Basics Flashcards
What does LAN stand for?
Local Area Network
What is the characteristic of best effort delivery?
No notification about packet receipt
Name two characteristcs of ethernet broadcast
- All transeivers receive every transmission
- Host interface filters among packets those intended for them.
What are the 4 layers of TCP/IP?
- Link
- Internet
- Transport
- Application
What does TCP stand for?
Transmission Control Protocol
What does CSMA stand for?
Carrier Sense Multiple Access
Explain CSMA
– multiple machines can access the Ethernet simultaneously
– each machine determines whether the “ether” is free by
sensing carrier wave propagation
– each transmission is limited in duration and needs a
minimum time between transmissions to prevent
monopolization of the network
What is CD and explain its purpuse in data networks.
Collision Detection
each transceiver monitors the cable while transmitting to search for foreign signal interferences
Explain Non-persistent CSMA
– Sensing is not continuously but repeated after random time.
– If no collisions are sensed, the station sends a packet.
Explain p-Persistent CSMA
Continuously sensing, but sending of packet probabilistically:
send in current time slot with probability p and in another
time slot with probability 1-p
What type of scheme does ethernet have?
1-persistent CSMA/CD scheme
What does the IP specify?
Internet Protocol:
– specification of basic unit of transfer (packet)
– IP software handles the routing functionality
– IP includes the rules for unreliable,
connectionless, best-effort delivery
How long is the IP adress of IPv4?
32 bits / 4 bytes
How large is the IPv4 header?
6x 32 bits
How is the fragmentation counter updated per IPv4 packet?
It contains the offset of the data in octets.
Ex. 320 bits in the first packet means 80 octets, so the offset for the second packet is 80
Why does IPv6 extended its adress size?
More digital devices calls for more adresses.
Why does IPv6 no longer provide checksum?
Speed up: data integrity is checked in other layers already
Explain why Header options are ommited in IPv6?
Because headers can now be appended using the “Next Header” option, which defines that there will be more header information before the packet data.
What makes IPv6 easier to manage than IPv4?
Auto configuration of Adresses, as well as more clearer structuring of subnets (using predefined adresses).
How can shorten IPv6 adresses?
You skip leading 0’s and replace large range of 0’s with ::
Ex. 2010:00A1:0000:0004= 2010:A1::4
this example is shorter than standard size
How is the global unicast adress composed ususually (for IPv6) ?
– First 64 bits: Global Routing Prefix (n bits) + Subnet-ID
(64 - n bits)
– Last 64 bits: Interface-ID
– Interface-ID often generated using the interface’s MAC
address
Explain what neighbor discovery is and how it works.
Every IPv6 interface has a link-local address that identifies it and
is used for link-local communication (e.g., for Neighbor
Discovery)
* Neighbor Discovery: IPv6 nodes can find systems they are
physically connected to (other nodes and routers)
– Routers send periodic Router Advertisements (RAs)
– Hosts can also solicit RAs by broadcasting a Router Solicitation
– RAs contain Global Routing Prefix and Subnet-IDs that hosts need for
generating global unicast IP addresses
* Duplicate Address Detection automatically resolves
address conflicts
Dia Dump
Give two examples of Distanc Vector protocols
RIP + Bellman-Ford
Give two examples of Link-state protocols
OSPF or Dijkstra
In Link-State Protocols, who constructs the forward table?
The local routing algorithm.
In Distance Vector Protocols, who constructs the forward table?
The protocol itself.
What kind of routing information does a Distance-vector protocol maintain?
maintain list of shortest distances
What kind of routing information does a Link-state protocol maintain?
maintain entire map of network
What kind of information does the Distance-vector protocol exchange with its peers?
a list of distances with neighbors
What kind of information does the link-state protocol exchange with its peers?
flood topology information
Which is more complex Link-state or distance-vector protocols?
Link state
Which is more vulnerable Link-state or distance-vector protocols?
distance vector
In routing hierarchy, explain what the intra-domain routes?
Routing within
one Autonomonous Systems (e.g., OSPF or IS-IS)
In routing hierarchy, explain what the inter-domain routes?
Glues together
different Autonomonous Systems (e.g., BGP)
What does BGP stand for?
Border Gateway Protocol
What are the 3 properties of BGP?
– glue between ASs
– complex protocol
– only unicas
In what way is BGP enhanced comapered to standard dinstance vector protocols?
Path vector contains entire path (list of ASs)
Explain how BGP decides best path:
- H receives the path vector from its BGP-neighbours about K:
from B: BHIJK from G: GHDK
from D: DK from I : IJK - H discards the path from B and from G that runs over itself
- H uses a module with policy information and computes a “distance” for each remaining path from H to K
- The path corresponding to ‘best’ distance is stored in H’s
routing table
What does MPLS stand for?
Multi Protocol Label Switching
What are MPLS use cases?
- traffic engineering
- scalable IP virtual private networks
Does TCP send acknowledgements?
Yes, the receiver does
In what layer is TCP applied?
Transport layer
What are the three ways TCP handles errors or packet loss after detection?
Retransmission: for rare errors or when time permits.
Forward Error Correction: For frequent erros or when short on time.
