ECE 4380 Final Exam Cards

Suppose the following error polynomial occurs: E(X) = x^8 + x^4 + x + 1. Can this error be detected with the generator polynomial x^4 + x + 1? Explain why or why not.

Yes, because there is a remainder

Most wireless networks organize the topology with base stations or access points. Describe an alternative topology that is seeing increasing interest and that does not use base stations.

"In an ad hoc network, nodes are peers; that is, there is no special base station node." pg. 133 in book. Ad hoc networks don't require preexisting infrastructure and routes between nodes may contain multiple hops.

The link-state algorithm requires that link-state packets (LSP's) be reliably flooded to all other routers. If a router receives an LSP from a neighbor X, how does it (a) determine if it should forward the LSP and (b) to which neighbors does it send the LSP?

(a) If the sequence number is larger than stored in local LSP-database (and LSP has not timed out) (b) All neighbors except for neighbor X

SONET uses clock-based framing and looks for a special bit pattern to synchronize to the frame boundaries. What is done to (a) ensure there are sufficient transitions to maintain timing and (b) guard against the special bit pattern occurring in the payload?

(a) Send dummy data to keep the clock NRZ encoding, data scrambled (XOR'd) with 127-bit pattern to create transitions (b) Float Frame -- payload frame floats within clock frame and part of overhead specified the first byte of payload Receiver looks for bit pattern consistently

Name the seven layers of the OSI protocol stack in top down order.

(note: OS kernel) Physical Data Link Network Transport Session Presentation Application (note: user level) Please Do Not Throw Sausage Pizzas Around

Why does the offset field in the IP header measure the offset in 8-byte units? (Hint: recall the offset field is 13 bits long).

- 2^13 = 8192 values - 8192 * 8 = 65536 - 65536 is the max size of an IP datagram. Measuring the offset in 8-byte units allows us to handle fragmentation for all datagram sizes

Explain the TCP model of use.

- Connection setup via 3-way handshake - Data transport -> sender writes some data -> TCP - breaks data into segments - sends each segment over IP - retransmits, reorders, and removes duplicates as necessary -> Receiver reads some data

Describe an advantage of an IP router compared to an Ethernet learning bridge.

Optimal routes, support for different types of LANs, scalability, reduced collision domains, etc.

The body of Ethernet frame must contain between 46 and 1500 bytes. What is the reason that the body must contain at least 46 bytes? Why does the design limit the maximum number of bytes in the body to 1500 bytes?

46 bytes: ensures frame is long enough to detect a collision 1500 bytes: limits max time one node can transmit to allow fairness. 9.6 us of idle time is required between consecutive frames from 1 node

The 4B/5B encoding and Manchester encoding are both designed to help a receiver synchronize its clock with the transmitted waveform. What is the advantage of utilizing 4B/5B instead of Manchester encoding?

4B/5B encoding is 80% efficient. Manchester is just 50% efficient.

For a node connected to an Ethernet LAN, CSMA/CD is used to detect a collision and a slotted protocol is used to determine when to schedule a retransmission. If the node has attempted to forward a frame four times and each time it has detected a collision, what is the probability that the node selects the first slot for the next forwarding attempt for this frame?

4 collisions -> window size = 2^4 = 16

To support the reliable transmission protocol, how many bits must be included in the header for each frame if 16 concurrent logical channels are used for a point-to-point link?

5 bits total; 4 bits for the channel number (2^4 = 16) and 1 bit for the sequence number

In the source routing example of Section 3.1.3, the address received by B is not reversible and doesn't help B know how to reach A. Propose a modification to the delivery mechanism that does allow for reversibility. Your mechanism should not require giving all switches globally unique names.

- To allow for reversibility, the address for every output port could be saved in the header alongside the addresses for input ports. - Could also assign each switch a locally unique name - forwarding switches would fill in the sequence of those names, and if packet is reversed, switches would use these names to look up the previous hop

What 3 events send segments in the TCP Data Transport Model?

- as soon as collect at least MSS bytes of data - explicit PUSH operation by application - periodic timeout

Why do you think IPv4 has fragment reassembly done at the endpoint, rather than at the next router? Why do you think IPv6 abandoned fragmentation entirely? (Hint: think about the differences between IP-layer fragmentation and link-layer fragmentation).

