels are used to organize the hosts, and calculate the numbers of entities at each level.) 2. Routing vs. switching: stateful or stateless, connection-oriented or connectionless, matching or indexing? (Associate these features with routing and switching.) 3. What may increase or decrease the latency inside Internet? (What are the factors that might increase or decrease the latency of queuing, transmission, processing, and propagation, respectively?) 4. What do Little s result and bandwidth-delay product tell us? (Hints: The former is about a node, while the latter is about a link or path.) 5. What does the end-to-end argument say about networking? 6. According to the end-to-end argument, at which single layer should we put error control for Internet? But then why do we put it in many layers including link, IP, and transport layers? 7. What types of mechanisms should be put into control-plane and data-plane, respectively? (Specify their type of packets, purpose, granularity of processing time, and example operation.) 8. What are standard and implementation-dependent components in a router? (Specify their types of components and example.) 9. What s inside a Linux distribution? (Specify what types of files you would find in a distribution and how they are organized.) 10. How does a Linux system boot up? (Describe its procedure.) 11. When do we implement a mechanism of a network device into ASIC, driver, kernel, and daemon, respectively? (Specify their guidelines and examples.) Chapter 2: Data Link Layer 1. What are the byte and bit orders in IP over Ethernet? 2. Why FCS at the tail? IP Checksum in the header? 3. Why is big bandwidth delay product (BDP) bad for CSMA/CD? 4. What is the problem in half-duplex gigabit Ethernet? 5. What is the minimum frame length, in meter, when transmitted in gigabit Ethernet? 6. Why not CSMA/CD for Wireless LAN? 7. What problem does the RTS/CTS mechanism solve for CSMA/CA in wireless LAN? 8. What are the differences between collision domain, broadcast domain, and VLAN? (Describe what they are, their scopes and whether they can overlap.) 1
9. Layer-2 bridging vs. layer-3 routing? (Compare their forward mechanisms, management, and scalability.) 10. Layer-2 bridging on a large campus network? Why not? 11. Why do we say bridges are transparent to hosts while routers are not? 12. Why do we need a spanning tree in transparent bridging? 13. How do we design a MAC in an IC? (Describe the general design flow and the variables used in programming) 14. How does a driver work to send and receive frames? (Describe the handling of outgoing and incoming packets, with hardware and interrupt handling.) 15. How do you compile, install, and run a Linux driver? (Describe its general procedure.) 16. What does a network adaptor driver want when it probes the hardware? For what? 17. What interrupts may lead a system to execute a network adaptor driver? Chapter 3: IP Layer 1. Why do we need both MAC address and IP address for a network interface? Why not just one address? 2. Why are MAC addresses flat but IP addresses hierarchical? 3. Why is netmask used inside a router and a host? 4. Routing vs. forwarding? (Compare their type of work and the used algorithm.) 5. Why there could be multiple matched IP prefixes in router s table lookup? (Explain what network configurations could result in this.) 6. How is the longest-prefix matching done in the Linux kernel? Why is the matched prefix guaranteed to be the longest? 7. How is the forwarding table organized in the Linux kernel? 8. What header fields are needed in IP reassembly at destination hosts? 9. What packet modifications are needed for FTP through NAT? 10. What packet modifications are needed for ICMP through NAT? 11. How is the NAT table implemented in the Linux kernel? 12. What header fields in IPv4 are removed from or added into the header of IPv6? Why? 13. How can IPv4 and IPv6 co-exist? 14. How does a host translate IP addresses to MAC addresses through ARP? 15. How does a host obtain its IP address through DHCP or ARP? 16. How are ping and traceroute implemented? 17. How does the count-to-infinity problem occur in RIP? 18. RIP vs. OSPF? (Compare their network state information and route computation.) 19. Distance vector routing vs. link state routing? (Compare their routing message complexity, computation complexity, speed of convergence, and scalability.) 20. RIP vs. BGP? (Summarize their similarities and differences.) 2
