Overview of the Layers
The layers collectively are often referred to as the protocol stack.
3.1.1 Physical Layer
This is concerned with the nature of the physical media (metal or optical cable, free-space microwave, etc.)
used to send signals, the nature of the signals themselves, and so on.
There is also the question of signal form; the signals themselves may be in the form of pure 0-1 bits, or may
be in the form of certain frequencies. In addition there are questions concerning how a receiver distinguishes
two bits which are adjacent in time.A major issue is the form of the medium, both in terms of the materials it uses and its topology. A basicwired Ethernet, for example, consists of cable conducting electrical signals; the connections could also be wireless. More complicated networks, including Ethernets, may consist of more than one cable, with all of them connected via a hub. The latter has become common even at the household level.
3.1.2 Data Link Layer
For example, in an Ethernet, this layer is concerned with ensuring that two network stations connected to
the same cable do not try to access the line at the same time.4 For this reason the Ethernet operation is an
example of what is called a Medium Access Control (MAC) Protocol.5
Here is an overview of how the Ethernet MAC protocol works, using a “listen before talk” approach. When
a network node has a message ready to send, it first senses the cable to see if any node is currently sending.
If so, it generates a random backoff time, waiting this amount of time before trying again. If the node does
not “hear” any other node sending, it will go ahead and send.
There is a small chance that another node actually had been sending but due to signal propagation delay the
transmission had not yet reached the first node. In that case a collision will occur, destroying both messages.
Both nodes will sense the collision, and again wait random amounts of time before trying again.
This layer also does the setting up of frames of bits (i.e. sets of consecutive bits sent along the wire), which
not only include the message itself but also information such as (say, in the Ethernet case) the Ethernet ID
number of the destination machine.Messages may be broken up into pieces before being sent. This may be handled at the transport level (seebelow), but may also be done at the data link level.
3.1.3 Network Layer
This is the routing layer. Questions addressed in this layer include: If in our example above saturn wants
to send a message to holstein, how is that accomplished? Obviously its first step is to send the message to
mars; how does saturn know this? How can alternate routes be found if traffic congestion occurs
3.1.4 Transport Layer
Suppose saturn’s message to holstein consists of a large file transfer, say 100 megabytes. This transfer will
take a long time (by network standards), and we certainly don’t want it to monopolize the network during
4In a small Ethernet, we would not even have two nodes on one cable; we would just have each station connected to a different port in the hub.And the address of an Ethernet or other LAN card is known as the MAC address of the card.This phrasing implies that the frame will be sent only to another machine on the same Ethernet. This is true, but it may be that the frame’s ultimate destination is on another LAN, and the current Ethernet destination ID is for a machine which plays the role of a intermediary router to other LANs. See the description of the Network layer below. that time. We also must deal with the fact that the buffer space at holstein won’t be large enough to deal with a 100-megabyte message. Also, one 100-megabyte message would have a sizable probability of havingat least one bit in error, and if so, we would have to retransmit the entire message!So, the file transfer must be done in pieces. But we don’t want to burden the user at saturn with the task of breaking up the 100 megabytes into pieces, nor do we want to burden the user at holstein with the reassembly of the messages. Instead, the network software (again, typically in the OS) should provide these services,which it does at the transport layer, as for example is the case with TCP.
3.1.5 Session Layer
This layer is concerned with management of a session, i.e. the duration of a connection between two network
nodes. The word connection here does not mean something physical, but rather refers to an agreement
between two nodes that some chunks of data with some relation to each other will be exchanged for some
time. Actually, TCP does this in some senses, as does the socket interface to TCP, which is very much like
the interfaces for reading or writing a file (described in more detail later).
3.1.6 Presentation Layer
This layer deals with such matters as translating between character codes, if the source uses one and the
destination the other. In the old days, this could mean ASCII at one end and EBCDIC on the other end.
Today, though, it could mean for example two different coding systems for Chinese characters, Big 5 and
GB.
3.1.7 Application Layer
You can write programs at the application layer yourself, and of course you use many programs written by
others, such as ftp, Web browsers, e-mail utilities, and so on.