- connections between routers can have different MTUs and can fragment packets differently. However, reassembling packets just to fragment them again at every router could be time-consuming, so it is more efficient to reassemble packets at the endpoint in IPv4. - IPv6 ensures that network equipment delivers packets of size 1280 B or less. So, fragmentation happens at the link layer, as packets should never have to be re-fragmented.

Explain how data is broken into segments in the TCP Data Transport Model

- limited by maximum segment size (MSS) - defaults to 536 bytes - negotiable during connection setup - largest size without causing local IP to fragment packet - typically set to MTU of directly connected network minus size of IP and TCP headers (40 bytes)

In TCP connection establishment, why is the initial sequence number randomly selected?

- prevent 2 incarnations from using the same sequence number too soon - reduce chance that segment from earlier incarnation of connection mistaken for later incarnation

Name an advantage of a level-3 switch compared to a level-2 switch

- support for different types of LANs - improved sorting, scalability, reduced collision domains, etc.

How is a connection established in TCP?

- three way handshake to set up sequence numbers - ACKs identify next sequence number expected - set timers for retransmissions - randomly select initial sequence number

What are the 3 pieces of information that are included in the configuration messages for the spanning tree algorithm?

1. The ID for the bridge that is sending the message 2. The ID for what the sending bridge believes to be the root bridge 3. The distance, measured in hops, from the sending bridge to the root bridge

Clemson University is assigned a block of class C addresses for all 802.11 networks. The block is 198.21.128/17. How many class C addresses are aggregated to create this block?

17 = 16 + 1, so most significant bit must be one. This leaves 8 - 1 = 7 bits for the rest of the class C address block. 2^7 = 128 class C address blocks are aggregated.

Consider an ad hoc wireless network, and assume that nine wireless nodes are arranged in a uniform grid as shown in the figure to the right. For example, r meters separate nodes 1 and 2. Assume that each node has the same signal coverage and a circular shape represents the coverage (eg, if the signal coverage is r meters then nodes 1 and 2 can transmit to each other). What is the minimum signal coverage so that none of the nodes suffer from the hidden terminal problem?

2(sqrt(2r)). In order for the hidden terminal problem to be avoided, all nodes must have signal coverages large enough to detect all of the other nodes. 1 and 9 and 3 and 6 are the furthest away and must be able to detect each other. Solve for the distance between these pairs; this will give you 2(sqrt(2r)).

For the sliding window protocol, if the sender's window size (SWS) equals the receiver's window size (RWS), what is the minimum number of sequence numbers that are required?

2*SWS sequence numbers, or, alternatively, SWS+RWS sequence numbers

Describe the actions performed by IP routers to catch packets that are going around in loops.

A router documents the TTL (hop count) field in an IP packet. If the hop count is zero, the packet is discarded. An ICMP message can be sent to the source.

Protocol-independent multicast (PIM) was developed in response to the scaling problems of earlier protocols. In PIM sparse mode (PIM-SM), routers explicitly join the multicast distribution tree using PIM protocol messages known as Join messages. Where does a router send the Join message?

A router explicitly adds itself to the tree by sending a join message to the rendezvous point (RP)

Protocol-independent multicast (PIM) was developed in response to the scaling problems of earlier protocols. One common problem with multicast protocols is how to determine if a router should participate in the multicast distribution tree. How does the PIM sparse mode (PIM-SM) handle determining if a router should join the tree?

A router explicitly adds itself to tree by sending join message to rendezvous points (RP). Routers that forward join messages add themselves to the tree.

Suppose 2 subnets share the same physical LAN; hosts on each subnet will see the other subnet's broadcast packets. Will ARP be affected by such sharing?

ARP won't be affected as hosts only broadcast ARP queries for other hosts on the same subnet. Hosts on the other subnet will hear these but will not answer. A host on one subnet would answer an ARP query from the other subnet if it were ever issued, but such an ARP query would never be issued.

The sliding window ARQ protocol utilizes cumulative acknowledgements. If the sender's window size is N, the sender could see N-1 duplicate acknowledgements if the first frame in the window is lost. The sliding window protocol we have studied does not attempt a retransmission until a timeout. Why does the sender ignore duplicate acknowledgements even though they indicate that a frame has been lost?

