Multimedia Networking Flashcards
(81 cards)
1
Q
What are the 3 application types of multimedia networking?
A
- Streaming, stored audio, video (i.e netflix youtube)
- Conversational voice/video over IP (i.e skype)
- Streaming live audio and video (i.e sports events)
2
Q
What is streaming?
A
Video or audio can begin playout before downloading entire file
3
Q
What is meant by stored audio in multimedia networking?
A
Audio or video that is stored at a server side, can be transmitted faster than audio/video is rendered aka there is buffering at the client.
4
Q
What is jitter?
A
Variable network delay
5
Q
In streaming if a client’s bit rate has jitter is this an issue?
A
No as there is a buffer so the video plays out at a constant rate to the client
6
Q
Describe how client side buffering and playout interact?
A
- Initial fill of buffer until playout begins at t_p
- Playout begins and continues at a constant rate
- Buffer is filled at variable rate so buffer level varies
7
Q
How does streaming over UDP work?
A
- Server sends at a rate appropriate for the client
- Sent rate = encoding rate = constant rate
8
Q
In streaming over UDP how does congestion affect transmission rate?
A
- It does not
9
Q
What is a normal length for playout delay?
A
2-5 seconds
10
Q
What is the purpose of playout delay and in what kind of multimedia use is it for?
A
It removes network jitter and is used in streaming
11
Q
What can be an issue with the UDP used in streaming services that are not faced by TCP HTTP
A
May not be allowed through firewalls, whereas TCP may be allowed through firewalls
12
Q
How does streaming work with HTTP? (TCP)
A
- Multimedia file retrieved with HTTP GET
- Send at maximum possible rate under TCP
- Playout delay
13
Q
Does the file rate fluctuate in HTTP streaming?
A
Yes it does, due to congestion control and retransmissions which ensure in order delivery
14
Q
What is the difference between playout delay with UDP vs. HTTP (TCP) streaming?
A
- TCP has larger playout delay ensures more smooth TCP delivery rate than UDP
15
Q
In adaptive playout delay what is the equation for packet delay estimation di?
A
di = (1-a)d(i-1) + a(ri - ti)
- di = delay estimate after ith packet
- a = small constant e.g 0.1
- d(i-1) = ith - 1 packet
- ri = time received
- ti = time sent
16
Q
In adaptive playout delay what is the equation for average deviation of delay, vi?
A
vi = (1-b)v(i-1) + b|ri - ti - di|
-vi = average deviation of delay after ith packet
- b = small constant
- v(i-1) = previous average deviation after i-1th packet
- ri = time received
- ti = time sent
- di = packet delay estimation after ith packet
17
Q
In adaptive playout delay what is the equation for the playout time of the first packet in a talk spurt?
A
Playout_time_i = ti + di + Kvi
- ti = time sent
- di = packet delay estimation after ith packet
- vi = average deviation of packet delay value
- K = some constant
18
Q
What is meant by a one-size-fits all, best effort service model? and what is good about it?
A
Increasing link capacity so that congestion does not occur.
- Low complexity of network mechanisms
19
Q
What is an issue with the one-size-fits all service model?
A
- High deployment cost: It is wasteful because the service will not always be busy - aka waste of money
20
Q
What is an alternative to the best effort service model?
A
- Multiple classes of service model
21
Q
What is the multiple classes of service model?
A
- Partition traffic into classes
- Network treats different classes of traffic differently
22
Q
In the multiple classes model what 3 things should be considered to be able to maintain required QoS guarantees?
A
- A service level agreement is necessary (SLA)
- A policing mechanism needs to be in place
- Flows being allowed to use another flow’s unused bandwidth whenever.(SLA
23
Q
What do packet classification and packet marking allow routers to do?
A
- Packet classification allows - router to treat different classes accordingly
- Packet marking allows router to distinguish among packets belonging to different classes of traffic
24
Q
Why are policing mechanisms necessary?
A
- Provide a certain DEGREE of traffic isolation among classes, so that one class is not adversely affected by another traffic class that is misbehaving
25
What is call admission?
- Call admission: a network may deny service if it cannot meet needs
26
What must flows do before they are allowed to join a network?
Declare their QoS requirements because sufficient resources will not always be available
27
What are 3 different types of scheduling policies?
- FIFO (first in first out)
- Priority queueing
- Weighted fair queueing
28
How does priority queueing work?
