Components in the end-to-end delay: Transmission Difference between mail from in SMTP and from in

delay, propagation delay, and queuing delay. Since the mail message itself: mail from is a command
we have fixed links from the source to the from the SMTP mail server, whereas from: is a
destination is fixed. Therefore, transmission and header in a mail message which informs the user
propagation are fixed while queuing varies due to agent how to render an email message.
the variation in the packet data arriving at the
network. Videos have both spatial and temporal redundancy.
Spatial redundancy refers to the fact that adjacent
(CND) improves the response time for the user. Web pixels in a single image of a video may be very similar
caching reduces overall delays due to two reasons. (e.g., similar color or luminance), hence it is possible
One is because the caches are closer to the user than to avoid sending repeated data for each pixel.
the origin server of content, so serving the content Temporal redundancy refers to the fact that frames
from the caches incurs a smaller propagation delay. adjacent in (playout) time may be very similar, due
The second reason is that they might have a high to limited motion in time between the two images.
bandwidth path available between the cache and Once again, it will be possible to reduce the amount
user compared to the user and origin server in this of information transmitted or stored by only
propagation delay to get objects from the cache are recording and transmitting the differences between
lower. Web caching is not useful for a dynamic or the frames.
personalized object that may not be stored
independently of the origin server.
There are four types of physical transmission on a The DNS hierarchy works its way down till it reaches
link: a name server with a mapping for the requested
Binary signal name.
Amplitude modulation
Frequency modulation With replication, a single set of domain names can
Phase modulation be served by multiple replica servers, each
The HTTP method types Get: the resources specified holding identical entries for the requested set of
in the requested URL domain names. Such load spreading helps the
Post: send entities to a data-handling process at the DNS system scale to a larger query load.
requested URL
Head: asks the server to leave the requested object Caching helps to reduce query load. Caching allows
out of response, but sends the rest of the request the DNS to store the IP address mapping
Put: update a resource at the requested with a new temporarily. As a result, more queries aren’t needed
entity specified in the entity-body for the same IP address until the cache entries time
Delete: deletes the specified in the URL out. This reduces query load.

The difference between post and put the URL of the The identification field helps relate the DNS response
request identifies the resource that processes the to a query sent out earlier.
entity-body while put contains in the entity-body.
An authoritative DNS server holds the “true”
GET: return the requested resource in the entity- mapping from a name to a value, e.g., the IP
body of the response along with response headers the address corresponding to a domain name. This
while HEAD: return all the response headers in the value is known to be always up to date.
GET response, but without the resource in the
entity-body Internet software and hardware are arranged in
layers because they provide modularity, and each
DNS helps the internet not just be convenient but layer performs a specific function and interface to
also work efficiently and loads whatever content the the layer above and below it. The functionality of
user inputs quickly. The problem that the DNS solves each layer is implemented in the protocols.
is that it turns human-readable addresses into IP
addresses. The client-side buffer is for downloaded videos to
absorb the variation in the network conditions. The
buffer helps the user to pause, fast forward, rewind Circuit switching is a good choice here,
or even jump through the video.  because once the network allocates the resources, it
will be available for the connection until it
Traditional applications like HTTP and SMTP have terminates. Further, since there are just a few pairs
delay tolerant but not loss tolerant and there you of communicating endpoints, it will be feasible
could only use the data once it’s fully transferred the to uphold all such reservations without running out
multimedia application are often real-time because of bandwidth resources.
you can watch or stream the video.
Many caches evict old items based on recency of use
Manifest file: it is a file that the user requests that is (i.e., evict the least recently used items (LRU)).
listed will all the different versions of the Assuming the DNS cache is limited in space, the
video(audio language, caption, etc.), and after longevity of a domain in the cache indicates how
choosing a version while a video is being watched recent and frequent accesses to this domain are.
does the header of the HTTP request have the Another possibility is using the frequency with which
version the user wants and according to that, the the domain reappears in the cache after a possible
server sends. request line (not the headers) of the eviction. Picking the highest longevity or the most
HTTP request contains the URL of the object frequent re-appearances are two methods that
corresponding to the specific version of the video or could provide an idea of the most popular domains
audio object the client wants. That is, the request accessed by the users
doesn’t specify the “version” (i.e.: give me “1080p in the department.
video”), it directly asks for a specific object
corresponding to that version One may conclude that a given domain was recently
accessed by another user if the time to obtain a DNS
There are two types of transport: response for a given domain is closer to Tl than to
Transmission Control Protocol (TCP) Ta, i.e., the local resolver cached the most recent
Connection-based: the application remembers the user’s response, and the item is still in the cache.
other process talking to it. Suitable for longer-term,
contextual data transfers, like HTTP, file transfers,
etc. Guarantees: reliability, ordering, congestion
control

User Datagram protocol

Connectionless: the app doesn’t remember the last
process or source that talked to it. Suitable for
single req/resp flows, like DNS. Guarantees: basic
error detection

A server is an “always-on” machine, that is supposed

to always be available to provide a ser-
vice. It is widely reachable from different parts of the
Internet. A client is a machine that may
come on intermittently and interacts with a server as
a “customer” of the service it offers. Client
machines may not always be reachable from large
parts of the Internet.

Delays:
Propagation delay: time to move one bit across.
Transmission delay: time from 1st bit @ sender to
last bit @ sender
Queuing delay: time that packet waits for
transmission