The layers collectively are often referred to as the protocol stack.
3.1.1 Physical Layer
This is concerned with the nature of the physical media (metal or optical cable, free-space microwave, etc.)
used to send signals, the nature of the signals themselves, and so on.
There is also the question of signal form; the signals themselves may be in the form of pure 0-1 bits, or may
be in the form of certain frequencies. In addition there are questions concerning how a receiver distinguishes
two bits which are adjacent in time.A major issue is the form of the medium, both in terms of the materials it uses and its topology. A basicwired Ethernet, for example, consists of cable conducting electrical signals; the connections could also be wireless. More complicated networks, including Ethernets, may consist of more than one cable, with all of them connected via a hub. The latter has become common even at the household level.
3.1.2 Data Link Layer
For example, in an Ethernet, this layer is concerned with ensuring that two network stations connected to
the same cable do not try to access the line at the same time.4 For this reason the Ethernet operation is an
example of what is called a Medium Access Control (MAC) Protocol.5
Here is an overview of how the Ethernet MAC protocol works, using a “listen before talk” approach. When
a network node has a message ready to send, it first senses the cable to see if any node is currently sending.
If so, it generates a random backoff time, waiting this amount of time before trying again. If the node does
not “hear” any other node sending, it will go ahead and send.
There is a small chance that another node actually had been sending but due to signal propagation delay the
transmission had not yet reached the first node. In that case a collision will occur, destroying both messages.
Both nodes will sense the collision, and again wait random amounts of time before trying again.
This layer also does the setting up of frames of bits (i.e. sets of consecutive bits sent along the wire), which
not only include the message itself but also information such as (say, in the Ethernet case) the Ethernet ID
number of the destination machine.Messages may be broken up into pieces before being sent. This may be handled at the transport level (seebelow), but may also be done at the data link level.
3.1.3 Network Layer
This is the routing layer. Questions addressed in this layer include: If in our example above saturn wants
to send a message to holstein, how is that accomplished? Obviously its first step is to send the message to
mars; how does saturn know this? How can alternate routes be found if traffic congestion occurs
3.1.4 Transport Layer
Suppose saturn’s message to holstein consists of a large file transfer, say 100 megabytes. This transfer will
take a long time (by network standards), and we certainly don’t want it to monopolize the network during
4In a small Ethernet, we would not even have two nodes on one cable; we would just have each station connected to a different port in the hub.And the address of an Ethernet or other LAN card is known as the MAC address of the card.This phrasing implies that the frame will be sent only to another machine on the same Ethernet. This is true, but it may be that the frame’s ultimate destination is on another LAN, and the current Ethernet destination ID is for a machine which plays the role of a intermediary router to other LANs. See the description of the Network layer below. that time. We also must deal with the fact that the buffer space at holstein won’t be large enough to deal with a 100-megabyte message. Also, one 100-megabyte message would have a sizable probability of havingat least one bit in error, and if so, we would have to retransmit the entire message!So, the file transfer must be done in pieces. But we don’t want to burden the user at saturn with the task of breaking up the 100 megabytes into pieces, nor do we want to burden the user at holstein with the reassembly of the messages. Instead, the network software (again, typically in the OS) should provide these services,which it does at the transport layer, as for example is the case with TCP.
3.1.5 Session Layer
This layer is concerned with management of a session, i.e. the duration of a connection between two network
nodes. The word connection here does not mean something physical, but rather refers to an agreement
between two nodes that some chunks of data with some relation to each other will be exchanged for some
time. Actually, TCP does this in some senses, as does the socket interface to TCP, which is very much like
the interfaces for reading or writing a file (described in more detail later).
3.1.6 Presentation Layer
This layer deals with such matters as translating between character codes, if the source uses one and the
destination the other. In the old days, this could mean ASCII at one end and EBCDIC on the other end.
Today, though, it could mean for example two different coding systems for Chinese characters, Big 5 and
GB.
3.1.7 Application Layer
You can write programs at the application layer yourself, and of course you use many programs written by
others, such as ftp, Web browsers, e-mail utilities, and so on.
0 comments:
Post a Comment