Data Link Layer Flashcards
Discuss pure Aloha
Pure ALOHA lets users transmit upstream frames whenever they like, reducing latency. However, this results in collisions making the frames garbage. When the checksum isn’t correct at the receiving side, the packet won’t be retransmitted via the downstream. The user waits a random time interval to retransmit, because otherwise synchronous retransmissions will keep colliding. To fix this issue, they came up with Slotted Aloha. Users can only send a frame at the beginning of a time slot. Now collision only occurs when two frames are sent in the same time slot, improving performance.
Discuss hidden and exposed terminals
When terminals are hidden from eachother, they won’t hear eachothers transmissions. Therefore, they can transmit at the same time resulting in collisions.
When terminals are exposed to eachother, one can back off its transmission because the other terminal started a transmission. It is however possible that the terminal backed off even thought the two terminals wanted to transmit a frame to different destinations.
Discuss non-persistent CSMA
To avoid the hidden and exposed terminals problem, we only sample the channel when data is ready. If the channel is idle, the data is transmitted. If the channel is already in use, the sender waits a random period of time and tries again. If a collision occurs the node waits a random time and retransmits.
Discuss p-persistent CSMA
Sample the channel when data is ready. If the channel is idle, there is a probability of p to transmit the data. If no ACK is received, the frame is retransmitted with a probability of 1-p.
Discuss BMAC
BMAC has no time-synchronization, so joining the network is quick and low-cost. Transmits of the senders are random, so we spend a little more energy when transmitting (= Low Power Listening). Nodes listen for short sample periods. If traffic is detected, the nodes stay active and listen for incoming packets. Long preamables are added to the messages so that the node is ready to receive the packet when it arrives. The length of the preamble must be greater than the length of the sampling period.
Discuss TSMP
Time Synchronized Mesh Protocol uses synchronized time slots to transmit frames. This is done with a centralised clock using Time Division Medium Access. This centralised clock is bad for resilience and because of load centralization. The collisions are minimized because the time slots aren’t overlapping. Less collisions means less retransmissions, saving a lot of power. However, a bad fine grained clock reduces performance. That’s why a guard time is added to compensate for this problem.
When a node joins the network it listens to transmissions so it can learn when the node itself can transmit data without colliding. The nodes can hop through different frequency channels. Packets are acknowledged at each hop to reduce overhead. Extra redundancy is added through spatial and temporal redundancy. The synchronized approach results in high latency for sporadic traffic.
What’s the problem with classic Ethernet?
All devices are connected to the same Ether cable. If something goes wrong in the cable, the entire system shuts down.
What is the preamble used for in the Ethernet frame?
The preamble is used to synchronize the sender and receiver clocks.
What does the padding do in Ethernet frames?
Padding is required to ensure collision detection for shorter frames.
Why is the Ethernet max frame size 1500 bytes?
DIX uses a length field, while IEEE 802.3 uses a type field. All the type values are greater than 1500. This means that if the field is smaller than 1500, it denotes a size. Otherwise, it denotes a type.
Why is a minimum frame size required for Ethernet?
Distinguishing valid packets from failed chunks.
Distinguishing packets from noise bursts from collisions.
Small packets can be sent across a cable before collision is detected.
What is exponential back-off?
If collision occurs, Ethernet uses exponential slotted back-off. Each station waits a random number between 0 and 2^i-1, with i the amount of collisions of the station.
What’s the problem with persistent channel sampling?
Everyone can ready all the traffic and see who the senders and the receivers are.
What problem of classic Ethernet do hubs solve?
A hub is a place where different links come together through different ports. If one link is broken, the other ones still function.
Why would we use Switched Ethernet?
A switch analyzes MAC addresses of frames, sending them to their corresponding port via the right link. This saves a lot of bandwidth and improves the performance.
