Transcript Streaming and Content Delivery

Streaming and
Content Delivery
SECTIONS 7.4 AND 7.5
Digital Audio

Audio waves are converted to a digital format by an ADC (Analog-to-Digital
Converter) by taking electrical voltage as input and generates a binary number
as output.

Reserve is to convert digital formats with DAC, so that sound can be projected
by an audio device.

Problem with this is that doing so creates high transfer times and uses up large
amounts of bandwidth.

In order to deal with this, we compress the audio files.

Compressing relies on encoding and decoding files for less intensive transfer.

Encoding is needed once whereas decoding is needed many times, so
encoding is allowed to take longer.

Waveform encoding – transforms signals mathematically. Perceptual encoding
– abuses certain flaws in the human auditory system.

Need to use different algorithms for live audio, otherwise we experience
interrupts
Digital Video

Easiest way of having video is with frames that consist of rectangular grids of
pixels – only allows for black and white (no shades of grey)

Using 8 bits allows us to use shades of gray for the pixels.

24 bits of color for the pixels allows us to produce 16 million different colors –
more than we can even see.

For TV’s and computer monitors, pixels are instead made of red, green, and
blue subpixels.

For television, which broadcasts at 25 frames/sec, motion isn’t steady, so
instead they use two fields, one which is broadcast on even numbers, and
the other broadcasting on odd numbers, a process known as interlacing.

Needs to be compressed like audio, done so using the MPEG standard.
Streaming

Rather than loading videos entirely before they are loaded, we use metafiles, things that
point to where we can access the video from.

Video players have jobs to do – mainly dealing with errors, decompressing, and eliminating
jitter.

For error handling, we attach parity packets to the end of our information, which
duplicates the data.

Alternate approach is interleaving, where we mix up the media and descramble it at the
end, so that if things go missing, its over a period of time rather than all at once, and hence
less noticeable.

To help eliminate jitter, we simply use a playout buffer – simply, we load some of the video
before we actually start playing it, rather than having to load and immediately display the
same data.

Livestreaming is different – data is sent as its being created, so buffer has to be larger so
things can be processed, in a way similar to the playout buffer.

Since livestreaming has many people accessing the same data, we use a method called
multicasting, where each viewer is sent a packet to a group address, with information for
audio/video being sent to this group address rather than many separate ones.
Traffic, Farms, and Proxies

Before 1994, most traffic was FTP, then became P2P(peer-to-peer) around 2000.

Traffic is highly skewed, small amounts of websites have large amounts of traffic, while
large amounts of websites have small amounts of traffic.

Server farm – collection of computer servers that provide functionality far beyond that of a
single machine.

Main problem with server farms is that the set of computers must look like a single logical
website to clients.

With server farms, we map using load balance to evenly distribute the workload.

Proxy server – Computer that acts as mediator between a web browser and the Internet.

Web proxies are used to fetch web requests on behalf of users, and are more beneficial for
popular requests.

Proxies are popular because they can also be used to filter content.
CDN’s
CDN – System of distributed servers that deliver
web content based on the location of the user.
 CDN’s use a structure that has three levels, an
origin server, a node, and worldwide clients.
 To support a CDN, we can use either a web
proxy, mirroring, or DNS redirection.
 CDN’s themselves can be used to support
dynamic page creation, images, videos, and so
on.

Peer-To-Peer Networks

Basic idea of Peer-to-Peer file sharing is that many computers come
together and pool resources to form a content distribution system.

Computers in this system do not need to be intense, and are
commonly home computers.

Peer-to-Peer networks are self-scaling, meaning their usable upload
capacity grows in tandem with download demands from their users.

Commonly used because CDN’s have large amounts of personal
information about many people, which can then be used to
provide a better service.