Because retransmissions based on duplicate ACKs can easily lead to the Sorcerer's Apprentice bug. That is, many unnecessary retransmissions.

For computer networks, statistical multiplexing at a switch can result in contention for an output port and congestion at the queue, which leads to dropped packets. Fixed (or synchronous) time-division multiplexing or frequency-division multiplexing do not suffer from dropped packets. Given the problems with congestion, why do computer networks employ statistical multiplexing?

Because statistical multiplexing allows much higher efficiency in utilizing a link for computer traffic

For a wireless network based on the IEEE 802.11 standard it is desirable that there is some overlap in the area that is covered by each of the access points (AP's). Explain one benefit from having some overlap by AP's. Give one drawback that occurs if the overlap is excessive.

Benefit of overlap: support for handoff; account for dynamic conditions. In addition, if an AP becomes disabled, a node can access the other nearby AP. Drawback of overlap: reduced performance. If either AP overhears an RTS or CTS from the overlapping AP, the overhearing AP will be blocked from transmitting. At most, one AP can transmit at a time.

The SONET standard uses a special 2-byte pattern to mark the start of each frame. Bit stuffing is not used, thus the pattern may occur in the payload. In sentinel-based framing approaches either bit or byte stuffing is used to keep the sentinel from occurring in the payload. Why is stuffing not used with SONET framing?

Bit stuffing cannot be used because SONET has fixed length frames

Consider an 802.11 ad hoc wireless network, and assume that nine wireless nodes are arranged in a uniform grid as shown in the figure to the right. For example, r meters separate nodes 1 and 2. Assume that each node has the same signal coverage and a circular shape represents the coverage (e.g., if the signal coverage is r meters then nodes 1 and 2 can transmit to each other but nodes 1 and 5 cannot). Assume that node 1 transmitted an RTS to node 2, node 2 transmitted a CTS to node 1, and node 1 is currently transmitting a data packet to node 2. At this instant, which nodes are permitted to transmit an RTS?

By the rules of the DCF channel-access protocol, when node 1 transmits the RTS, node 4 overhears it and is blocked. Node 2 responds with a CTS, and both nodes 3 and 5 overhear this transmission and are blocked. Nodes 6, 7, 8, and 9 are not blocked and thus are permitted to transmit an RTS. (However, if the RTS specifies nodes 3, 4, or 5 as the destination, then no CTS is sent and the host that transmitted the RTS must wait for a timeout and try again later. Unfortunately, the host does not know how long to wait since it does not know the data rate or the packet size that host 1 is using for its transmission to host 2.)

For a CRC code, what property must the polynomial C(x) have to ensure that all odd number of errors are detected?

C(x) must contain the factor (x+1)

The Ethernet protocol was developed for a network in which the nodes are connected by a coax cable. Many current implementations use a star configuration in which each node is connected to a hub by a twisted pair cable. Describe two functions that are performed by the transceiver for a coax-based network but that must be performed by the hub for a star-based network.

Collision detection Jamming after a collision

In selecting an error detection mechanism for a channel in which bit errors occur randomly and independently which approach is the most likely to correctly detect errors: cyclic redundancy check, two-dimensional parity, or checksums?

Cyclic redundancy check is best for independent and random bit-errors. Two-dimensional parity is better for burst errors, which are often adjacent to one another, and checksums often serve as backup to CRC.

Suppose 2 subnets share the same physical LAN; hosts on each subnet will see the other subnet's broadcast packets. How will DHCP fare if two servers, one for each subnet, coexist on the shared LAN? What problems might (do!) arise?

DHCP will have a hard time determining which hosts belong on which subnet. Subnet assignments depend on which server answered first. Full DHCP implementation deals with this by allowing servers to be manually configured to ignore address-assignment requests from certain physical addresses on the other subnet.

For a local area network (LAN), the media-access control (MAC) protocol coordinates access to the shared link. In the OSI architecture, which layer is responsible for MAC protocols?

Data link layer (layer 2)

What mechanism is used to ensure that each Ethernet network interface card attached to a local area network has a unique MAC address?

Each adapter has a unique, 48-bit unicast address.

How are forwarding tables built for learning bridges?

Each bridge monitors the source address for all frames it receives. The port the frame is received on is added to the table for this address.