- Packets are sorted into high and low priority queues (waiting areas) and then packets in the higher priority queue are priorities to send out to departures through the link server
29
How does WFQ work?
- Packets are classified into classes at arrival and each class gets a specified weighted amount of service each cycle
30
How can end-to-end delay be bounded with queues?
By bounding the queue
31
What are the 2 queue manager tasks?
- Remove packets from a queue upon the request of packet scheduler
- Discard and remark packets if queue is full or approaching saturation
32
What is an example mechanisms for discarding and remarking packets when queue is at saturation?
- Random early detection with in/out (RIO)
33
What is the goal of queue management?
- Proactively avoid the queue becoming congested
34
What are 3 packet removal policies? And what do they mean?
- Tail drop (drop arriving packet)
- Priority (drop on a priority basis i.e lowest priority)
- Random: (drop randomly from queue)
35
What is Random Early Detection? (RED)
- Process of dropping packets randomly from selected flows with some drop probability whenever the queue length exceeds given threshold
36
In an example where a queue uses Random Early Detection describe how the queue operates?
- If buffer average is below the minimum threshold - deduce no congestion.
- If buffer is above maximum threshold - high congestion aka drop packets with probability 1
- If buffer is BETWEEN min and max threshold - drop packets with probability p
37
What is the difference between RED and RED In/Out (RIO) and how does RIO work?
- RIO scheme assumes that an edge router has marked packets in the systems conforming to an SLA (service level agreement)
- Packets are marked as either : non-conforming (out of profile) or conforming (in profile)
- Packet out of profile are dropped first if a CORE ROUTER is congested
38
What parameters can be used to mark different packets as in/out of profile?
- things like standard vs. premium service
39
What is the goal of policing mechanisms?
Shape traffic so it does not exceed declared parameters
40
What are 3 commonly used criteria for policing mechanisms? and describe them?
- Average rate : (long term) how many packets can be sent per unit time in the long run
- Peak rate: What is the peak rate packets can be sent at
- Burst size : Max number of packets sent consecutively
41
In the token/leaky bucket scheme what does the peak rate correspond to?
rate that the bucket is filled up (unless the bucket is full)
42
Over an interval of length t what is the maximum number of packets admitted by the leaky bucket scheme? Equation
Total_num = rt + b
r: peak rate that tokens fill bucket
t : length of time measured in
b: number of tokens already in bucket when time started
43
What 2 things can happen when a packet is no met by a token?
- Packet is dropped
- Packet is tagged as a "naughty packet"
44
What happens when you combine token/leaky bucket scheme with WFQ?
You can provide a guaranteed upper bound delay on the service because you can bound the queues
45
If rate that packets are leaving a queue is R and bucket size is b then what is the maximum token rate, r?
max r = b/R
46
What is the fine grained approach for internet QoS? And what is an example of this?
- Providing QoS for individual applications or flows
- I.E INTSERV (integrated services)+ RSVP
47
What is the coarse grained approach for internet QoS and what is an example of this?
- Provide QoS to large classes of data or aggregate traffic
- E.G differentiated services (DS)
48
What is resource reservation protocol (RSVP)
- Network-level signalling protocol operating on top of IP
49
Generally how does resource reservation protocol (RSVP) work?
Hosts request a specific QoS for a particular flow
- Routers provide requested QoS along the path(s)
50
What 4 things are RSVP routers responsible for?
- Dedicating requested resources (e.g bandwidth)
- Maintaining the soft state for signal connection (i.e each flow treated individually)
- Accept/refuse new connections based on available resources
- Guarantee service level agreement (SLA) of existing calls
51
Where can resource reservation protocol be transported (RSVP)?
Inside an IP packet
52
What happens on sender side of resource reservation protocol? (RSVP)
Sender sends a message containing TSpec (flow traffic characeristics) to the receiver
53
What happens on receiver side of resource reservation protocol (RSVP) ?
- Looks at sender's characteristics specification
- Needs to know what path the packet will travel
- Initiate the establishment of resource reservations at EACH ROUTER ON PATH
- Reservations sent back to sender
- If reservation CAN be made a RESV message request is passed to next router all the way back up to sender
- RESV message contains the RSpec aka service request from networks
54
What happens on receiver side of resource reservation protocol (RSVP) ?