Discard: When strict reliablity is not required or able.
What is RTO and how is it applied in TCP?
Retransmission Timeout is maximum time between transmission and ACK before the transmittor retries the transmission. The value of RTO is updated during runtime in order to adapt for performance in the system.
Explain how duplicate acknowledgements work
When ordered data is sent, the receiver responds with the ACK N+1 (the next expected packet). When a packet is lost, and an out of order packet is receieved, then the receiver keeps sending dublicate ACKs. When 3 dublicate ACKS (so 4 total) are received by the sender, it resends the packet that has been lost.
After the missing packet has been received by the receiver, it Acknowledges with the updated counter of prereceived packages. (so packets in buffer are read and processed before ACK is send).
ACK 45 -> ACK 50
How does TCP implement flow control?
Using (sliding) receive window: it limits the packets that are ‘inflight’ such that the receiver is not overwhelmed. It does this by pacing the ACKs of the receiver.
How does TCP implement congestion control?
Using (sliding) congestion window: it limits the packets that are ‘inflight’ such that the network is not overwhelmed. It does this by implementing a Congestion Control Algorithm
What is CCA?
Congestion Control Algorithms
In TCP, what is refered to using cwnd?
Congestion Window (for congestion control)
What is TCP Reno responsible for?
Congestion Control in TCP.
What type of congestion avoidance does TCP Reno have?
Additive Increase, Multiplicative
Decrease (AIMD)
Explain the phases of TCP Reno (in detail)
- Slow start: For every ACK received, cwnd +=1. Until above allowed threshold or packetloss -> go to step 2.
- Packet Loss. Set ssthreshold to inflight/2. In case of heavy congestion: Go back to phase 1. In case of mil congestion, go to phase 3.
- Fast Recovery. Set cwnd to ssthresh + 3 and retransmit lost packets. If another dub. ACKs. occur, set to ssthresh +1. Go to step 4.
- Congestion Avoidence. Linearly increase the cwnd by one each round, untill packet loss or erro occurs.
What problem does TCP Reno have?
TCP Reno’s linear increase is too slow in networks with high bandwith or long RTTs.
What is TCP Cubic?
Another congestion control part of TCP, but with faster increase than TCP Reno.
What is the problem for loss-based CCAs and small buffers?
low throughput: CCA cuts cwnd too often
What is bufferfloating?
When buffers are too large: high queueing delay.
What is BBR?
Bottleneck Bandwith and Routrip propegation time.
What is the core idea of BBR?
adjust behavior based on a model of the
network path
In BBR, what is BDP and how is it calculated?
Bandwith Delay Product = bottleneck (max) bandwith x RTT
What is the optimal operating point for BBR?
1 Bandwidth-Delay Product
(BDP) of data is inflight
How does BBR scale during runtime?
- Regularly raise sending rate to probe for bandwidth
- Regularly lower sending rate to obtain samples of the RTT
Why does BBR not use ACKs?
self-clocking through ACKs no
longer possible nowadays
– DOCSIS or WiFi aggregate ACKs and send them out in a burst (“delayed ACKs”)
What are some big problems with BBR?
- fairness issues
- high queueing delay
- no reaction to packet loss
How are classic buffers usually managed. And what are the consequences?
Tail drop It drops the latest packets when buffer is full.
Leads to higher queueing delay and specific flows being targeted.
Concurrent packets usually belong to the same flow, who are dropped
What is AQM and how is it used?
Active Queue Management.
it manages the buffer by randomly select and dropping packets before the buffer is full.
Is AQM faster than classic buffers?
Yes, certain flows are no longer targeted.
What is RED and how does it work?
Random Early Detection introduces a buffer min and max capacity. It then drops packets between this threshold with probability p, as a prepocaution. p is increased with the queue’s occupancy.
What is CoDel and how does it work?
Controlled Delay.
Measure the sojourn time of packets in buffer.
After an interval of 100 ms, if minimum sojourn time was over 5 ms, then drop a packet and shorten interval with sqrt of times interval was shortened. If min sojourn time was under 5 ms, reset back to 100 ms.
What is ECN and how does it work?
Excplicit Congestion Notification. Routers mark packets with ‘congestion experienced’ (CE) such that it is received by the receiver. The receiver marks an ACK with an echo of this message (ECE) and sends it to the transmitter. The transmittor cwnd is reduced (CWR), and marks the next packet with this info.
What system is a prerequicit for ECN?
AQM
What is UDP?
User Data Protocol
What does QUIC stand for?
Quick UDP Internet Connections
What are the goals of QUIC?
– Low latency
– Security
– Multiplexing without head-of-line blocking
– Connection migration
Where does QUIC implement congestion control?
Application layer
Does QUIC send an acknowledge?
No, recievers only respond to initial message.
What is multiplexing and what is a major disadvantage with TCP?
Multiple data streams
within one connection to access multiple endpoints at once.
If there is an error in one stream, then TCP blocks all other streams until this is resolved (major delay)
How does QUIC implement multiplexing?
Using designated UDPs per stream, then streams are independent.
How does QUIC solve connection migration?
packets contain connection ID, so it can reestablish connection fast. (In TCP connection just breaks and a workaround is needed)