Consider the case in which DHCP is used to dynamically assign IP addresses to hosts from an address pool. Describe how DHCP can determine that it should recover an IP address from a host that is suddenly disconnected from the network.

Each host leases the address from the server. If the host does not renew the lease, the server times out and reclaims the address.

What is an advantage of the internet checksum algorithm compared to the cyclic redundancy check?

Easily implemented in software, simple addition

Typically, stuffing is required when sentinel-based framing is utilized. Explain why stuffing is not required if 4B/5B encoding and sentinel-based framing is used?

Every 4 bits of data are encoded in a 5-bit code with at most one leading 0 and at most 2 trailing 0's, but with byte stuffing, every five 1's are followed by a 0. It is not necessary for 4B/5B.

What distinguishes forward error correction (FEC) from automatic repeat request (ARQ) in providing reliable delivery of frames?

FEC uses error correction code rather than error detection. It is less efficient because it includes redundant information in each frame. It is used when the link has high latency, a high error rate, or retransmissions are difficult.

True or False: The stop-and-wait ARQ protocol requires two bits in the header of each frame to be able to distinguish between the four scenarios that can occur: the acknowledgement is received before the timer expires, a lost packet, a lost acknowledgement, and an acknowledgment that is delayed and arrives after a timeout and retransmission.

False, stop and wait sends just 1 header bit

True or False? In the OSI architecture, the transport layer handles forwarding of packets among nodes within a packet-switched network, and is implemented on all nodes.

False; the network layer handles forwarding of packets among nodes within a packet-switched network

True of False? The concurrent logical channel protocol (CLC) requires fewer overhead bits in the header of each frame than for the sliding window protocol because CLC is based on stop-and-wait.

False; they require the same number of bits in the header

True or False: Two-dimensional parity catches all 1-bit and 2-bit errors, and most 3-bit and 4-bit errors.

False; two-dimensional parity catches ALL 1, 2, and 3-bit errors

What type of framing approach requires stuffing and why is stuffing needed?

Framing with sentinels (for example, HDLC) requires stuffing. Stuffing is needed so that data that uses the same characters/bits as flags in a message are not confused for being actual flags.

How does SONET ensure that there are plenty of transitions in the encoded signal to allow the receiver to recover the sender's clock?

From the book: "To ensure that there are plenty of transitions to allow the receiver to recover the sender's clock, the payload bytes are scrambled. This is done by calculating the exclusive OR (XOR) of the data to be transmitted and by the use of a well-known bit pattern. The bit pattern, which is 127 bits long, has plenty of transitions from 1 to 0, so that XORing it with the transmitted data is likely to yield a signal with enough transitions to enable clock recovery."

Experience shows that Ethernets work well only under lightly loaded conditions. Why don't they work well when more heavily loaded?

Heavy loads lead to frequent collisions and large backoff windows. The extra delays reduce throughput

A host attached to an Ethernet LAN is configured with a single default route in its forwarding table. Assume two routers are connected to this LAN, R1, and R2, and that R1 is the default router for the host. If a destination is reachable through R2, the first time a packet from the host is forwarded it travels to R1, and R1 forwards the packet to R2 (the packet travels on the LAN twice). What mechanism is available with IP to update the host to forward packets for this destination directly to R2?

ICMP redirect message to have the host install a host-specific entry in its forwarding table

A host has one network to a wired LAN, and there are two routers, R1 and R2, that are also attached to this LAN. The forwarding table for the host is configured with a single default router, R1. Suppose a particular destination is reachable through router R2 only (this implies that when R1 receives a packet for this destination it forwards the packet to R2). Describe the mechanism that is available with IP so the host can send packets for this destination to R2.

ICMP-Redirect. R1 sends ICMP to host informing it to add a route to its forwarding table to R2 for this destination

Describe an additional mechanism (besides ARP caching) that is available in IP to reduce the number of frames required for one host to send a sequence of packets to another.

ICMP-redirect. This informs the source host that a better route to the destination host is possible. This allows entries in forwarding tables to be changed.

For the sliding window protocol with a send window size and receive window size each equal to four, what is the smallest number of bits that must be allocated in the frame header for sequence numbers?