- Looks at sender's characteristics specification
- Needs to know what path the packet will travel
- Initiate the establishment of resource reservations at EACH ROUTER ON PATH
- Reservations sent back to sender
- If reservation CAN be made a RESV message request is passed to next router all the way back up to sender
- RESV message contains the RSpec aka service request from networks
55
What are the pros of RSVP - resource reservation protocol?
- It is simple to increase or decrease the level of resource allocation.
56
What are the cons of resource reservation protocol (RSVP) ?
- Requires each flow state information to be STORED on each particular route.
- Each reservation needs some memory
- Each router has to classify police and queue each flow
- admission control decision are needed
57
What is Tspec in integrated services mechanisms?
Describe flow's traffic characteristics - give network info on bandwidth requirements of incoming flow
58
What is RSpec in integrated services mechanisms?
Describe service request - request for bound delay etc.
59
What is call admission control in integrated service mechanisms?
- Looks at TSpec and RSpec and decides on a per-flow basis if the service can be provided.
60
What is resource reservation?
- Bandwidth and buffers are reserved for a data flow
61
What is traffic policing in integrated services mechanisms?
- Per-packet decisions to ensure that all flows conform to their TSPec
62
General description of INSERV (integrated service) ?
- Uses RSVP (resource reservation protocol)
and
- Per flow QoS-sensitive scheduling
(once RSVP reply is sent then management of queues begins)
63
What is difference between differentiated service (DS) and integrated service models (INSERV)?
- DS model is a simpler and more scalable solution
- Per flow service is replaced with per aggregate service
- Complex processing is moved from the core routers to edge routers
64
How is differentiated service (DS) transmitted?
In an IPv6 datagram, there is a DS field which contains a packets class, so the DS field is used to identify per hop behaviour I.E how a router will treat a packet depending on the class it's from
65
Does a differentiated service (DS) need a service level agreement?
Yes. Customer wishing to receive a DS must first have a SLA
66
What does a service level agreement (SLA) include for a differentiated service (DS)? And what is this?
It includes a traffic conditioning agreement (TCA) - it defines the nature of the various traffic classes and services
67
In a differentiated service (DS) what is an edge router responsible for?
- Per-flow aggregate traffic management
- To ensure traffic conformance in respect to TCA (traffic conditioning agreement at the edge-router)
- Marks packets as either in-profile or out-profile (e.g using the leaky bucket)
68
In a differentiated service (DS) what is a core router responsible for?
- Per class traffic management
- Queueing and scheduling based on marking at edge (no soft knowledge)
- Preference is given to in-profile packets over out-profile ones
69
What is meant by per-hop behaviour? PHB
- Routers class differentiation
70
What happens when an edge router in DS does traffic classification?
Directs a packet into various output based on content of packet header (DS, ID number, dest address etc.)
71
What is the role of an edge router as a traffic conditioner in differentiated services?
Metering, marking, shaping and dropping packets
72
What are the different ways that packets can be marked by an edge-router used in differentiated services?
- Class-based marking: packets from different classes marked differently
- Intra-class marking: conforming the (in-profile) portion of flow marked differently than non-conforming (out-profile) one
73
What are the effects of PHBs in core-routers in differentiated services?
- Per hop behaviour
- PHB result in a different measurable forwarding performance behaviour (each forwarding strat has diff performance)
- PHB doesn't specify what mechanisms to use to ensure required PHB performance behaviour
74
What are the 3 different classes of PHB forwarding?
- Expedite forwarding
- Assured forwarding
- Default PHB
75
What is expedite forwarding PHB?
- Low-loss, low-latency, low-jitter (not much randomness), assured bandwidth end to end services
- Limited by speed link of router
- Can be implemented using priority queueing or WFQ
76
What is assured forwarding PHB?
- Delivers the aggregate traffic with high assurance as long as the traffic does not exceed the traffic profile
- Not-intended for low-latency , low-jitter
77
What is default PHB?
Best effort treatement
78
WWhat is link virtualisation?
Viewing the internet as a logical link connecting IP routers
79
What is multiprotocol label switching? (MPLS)
- Integrates layer 2 switching with layer 3 switching
80
What is key feature in a label switching router (LSR)?
- Key feature is separation of CONTROL PLANE (sets up routes) and FORWARDING PLANE (carries information)
81
Why are multiprotocol label switching routers good? (MPLS)
- Improve forwarding performance (simplified lookup)
- Improve scalability (label stacking and merging)
- Provide traffic engineering (through explicit routing)