21. Why do RIP, OSPF, and BGP run over UDP, IP, and TCP, respectively? 22. Can you estimate the number of routing table entries in intra-as and inter-as routers? (Estimate the range or order of magnitude.) 23. How do routing protocols in zebra exchange messages with other routers and update routing tables in the kernel? 24. How does a router know through IGMP whether hosts in its subnet have joined a multicast group? 25. Does a router know exactly which hosts have joined a multicast group? 26. Source-based vs. core-based multicast tree? (Compare their number of states and scalability.) 27. How many pieces of state information are kept in routers for source-based and core-based multicast routers, respectively? (Consider the numbers of multicast groups and sources.) What kinds of state information might be kept? 28. Do multicast packets really flow on the shortest path in the reverse path multicasting in DVMRP? 29. What state information is kept in routers for the reverse path multicasting in DVMRP? 30. How is the RP of a multicast group determined in PIM-SM? Chapter 4: Transport Layer 1. Layer-2 channel vs. layer-4 channel? (Compare their channel length, error, and latency distribution.) 2. TCP vs. UDP? (Compare their connection management, error control, and flow control.) 3. Why does most real-time traffic run over UDP? 4. What mechanisms are needed to support error control in TCP? 5. Why does TCP need 3-way handshake, instead of 2-way, in connection setup? 6. When is a lost TCP segment retransmitted? 7. What factors are considered in deciding the TCP window size? 8. How does the window grow in slow-start and congestion avoidance? 9. Why are fast retransmit and fast recovery added to TCP? What major change does New Reno make? 10. How is the socket implemented in Linux? (Briefly describe the processing flow of socket functions and the data structures of sockets.) 11. What are the available kernel-space and the user-space programs of Linux to filter and capture packets? 12. What extra support can be done over RTP and RTCP than UDP? Chapter 5: Internet Services 1. Why are protocol messages of most Internet applications ASCII and variable length? 3
2. Why do servers running over TCP and UDP have concurrent and iterative implementations, respectively? 3. How do DNS servers resolve a domain name into an IP address? 4. How do DNS servers resolve an IP address into a domain name? 5. What resource records are used in forward-dns and reverse-dns, respectively? 6. What entities and protocols are involved if you send a mail for your friend to read? 7. POP3 vs. IMAP4? (Compare their number of commands, flexibility, and usage.) 8. What POP3 messages are exchanged when you download mails from the server? 9. What HTTP messages are exchanged in downloading, filling, and uploading a Web form? 10. What does the connection persistence mean in HTTP 1.1? 11. Forward-caching vs. reverse-caching? (Compare their location and cache content.) 12. How does an HTTP proxy intercept HTTP requests destined to HTTP servers? 13. What does an HTTP caching proxy do if it has a cache miss? How many TCP connections does it establish for a specific client? 14. Active mode vs. passive mode in FTP? (Describe from whose perspective is the mode and how the data connection is established.) 15. Control and data connections in FTP? (Explain why we need two connections.) 16. What protocol messages on the control connection are exchanged in downloading and uploading a file in FTP, using active mode and passive mode, respectively? 17. Why is streaming quite robust to Internet delay, jitter, and loss? Chapter 6: Internet Quality of Services 1. What control-plane and data-plane mechanisms are needed to provide QoS guarantee in Internet? 2. WFQ (Wieghted Fair Queuing) vs. WRR (Weighted Round Robin)? (Compare their complexity and scalability.) 3. Why is RED (Random Early Discard) better than FIFO (First-In First-Out), especially for real-time traffic? 4. IntServ vs. DiffServ? (Compare their QoS granularity, complexity at edge routers, complexity at core routers, and scalability.) 5. What are the barriers to deploy InteServ? (List three barriers.) 6. What are the barriers to deploy DiffServ? (List three barriers.) Chapter 7: Internet Security 1. Private key algorithm vs. public key algorithm? (Compare their computation complexity, security, and usage.) 2. How can we combine the private key algorithm and the public key algorithm? 3. Where do we use transport-mode IPsec and tunnel-mode IPsec, respectively? What parts of 4
the packets do they encrypt? 4. In the tunnel-mode IPsec, if authentication is done before encryption, what sequence of headers would we have in the packets? (The headers include AH (Authentication Header), ESP (Encapsulation Security Payload), IP, TCP or UDP.) 5. SKIP (Simple Key management for IP) vs. IKE (Internet Key Exchange)? (Compare their operations and barriers to deployment.) 6. Packet-filter vs. application-proxy firewalls? (Compare their purpose and where they are implemented in Linux systems.) 7. Virus vs. worm? (Compare their characteristics and model of replication.) 8. DoS (Denial of Service) vs. buffer overflow attack? (Compare their purpose and operations.) 9. Under what situation will an IDS (Intrusion Detection System) such as snort have false positive or false negative? 5