If SWS = RWS = 4, bits must cover values up to 2*SWS, or 8. 4 bits are needed to represent this value. SWS <= (MaxSequence +1)/2

For computer networks it is argued that statistical multiplexing is preferable to fixed (or synchronous) time-division multiplexing or frequency-division multiplexing. What is a major disadvantage of statistical multiplexing?

If all message data cannot be sent in one packet, the message is split up into multiple packets which have to be properly reassembled back into the original message when they reach their destination. Packets waiting to be transmitted on a link are also buffered, and buffer overflow can occur.

The 4B/5B encoding approach selects the 5-bit codes so that a codeword has no more than one leading 0 and no more than two trailing 0s. Why is this encoding approach only concerned about consecutive 0s and not consecutive 1s?

If too many 0's, the signal will read low, and this may be interpreted as no signal at all. In addition, the resulting 5-bit codes are transmitted using the NRZI encoding, which already takes care of the issue of consecutive 1's.

In a sliding window protocol with RWS=SWS=5, a very large set of possible sequence numbers (assume no wrapping), and in-order packet arrivals, why can the receiver be assured that it will never again receive the frame with sequence number 10 if it is currently expecting frame 17?

In this case, RWS is [17, 18, 19, 20, 21]. We will never receive a frame with a sequence number less than 17 because in order for the receiver's window to have shifted to expect 17, all sequence numbers below 17 must have been received. In addition, 10 is outside of the window.

What does a host do when it receives an ARP packet from an unknown host to a second unknown host?

It ignores the ARP packet (to avoid cluttering on its ARP table)

What is the purpose of the Type field in the Ethernet frame?

It serves as a demultiplex key. It demultiplexes to higher-level protocols.

What does congestion control do?

Keeps sender from overrunning network

IPv6 provides a 128-bit address space, compared to the 32-bit class-based address space with IPv4. How does IPv6 address scaling concerns given that there are so many more possible address locations?

Like IPv4, IPv6 controls routing table size by using classless address space (ie, CIDR). Careful planning of address prefix assignments is needed.

How does Manchester encoding enable a receiver to synchronize its clock with the transmitter's clock?

Manchester encoding works based on a clock cycle, so by its nature, the receiver will be synchronized with the clock of the transmitter. In Manchester encoding, 0 = low to high, and 1 = high to low.

State an advantage of 4B/5B encoding compared to Manchester encoding.

Manchester has a bit change per clock cycle, or two bits per cycle, and is 50% efficient, while 4B/5B has one extra bit for every 5 bits and uses NRZI, which is 80% efficient.

For the go-back-N protocol with N, the send window size, equal to 8, what is the smallest number of sequence numbers that are needed to ensure reliable delivery of the frames?

Nine sequence numbers are required, as the receiver's window size is one. Also, in the case of go-back-N, MAX = SWS + 1

Under what conditions may a CRC code fail to detect an error?

One condition is when the message's polynomial representation happens to be evenly divisible by the CRC code's polynomial representation. (For example, when a message has an odd bit error but (x+1) is not a factor of the CRC code)

What are the three components of latency?

Propagation, queue, and transmit

Assume router R8 receives a broadcast ARP query that is requesting host H25's MAC address. The other fields in R8's ARP table include H25's IP address and the IP and MAC addresses for the source of the ARP packet, which is host H19. If R8 finds that H25's IP address is in its ARP table but H19's IP address is not in its ARP table, how does R8 update its ARP table and what ARP reply packet does R8 send?

R8 ignores the ARP query and makes no reply. H19 is unknown and therefore is not added to the table. H25 is not the source of the query, so it does not update the time. (H25 is responsible for responding to the query).

A network architecture guides the design and implementation of a network. The architecture consists of layering and protocols, and the objects that make up the layers are called protocols. Each protocol defines two different interfaces. Name the interfaces and briefly describe their purpose.

Service interface: operations on this protocol Peer-to-peer interface: messages exchanged with peer

The traceroute function allows a network engineer to discover to IP addresses for the routers in the path to some destination. What features of IP and ICMP make this capability easy to implement?

Set the TTL field in an IP packet to 1. At first router, it is dropped and ICMP error message is returned to source. Repeat for successively larger hop counts.

