Quality of Service in Packet Networks Flashcards

1
Q

Explain the internet

A

The Internet is an IP switched network that makes use of optical lightpaths for point to point transmission.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Explain packet loss

A

Packet sent by an application does not get to its destination

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

Explain latency

A

The time it takes for a packet to reach its destination

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Explain jitter

A

The variation in the latency of received packets

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Explain capacity

A

How many bits that are packed into packets can be received per second

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

Draw delay as a function of load

A

ipad

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Draw throughput as a function of load

A

ipad

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is latency typically a problem for?

A

voice or video conferencing

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is jitter typically a problem for?

A

Problem for some real time applications that expect a packet within a certain time from the previous, like Voip and video.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What is TCP?

A

Transmission control protocal is one of the main protocols in the IP suite. Provides reliable, connection-oriented communication between devices on a network.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Explain the effects of packet loss

A

Unrecovered packet loss is catastrophic for file transfer (piece of file is missing, some info is lost and might not even open)

Not too bad for voice or video, where you only lose one piece.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the issue with capacity?

A

Capacity relates to all the metrics above (jitter, packet loss, latency). Not enough capacity will increase all of these problems.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Which layer is capacity an issue?

A

Layer 1

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is QoS usually referred to and what does this entail?

A

QoS differentiation (prioritisation)

  1. If we do statistical multiplexing at times things will get congested
  2. When that happens QoS gives priority to pakcets from applications that would suffer most from increase in packet loss, latency and jitter
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

How does prioritisation work

A

Packet heads have a field that indicate priority (Ethernet, IP, MPLS)

Based on these fields, when congestion occurs router/switches can give priority to packets with higher priority values

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Name the QoS tools

A
  1. Policer
  2. Shaper
  3. Classifier
  4. Metering and coloring
  5. Queuing differentiation
  6. Scheduler
  7. Rewrite
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Explain the classifier

A

Inspects a packet (for its incoming port, address, COS marking) and assigns it a COS class inside the router

If a packet has COS marking three actions are possible:

  1. Trust it and use it for its internal router operations
  2. Increase its granularity (for example provide more in depth differentiation)
  3. Change it completely
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Explain metering and colouring

A

Customer have a contract with an operator through a Service Level Agreement

For example, have a certain capacity of:

  1. 5 Gb/s of committed information rate (must be satisfied)
  2. 8 Gb/s of peak information rate (is satisfied if capacity is available)
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

Explain Two rate Three Colour Marking

A

There is colour blind and colour aware:

ipad q

20
Q

What is a policer?

A

A policer performs an action on packets based on data rate threshold limit

e.g. Any traffic above a certain rate is automatically dropped, typically performed at the ingress of a router

21
Q

What is a shaper

A

A shaperperforms an action on packets based on data rate threshold limit

HOWEVER, the action is to delay, so that the threshold is not exceeded

22
Q

Explain the policers implementation

A
  1. Policer implementation:
  • The bucket gets filled with tokens at a constant rate
  • If there are enough credits (tokens), then it enters one of the queues
  • If there aren’t enough credits (the bucket is empty), then the packet gets discarded.
23
Q

In the policer implementation, what does the depth of the bucket tell us?

A

It determines the maximum number of credits that can be stored at any one time -> determined the burst size limit

24
Q

What is the burst size limit?

A

The burst size allows for short periods of traffic bursting.

25
Q

Explain the leaky bucket algorithm

A
  1. Incoming packets are stored that accepts new packets at whatever rate they come in
  2. The bugger is the depleted at a constant rate
  3. The rate at which packets taken from the buffer determines the bandwidth limit
26
Q

What does the size of the bucket (buffer) tell you?

A

Determines for how long excess traffic can be sent before the bugger overflows and starts discarding incoming packets.

The ability to keep the traffic rather than discarding it is at the expense of increased latency.

27
Q

Queues work in what mode?

A

Queues work in FIFO mode and packets proceed linearly from input to output

  • Packets cannot pass other packets once they’re in the queue
28
Q

Can packets be dropped before entering a queue?

A

Yes, it’s based on the queue fill level.

  1. If a queue is 100% full, then all new packets towards that queue are dropped (no physical place to store it)
  2. Methods exist to drop packets before a queue is 100% full
29
Q

How many ports does a router have

A

8 different queues per port -> can identify up to 8 different classes, DSCP up to 64

30
Q

Explain the Scheduler

A

It’s associated with the use of differentiated queues, as it implements an algorithm to decide how to serve the queues.

31
Q

Explain a simple scheduler

A

Serve first all packets in the queue with highest priority, then the queue with second higher priority

32
Q

Explain Rewrite and why we need it

A

Rewrite changes the COS marking of an incoming packet

One straightforward reason is if the router does not trust the entity that has marked the packet.

  • It carries out its own classification and assigns a new marking.
33
Q

Explain how rewrite can be used to signal information to a downstream router

A
  1. Two packets, A and B, arrive with a trusted marking of ‘1’
  2. A -> green, B -> yellow
  3. Router ‘rewrites’ the marking of B from 1 to 2 to signal the downstream router that packet B had exceeded the CIR rate
  4. This implies that both router have a similar interpretation of the COS marking
34
Q

Explain the effects of QoS on packet loss, latency and jitter

A

Packet loss is caused by -> traffic policer, drop by the queue

Latency is caused by -> shaper, scheduler

Jitter is caused by -> shaper, scheduler

35
Q

Explain the effect of queue size on delay and jitter

A

A longer queue is able to absorb heavier traffic bursts:

  1. Lower chance of queue filling up and traffic being dropped
  2. However, packets spend longer in the queue and thus the delay increases (and so the jitter)
  • For some applications it makes sense to drop a packet rather than having too high a delay.
36
Q

Name some schedulers implementation

A
  1. FIFO queuing
  2. Fair queuing
  3. Round robin
37
Q

Explain FIFO and FQ

A

FIFO is a simple queue with no service differentiation.

Fair queuing:
- It isolates different flows into different logical queues
- The scheduler picks a packet from the logical queue in round robin.

38
Q

Explain Strict priority

A

Strict priority queuing has multiple queues of different priority and packets are taken first from the highest priority queue, then from second highest, and so on

39
Q

Explain weighted fair queueing

A

WFQ assigns proportional weights rather than giving strict priority to queues.

  • Q1 is served less often than Q0, but it doesn’t have to wait until Q0 is empty
  • It also operates bit-by-bit to avoid that flows with larger packets get to transmit more data.
40
Q

Explain weighted round robin

A

Similar to WFQ, as it assigns different weights to different queues, but it considers packets rather than bits

  • Less precise in terms of respecting the weights, but OK in average
  • Less computationally intensive than WFQ
41
Q

Explain deficit weight round robin

A

Tries to bring back fairness regarding packet size to WRR without having to calculate the exact size:

  • Uses tokens which are assigned to queues proportionally to the weight at every turn
  • When a packet transmits it consumes token (quantums), in an amount that is proportional to the packet size in bytes
  • If not enough quantums are available the queue is skipped until the next round

-

42
Q

Explain priority based deficit weighted round robin

A

Adds stricter priority for services that need very low delay and jitter

  • some queues served strictly before others while other queues follow DWRR
43
Q

What is the issue with PB-DWRR?

A

Issue if high-priority queue monopolise the capacity

44
Q

Explain what RED is?

A

Random Early Discard:

  • RED drops packets with a certain probability as the queue starts filling up.
45
Q

Explain what WRED is

A

Weighted Random Early Discard allow different drop profiles for different type of traffic

  • Gives lower priority traffic a higher drop percentage