Lecture 8: 19th November 2019 Flashcards
Adaptive streaming for multimedia (esp video)
What is adaptive streaming?
!
Why is video streaming omportant?
!
What is the problem with a single video stream for all devices?
!
What is the end-to-end workflow path for video streams?
!
How should streams of stored video files work?
!
What does streaming mean?
!
What does streaming mean with regard to sending video files?
!
What are the challenges of streaming stored video files?
!
How do stored video streams differ from their ideal scenario in practice?
!
How is network delay and jitter compensated for?
!
What is jitter?
!
How does client-side buffering work?
!
What are the mathematical parameters of client-side buffering?
!
How does client-side buffering work, with respect to its mathematical parameters?
!
What is the average fill rate of a client-side buffer?
!
What is the playout rate of a client-side buffer?
!
What is the initial playout delay tradeoff?
!
How does multimedia streaming work with UDP?
!
How does multimedia streaming work with HTTP?
!
What are the issues with HTTP-based streaming in the application layer?
!
What are the issues with HTTP-based streaming in the networks and systems?
!
Why may there be a tradeoff between quality and buffering?
!
How can we avoid a tradeoff between quality and buffering?
!
What are the three steps by which we can avoid a tradeoff between quality and buffering?
!
How does encoding work when avoiding a tradeoff between quality and buffering?
!
What is a bitrate ladder?
!
How does a bitreate ladder work?
!
How does video encoding work with bitrate ladders?
!
What is a manifest file?
!
How are manifest files used with video encoding?
!
What is a CDN?
!
What is rate adaptation?
!
What do servers and clients have to do to perform rate adaptation?
!
What is adaptive streaming known as over HTTP?
!
What is HAS?
!
What is DASH?
!
What measures allow you to perform rate adaptation?
!
What is capacity-based rate adaptation?
!
How does capacity-based rate adaptation work?
!
How easy is it to estimate a video flow’s capacity?
!
How consistent is the end-to-end throughput of video flows using capacity-based rate adaptation?
!
What is buffer-based rate adaptation?
!
How does buffer-based rate adaptation work?
!
How are the rate of the next chunk to be sent and buffer occupancy related in video flows using buffer-based rate adaptation?
!
When is a video flow using buffer-based rate adaptation safe from unnecessary rebuffering?
!
What is fairness?
!
What is QoE?
!
What is QoE fairness?
!
What is QoE fairness in video streams?
!
Why is it difficult to quantify fairness?
!
Why may rate adaptation remove unfairness in video streams unfairness?
!
Why does rate adaptation not remove unfairness in video streams in practice?
!
Which QoS measures suffer when there are multiple competing video streams on a network?
!
What is the difference between the bitrate and transmission rate of a video stream?
!
Why can rate adaptation lead to unfairness with competing video streams?
!
What are the contributing causes of unfairness with cmpeting video streams?
!
What is the fundamental root cause of unfairness with cmpeting video streams?
!
What is feedback control?
!
Why is the transmission rate of TCP like a sawtooth?
!
What is AIMD in TCP?
!
What are the properties of AIMD in TCP?
!
What is the feedback control loop in the network layer in TCP?
!
How does the feedback control loop in the network layer in TCP work?
!
What is the feedback control loop between the network and application in TCP?
!
How does the feedback control loop between the network and application in TCP work?
!
Why is bitrate not a good measure for fairness?
!
How are traditional TCP flows and media streaming flows over TCP different?
!
What are some metrics by which you can assess the quality of a video? Which are objective? Which are subjective?
!
What do metrics to assess video quality not consider?
screen size or resolution, showing it is not important
What is bandwidth utility?
!
How do you find bandwidth utility?
!
How may bandwidth utility be applied to find video stream fairness?
!
What two measures of fairness can you find using the max-min principle?
!
How can you find bitrate max-min fairness?
!
How can you find utility max-min fairness?
!
When is the quality of video on a device optimal?
!
What is pixel density?
!
What is the PPI measure?
!
What is the normalised PPI measure?
!
How can QoE be quantified in a way that includes the relative resolution of the video vs the screen?
!
What is SSIM?
!
What is bandwidth utility?
!
How does the bandwidth available to a device improve the utility/QoE of its video stream flows?
!
How do you find SSIM - process not formula?
!
How can bandwidth-utility and QoE metrics implement fairness?