A network that employs virtual-circuit forwarding must have a protocol to setup the virtual circuit before it can be utilized. One of the tasks during the setup phase is to assign the local virtual-circuit identifier (VCI) at each switch. Once the virtual circuit is setup, the packets are forwarded based on the VCI. How is the setup packet forwarded given that the local VCI's have not yet been established?

Setup requests are routed like datagrams, using destination-based routing

For a node connected to an Ethernet LAN, CSMA/CD is used to detect a collision and a slotted protocol is used to determine when to schedule a retransmission. If the node has attempted to forward a packet five times and each time it has detected a collision, what is the probability that the node selects the fourth slot for the next forwarding attempt for this packet?

Slot choices for this node range from 0 to (2^N)-1, where N = number of collisions. Thus there are 32 possible slots that this node can choose for retransmission. The probability that the node chooses the 4th slot is 1/32, or 0.031 (about 3%)

In the distributed Spanning Tree Algorithm, why does the root bridge periodically send configuration messages even after the system has stabilized?

So that is a bridge or link fails, the down stream bridges can detect the failure and restart the selection algorithm

Multiprotocol label switching (MPLS) combines some of the properties virtual circuits with the flexibility and robustness of datagrams. Name the two main applications of MPLS that have led to its widespread development.

Support for VPNs and traffic engineering

Suppose two nodes in an Ethernet local area network both have a very large number of frames to transmit. How does the protocol ensure that both nodes have an opportunity to transmit frames (instead of one node gaining access to the network and transmitting all of its frames continuously)?

The Ethernet protocol pauses for 9.6 μs between back-to-back packets. Ethernet also allows nodes to send when they need to, and uses exponential backoff when collisions occur so that nodes can re-transmit.

A host attached to an Ethernet LAN is configured with a single default route in its forwarding table. Assume two routers are connected to this LAN, R1 and R2, and that R1 is the default router for the host. If a destination host is reachable through R2, the first time a packet from the host is forwarded it travels to R1, and R1 forwards the packet to R2 (the packet travels on the LAN twice). What mechanism is available with IP to update the host to forward packets for this destination directly to R2?

The first host mentioned in this problem is the source host. After R1 forwards the packet to R2, it then sends an ICMP redirect message to the source host. The source host installs a host-specific entry in its forwarding table.

What does a host do when it receives an ARP query packet from a host listed in its ARP table that is addressed to another host that is also listed in its ARP table?

The first host updates the timer associated with the second host (source host) (the ARP is proof the source is alive). But it does not update the timer for the third host (the destination host) (since it doesn't know if the destination host is alive). The host does not reply to the ARP query.

Mobile IP is designed to work without any changes to non-mobile host and the majority of routers. Describe how a packet for a mobile host is forwarded from the home agent to the foreign given that none of the routers between the agents have been changed.

The foreign agent sends care-of-address to home agent. The home agent tunnels the packets to the foreign agent.

Mobile IP is designed to work without any changes to non-mobile hosts and the majority of routers. Describe how a packet for a mobile host is forwarded from the home agent to the foreign agent given that none of the routers between these agents have been changed.

The foreign agent sends care-of-address to home agent. The home agent tunnels the packets to the foreign agent.

An organization has a large number of LAN's and one DHCP server. How does a newly booted host contact the DHCP server when the server is not located on the host's LAN?

The host broadcasts a DHCP discovery message on the LAN. A DHCP relay sends a unicast message to the DHCP server.

A host that receives an IP packet that has been fragmented must wait until it collects all of the fragments associated with this packet. What actions does that host take if, after a suitably long period of time, there are still one or more fragments missing?

The host discards the fragments. Host can send an ICMP reassembly failure message, but IP does not handle retransmissions.

A SONET switch can support multiplexing of several low-speed incoming links onto one highspeed outgoing link. What feature of the framing scheme employed by SONET helps simplify the buffering and synchronization tasks required by the multiplexing?

The input links are merged byte wise round-robin into the single outgoing link. A float frame is used to solve synchronization. The payload frame floats within the clock frame and part of the overhead specifies the first byte of the payload.

An Ethernet can be configured with up to 4 repeaters and 5 segments. What limits the maximum length of a segment? What limits the maximum number of segments?

The length of the segment is limited by how long the signal will travel without being re-amplified. The number of segments is limited by how many times the signal can be amplified without being distorted.

A manufacturer of an Ethernet adapter must be assigned a unique 24-bit number. What is the number used for and why is it needed?

