application layer Flashcards
process
a program running code within a host on the application layer
encapsulation
the message going through each layer with each one adding its header to it
decapsulation
the message going back up the layers with each one processing its header then removing it
what are the benefits of encapsulation and decapsulation
allows each layer to operate independently
each layer only has to be concerned with its corresponding header
client process
initiates communication
server process
waits to be contacted
why do we need interfaces
so that the layers can communicate with each other
what is the interface used between the application and transport layer
socket
which 2 pieces of information do we need when sending a message
ip address
port number
ip address
network identifier
ipv4; 32
ipv6; 128
port number
transport layer identifier that delivers the message to the right process
why do we need both the ip and port number when sending messages
many processes can be on the same host so we need to specify which one
5 tuple
uniquely identifies a communication between two processes
what do we need to create the 5 tuple
source and destination ip
source and destination port number
transport protocol
how does the client know which port to use
popular applications have well know ports
ephemeral ports are arbitrary numbers automatically assigned by the os and are used for temporary communication
what are some standard api system calls
read()
write()
close()
what are some socket specific calls
send()
recv()
bind()
application layer protocols
defined by a set of messages with their type and when and how processes send and respond to protocols given the value of the field
open protocols
have public documents explaining the messages and fields
e.g. http
propriety protocols
not public and are tied to specific products
e.g. skype
what might an application need form the transport layer
data integrity
timing
throughput
security
tcp service
reliable
provides congestion and flow control
connection oriented
tcp as reliable
guaranteed that messages are delivered in the order that they were sent
tcp and congestion control
can delay or reduce the number of packets sent if theres high congestion
tcp and flow control
stops the sender from overwhelming the receiver when its sending more messages than it can process
tcp as connection oriented
a handshake establishes the connection before real messages are sent
udp service
v minimal
the sender has total control over when the packets are sent therefore they’re transmitted faster
the sender has to explicitly attach the ip and port# and the receiver has to extract them
what are messages called in udp
datagrams
raw sockets
network sockets that allow direct access to lower level protocols
what are some benefits of raw sockets
additional power as you get direct access to network packets
provide access to the network and transport layer header
what is a negative of the client server model
not scalable; when the server gets a lot of client requests you will need to duplicate it as it cannot handle that many
client server model
communicate with the server for recourses
clients don’t have a permanent ip address only an ephemeral one
the server is always on
p2p
any node provides a service
single code executed by all hosts
self scalable
self scalable
the more participants the more scalable as they each provide recourses
in which two ways can layering sometimes be harmful
lower performance due to data and processing overhead
the same process may be duplicated my different layers e.g. error checking
what is the end-to-end principle
don’t implement anything in the network that cant be implemented by the host
how does content sharing not follow the end-to-end principle
content sharing must be on the core network to enable services making content a first class object
what is a webpage made up of
a base html file and intependant referenced objects (can be addressed by url)
why do we use the tcp protocol when requesting web pages
need reliable transfers due to web transfers
how do we use tcp to request and receive web pages
the client initiates the tcp connection
the server accepts it
a http message can be exchanged
tcp connection closed
what are the two types of http connections
non persistent and persistent
non-persistent http connection
only one object can be sent per tcp connection
what is the response time for a non-persistent connection
2 rtt + file transmission time
one rtt to initiate the connection and another for the request
persistent http connection
multiple objects can be sent on a single tcp connection sequentially
concurrent tcp connections
multi-threading that makes receiving resources faster (in parallel)
what is the response time for persistent connections
2 rtt per object + the os overhead for the connection
pipe-lining
requesting multiple objects without waiting for each one to be recieved
what are some http method types
get
post
put
delete
head
http get
requests data from a specific source
http post
sends data to a server to create or update a recourse
http put
http post + uploads the file in the entity body to the path in the url field
http delete
deletes the file in the url field
http head
asks the server to leave the object out of a response
usually for debugging
what is the difference between using post and url to upload an input
post; input uploaded in the server in the body
url; uploaded in the url field which cant be encrypted
what are cookies
a small file with data up to 4kb
maintains the state of your interaction with the browser
what can cookies be used for
authorization
shopping carts
reccomendations
what is web caching
satisfying the client requirements without involving the origin server
how does web caching work
a proxy server receives all browser requests and the cache returns the objects if it has them stored
if not, then the proxy server acts as a client and requests the object from the origin server and returns it to the client
why do we use before caching
can save frequently used large content
what did we do before caching and what were the issues
one server duplicated in the same location and all identifying with the same ip, which server requests went to would be based on round robin, random or load-based
scalability; must support the traffic for many hosts
state; must keep track of previous decisions of all servers
conditional get
used with a proxy server
if it has up to date content then it is sent to the client
if not you get the “304; modified” message then the content is retrieved from the origin server and is sent back to the client
http 2
divides bigger projects into smaller pieces via head of line blocking
head of line blocking
when a client requests a large object amongst smaller ones the large one will prevent the smaller one form being sent until its done so to avoid this the large one is divided into smaller pieces to minimise hol blocking
http 3
adds security to http2 via encryption with https
domain name
a broader identifier that puts together hostnames and organization names and follow an organizational structure
gethostbyname()
library function that gets the ip address for a host name
why do we need the dns
humans cant remember ip addresses but computers prefer them as theyre a fixed 32 bit value
what are the 6 design objectives for the dns system
efficiency
scalability
resilience
security
ownership and control
governance and namespace
namespace
a set of all names that can be used within the system
dns as efficient
name resolution has to happen quicky as you cannot begin communication without it
dns as scalable
needs to handle large volumes of traffic quickly
dns as resilient
has to maintain high availability as if the dns fails then the internet isnt accessible
dns as secure
should return the correct ip as malware practitioners may take advantage of it and direct you to a malicious site
ownership and control in dns
domain name owners should be able to manage the ip-to-hostname mappings of their name; updates and choosing the order of ip
how do we enable loadbearing in dns
the dns server can send a different order of the duplicated server ip and the client can pick the top one which randomises which server is having queries sent to it
governance and namespace in dns
control over domain ownership and registration making sure that some domain names are reserved for specific organizations or companies
subdomains
domain names that end with another domain e.g. lancs.ac.uk
why dont we use a centralised server for the dns system
not resilient as there would be a single point of failure
not scalable as it wouldnt be able to handle large amounts of traffic
would be difficult to maintain a single server
how many root servers are there
13
root server
updated twice a day, has a backup and is replicated across the globe so hosts can access the closest one
tld
top level domain; eg. .com and countries
authoritative dns servers
organizations own servers that can be maintained by them or a service provider
local server
is not strictly in the heirarchy
queries are sent to it and they can act as a proxy that forwards the query into the heirarchy
recursive query
recursively going through the hierarchy until the right ip is found
caching and dns
once any server learns a mapping it is cached and a ttl is attatched to it so that when the site updates the cache also can
what are the 4 types of dns records
type A
type CNAME
type NS
type MX
type A records
hostname to ip mapping
type CNAME
alias to real name mapping
c = canonical
type NS
domain to authoritative name mapping; tells you the server it uses
type MX
name to mail server mapping
when requesting a mail server you are also given the domain so you don’t have to query twice