Find bandwidth on each device for a utility defined by a QoE metric value and then give them to all devices. Choose max obtainable utility value?
What is VHS?
!
How does the use of VHS affect the bitrate of competing video stream flows?
!
How does the use of VHS affect the bitrate of competing batch video download streams?
!
What is CDF?
!
What is JFI?
!
How does the CDF of video stream flows change as JFI increases with and without VHS?
!
How does the CDF of video stream flows change as link utilisation increases with and without VHS?
!
How does VHS affect the bitrate instability of video stream flows?
!
How does the CDF of batch video download flows change as JFI increases with and without VHS?
!
How does the CDF of batch video download flows change as link utilisation increases with and without VHS?
!
How does VHS affect the bitrate instability of batch video download flows?
!
Why does the use of HTTPS/encrypted communications affect adaptive video streaming?
!
What is the difference between session-rate and experience equality?
!
Where should QoS and QoE information generated by client hosts be sent? How should it be handled?
!
Can adaptive video streaming be implemented over HTTPS with client interaction? Why?
!
What is cDVD?
!
What is the architecture of cDVD?
!
How does cDVD work?
!
How does cDVD affect video stream bitrates?
!
Why does QoE fairness matter?
!
What is DASH-S?
!
What are the challenges with implementing DASH-S?
!
How may DASH-S be implemented?
!
What are the key challenges of adaptive video streaming?
!
How could video stream flow fairness return to the TCP-based rate equality model?
!
Why is a fixed network delay unrealistic with video stream flows?
!
How may network conditions change the playout buffer occupancy?
!
What is a common video segment size?
4 seconds
When should video stream flows start playing?
!
Where is Netflix’s CDN present? How?
!
Why is the quality of video streams initially poor(er)?
!
Which host in a video flow controls it? How?
!
How can you crudely estimate throughput from sending a single video segment? What does it not account for?
just assume throughput for all same as just that one
multi path
changing network conditions
changing bitrate
What is the priority in terms of video stream flows - quality or a lack of buffering?
lack of buffering
What’s the overall priority of video stream flows?
To deliver video sections in less than the time taken to send them. Segments usually of playback time 4s.
How can you estimate capacity from throughput?
!
How does receiver buffer occupancy influence the sending video bitrate?
!
Why are the increases in bitrate for increases in buffer occupancy non-linear?
!
How are video segments distributed into HTTP get requests?
Each one forms a separate GET request.
How do clients request video segments to arrive more rapidly or slowly or of a different bitrate?
In HTTP GET request parameters.
How may a large increase in the capacity of a link affect the existing buffered segments in the receiver?
They could be replaced with equivalent higher-quality segments.
Why does an equal bitrate of video streams not imply fairness between them?
!
What is feedback loop?
!
Why is TCP’s ack response system a feedback loop?
!
Why is there a delay between the bottleneck throughput changing and the bitrate of a video flow increasing?
rec buffer lol
What components’ feedback control loops affect the bitrate of video stream flows, and in what order?
application client then TCP (sawtooth)
What do gaps in feedback control graphs indicate with TCP-based video stream flows?
!
Why can simultaneous changes from the feedback control systems in the application and network layers cause issues?
They interfere and can make the flow performance worse rather than better
Why do VPNs use UDP?
TCP in connection to gateway and gateway-gateway would casuse 2x retransmissions in the event of loss
How can you define the utility of data by each bit?
“value” of each bit to the user
How can the viewing device or its properties be known?
HTTP headers or app-level logic
What is bitrate instability?
A measure of how much bitrate is chaging.
Why are QoS mechanisms hard to implement over encrypted channels?
Can’t do DPI on headers as they’re encrypted
Why may client interaction be a good way of implementing adaptive video streaming?
Only they can know (estimate) QoE
What does VHS stand for?
Video Home Sharper
What does cDVD stand for?
Client-Driven Video Delivery
What do max-min algorithms do in general?
Maximise the minimum anyone recieves of some metric or resource
Do max-min algorithms scale?
no
Can max-min algorithms be implemented for QoE QoS in the network core? Why?
No. Not scalable.
The exception to the latter may be in mobile providers. This is because they have a small number of gateways to allow for charging and resource reservations
What does fairness not account for?
Users not being fair in their usage; it only considers resource fairness to each host/flow