The number is the prefix for the node's destination address. It guarantees that all nodes connect to an Ethernet LAN have a unique address.

Describe a problem associated with communicating between heterogeneous architectures (eg, a mixture of Sun and Intel hosts) on a network.

They may store bytes differently. Intel uses little endian, while Sun uses big endian.

Explain the main drawback of the stop-and-wait ARQ algorithm.

This algorithm creates a low throughput as the sender must wait for acknowledgements and/or timeouts.

What is the purpose of the time to live (TTL) field in an IP packet header?

To catch packets that are traveling in loops. Each router decrements this field when forwarding a packet. When it reaches zero, the packet is dropped (an ICMP error message can be sent)

An autonomous system (AS), such as a large corporation may have connections to more than one other AS, but it may refuse to carry transit traffic. With BGP, how does AS implement a policy of not providing transit?

To implement a policy of not providing transit, the BGP speaker doesn't advertise routers to prefixes that are not contained within its AS

For the Ethernet protocol, why must a node wait 9.6 μs between sending back-to-back frames?

When a collision is detected, a minimum of 96 bits will be transmitted (64-bit preamble and 32-bit jamming sequence). 9.6 μs is the time for this to be sent at 10 Mbps. This allows a collision to be detected if there is one.

What does flow control do?

keeps sender from overrunning receiver

The 802.11 standard specifies that a node configured to use the distributed coordination function (DCF) for channel access can use one of two modes: one mode is the MACAW protocol as we studied in class and the other mode allows a node to access the channel without first exchanging RTS and CTS frames. Why is the second mode supported and under what conditions might a node elect to use the second mode instead of MACAW?

maybe for broadcasting? (needed for LAN)

Switching with source routing requires a variable length header. When a packet arrives at a switch, the switch must determine how to read the header. There are a number of possible solutions to handling headers for source routing. Describe one approach.

pg. 188-189, fig. 3.8 - there are three approaches: rotation, stripping, and pointer. - Rotation: each switch rotates the list after it has read its own entry - Stripping: each switch strips the first element from the list as it uses it - Pointer: carry a pointer to the current "next port" entry so that each switch updates the pointer rather than rotating the header

What traditional network class was under the most pressure before classless interdomain routing (CIDR)? Why?

pg. 225 - Class B, because the efficiency of Class B networks are less efficient for relatively small numbers of hosts that are still too large for Class C networks (i.e. say you have 256 hosts; this can not fit into Class C but can fit into Class B, but Class B can hold 65,535 hosts. You would therefore get 0.36% efficiency).

Clemson University's network has a very large number of LAN's but just one DHCP server. How does a newly booted host find the DHCP server?

pg. 233 and fig. 3.24 - the host sends a broadcast DHCPDISCOVER message, which the DHCP relay agent receives. The relay agent then sends a unicast DHCPDISCOVER to the DHCP server

An IP tunnel is a virtual point-to-point link between a pair of routers that are actually separated by an arbitrary number of networks. What information in the IP header indicates to the router at the destination of the tunnel that an IP packet that it has received has arrived through the tunnel?

pg. 238, fig. 3.27

When using distance-vector routing, certain circumstances can prevent the routing tables for the network from stabilizing. What is this problem known as?

pg. 248 - the count to infinity problem

Suppose a dynamic routing algorithm is employed to try and make routing tables correspond to least cost paths. What types of routing metrics are prone to producing load oscillations?

pg. 264-265 - heavy loaded second version of the ARPANET routing algorithm. This algorithm would advertise a high cost for a congested link, causing packets to leave the link, but then advertise a low cost since all of the packets had just left, causing them to flood back. The weight assigned to the link was derived from the average delay experienced by packets recently sent over that link.

How does a BGP speaker know if a route advertised by another speaker contains a loop?

pg. 318-319 - BGP advertises complete paths as an enumerated list of autonomous systems to reach a particular network. In other words, it advertises reachability information. If a provider sees itself in an advertised path, it knows that there will be a loop.

What is the purpose of the CRC field in the Ethernet frame?

to detect bit errors in a frame

What purpose does checksum serve?

to detect errors which may have been introduced during the transmission or storage

Why might someone object to use of pseudo-header?

used for checksum calculation