Transcript OSI Layer Model - University of Maine System

ISO Layer Model
Lecture 9
October 16, 2000
The Need for Protocols




Multiple hardware platforms need to have the ability to
communicate.
Writing communications code in this instance is nonportable (platform-dependent).
Protocols provide a high-level interface method to the
communications medium.
The protocol itself is a set of rules by which the
software and hardware must follow in order to
communicate properly.
Need for Protocols (cont.)



Multiple types of protocols exist.
RS-232 is form of a protocol in the sense that strict
rules must be followed for communication to occur
between devices.
Network protocols cover almost all communications
between two or more computers.
–
–
Transferring an ASCII file
Point-to-point encryption
Protocol Suites



A single, all-encompassing protocol is not only
inefficient, but also wasteful on CPU, memory
resources.
Programmers have created smaller suites
which provide the rule sets for communication
within that protocol.
The suites are also divided into smaller pieces,
called layers.
The ISO Layer Model
Why the Layer Model?



It is simple to visualize how software interacts
with hardware.
Topmost layer is closest to the user.
Bottom layer is the furthest away from the user.
Layer Breakdown

Layer 1: Physical
–

Layer 2: Data Link
–

The physical network hardware, medium.
How data is organized into frames, and how to
transmit those frames. Byte/bit stuffing, checksums.
Layer 3: Network
–
How addresses are assigned, and how packets are
transmitted from one end of the network to the
other.
Layer Breakdown (cont.)

Layer 4: Transport
–

Layer 5: Session
–

How to start sessions (connections) with remote devices,
machines. (Sockets)
Layer 6: Presentation
–

How to reliably transfer data.
How to represent data. Does int, char replacements.
Layer 7: Application
–
How one user-level program requests a connection to another
machine, and how the machine responds.
Protocol Stacks




The protocol is divided into modules.
Each module corresponds to a different level in
the layer model.
All of the modules put together is the “stack.”
Theoretically, a given module can only
communicate with the module above and
below it.
Stack Examples
Communication With Stacks
Stack Interaction



Because the stack dictates how computers
communicate, machines utilizing one type of
stack cannot communicate with a machine
running another type of stack.
Netware v. TCP/IP
But machines can concurrently run more than
one stack!
Layering



At each level where the packet is processes,
additional information beyond the actual data is
added or removed.
At the Layer 2, a checksum is added before
being sent to Layer 1.
When the Layer 2 of the receiving machine
receives the packet, it strips off the header and
the checksum footer.
A Layered Frame
Protocol Dependent Information


The previous picture is a general idea.
Some protocols specify that all additional
information must be appended instead of
prepended, etc.
Layering Overhead


Because the sending computer adds
information to the frame, the receiving machine
must undo the changes before the frame is
passed to the next layer.
Because layers must undo changes that the
counterpart creates, code for the other layers
can be independently created and updated
without affecting the other layers!
Data At Each Layer
Protocol Optimization Techniques


Protocols utilize different methods depending
on the focus and goal of the protocol.
Sequencing
–
–

To avoid out-of-order delivery
To avoid duplicate packets
Packet Retransmission
–
When a packet is received, an ACK is sent back. If
the ACK isn’t received by the transmitter, packet is
resent. (Bound by time.)
Optimization Techniques (cont.)

Avoiding Excessive Delay
–

Each transmission is assigned an ID. Packets must
contain a valid ID to be accepted for a given
“conversation.” p. 213
Flow Control to Avoid Overrun
–
–
Machines do no communicate at the same rate.
Sliding Window method.
The Sliding Window



The sliding window concept greatly increases
protocol throughput.
All packets in a given “window” are sent
sequentially, without waiting for an immediate
ACK.
Once all packets in a window have been sent,
transmitter waits for all ACKs to be received
before continuing on to next window.
Network Congestion


If a single computer
sends, no congestion on
network link.
If more than one
computer sends, packet
queuing must occur, so
some congestion exists.
Network Congestion (cont.)



If congestion persists over time, packets fill the
switch’s memory, and will begin to be
discarded.
Retransmission can fix this problem, but if the
congestion still exists when the retransmission
occurs, those packets will also be discarded.
This scenario is called congestion collapse.
Avoiding Congestion



If a switch is having congestion, send a
message back to the sender informing them of
the congestion.
Packet loss as an estimate of congestion.
Rate control
–
–
Temporary rate reduction
Sliding window size reduction
Protocol Design

Flow control vs. congestion control
–
–

Flow control sends more packets to improve
throughput
Congestion control reduces the number of packets
Sequence number
–
–
Too small, miscommunication, confusion
Too big, wasted bandwidth, overhead