Transcript Network Organization

Network Organization
Max Gorinevsky
56847972
NETWORK ORGANIZATION
The network is the computer™ ~ Sun
Microsystems

Sun’s 1980s advertising campaign.
Nowadays, networks are everywhere, and
networks are all-important.

How many hours a day do you spend on the
internet? 3? 5? 10?
Sun was right. In the modern world, the
stand-alone/non-networked computer is
entirely obsolete.

Would you rather have an iPod Touch with WiFi
access, or a top-of-the-line computer without
internet access?
From Local Area Networks to the World
Wide Web, networks are everywhere and
they tie into each other. Understanding
them is more important than ever, and will
be more important still.
The evolution of computer networks: two
different paths


Corporate: fast, accurate business transactions
Academic: facilitate collaboration, share knowledge
All networks want to share resources as simply,
quickly, and cost-effectively as possible
Computers used to cost millions of dollars:
companies could only really afford one system…
but every location needed the system.


Thus, you need connectivity
IBM developed the Systems Network Architecture,
which has lasted over 35 years
Systems Network Architecture



Describes a protocol stack for connecting computers
Original objective: increase hardware sales… by
reducing the cost of operating lots of computers.
Specifics for communication between physical units,
over which logical sessions happen.
Original version of SNA



Not flexible enough
Lots of overhead
Changed over time
But kept the same basic structure
SNA was useful, but its time has more or less
come: the replacement is OSI and other open
internet protocols.

SNA was designed before the dawn of layered
communication; many things were mixed into the
same protocol
Difficult to manage

SNA installation and maintenance are complex
Also expensive! SNA license - $10K/month for high-end
systems, like a corporation would want


Connecting to non-SNA networks was difficult
Every computer manufacturer used to have its own
communication protocols! Just try getting them to
work together…
Tower of Babel
Open System Interconnected Reference Model

Divides network architecture into 7 layers:
Each OSI layer has functionality, which is
implemented by one or more entities (and
broken down into other protocols)
Info starts at “layer 8”, the user, and gets passed
down through the layers, processed until it’s
turned into 1s and 0s, when it’s sent to another
computer


Wrapped up in headers as it goes
Receiving computer “unwraps”
headers
Like Matryeshki, Russian dolls
Overview of the 7 layers:





L7 (Application): applications, end-user processes.
Browsers, email, etc.
L6 (Presentation): translates application format to
network format (and vice-versa). Formats and
encrypts data to send it across network. Syntax layer.
L5 (Session): Establishes, manages, closes
connections between apps. Session coordination.
L4 (Transport): top (TCP) half of TCP/IP. Transfer of
data between systems.
L3 (Network): bottom (IP) half of TCP/IP. Handles
switching and routing, addressing, packet sequencing
Overview of the 7 layers (cont):

L2 (Data Link): Ethernet, data frames, etc. Handles
physical layer errors, flow control.
Media Access Control sublayer: controls how a computer on
the network gets access to data, permission to transfer
Logical Link Control sublayer: frame synchronization, flow
control, error checking

L1 (Physical): bit stream—electrical impulses, light
(optic fibers), radio (wireless). Hardware. All
physical—1s and 0s transmitted as current. Covers
cables, wireless cards, etc.
Example of how this works follows…
OSI headers: important – standardization!
Data is broken up into packets…




Each packet is “wrapped up”, like a letter
The name, address and return address on the
letter would be the header!
Then, UPS assigns a tracking code to the
letter: you still have the name/address, but
now it’s “wrapped up” in a tracking code
That’s kind of how OSI headers work! Each
header is wrapped around the previous ones,
containing information relevant to that layer’s
role
Each layer has its own header

Encapsulation
This does mean that each packet is
wrapped in a lot of non-data information
Any given amount of data breaks down into
more packets than you’d expect if you didn’t
account for the headers
TCP/IP: don’t confuse the protocols with the
TCP/IP model. The TCP/IP model competes with
OSI, and has 4 layers (Application, Transport,
Internet, Physical). Slightly simpler.


Two ways of describing the same things!
IP header
Back to TCP/IP: this is the protocol stack many
people who only know a little about networking
are likely to have heard of.
TCP/IP spread while the International
Organization for Standardization was still
perfecting the OSI model
Efficient and “lean”: IP layer is coupled with
OSI’s Data Link and Physical layers, so TCP/IP
can be used with any type of network. Even
different types in the same session!

As long as all networks are running IPv4 (or greater)
-Greater? Yes, IPv6 is coming.

And has been coming for years…
IP was designed before anyone knew just how big the
Internet would be.
 Originally, people at the ARPAnet project weren’t even
sure there would be a civilian use for it!
And now, some of us might say “LOL” to describe this doubt.
Because we’re used to saying that on the Internet.

Packet length (65,536 bytes) has become a
problem… packets move too fast to check for damage
between nodes… but addressing is a bigger problem.
Every host, every router, needs a unique address. That’s a
LOT of IP addresses. 32 bits of space in the header, 232
addresses… about 4.3 billion.
Seems like it should be enough… but that ignores routers,
switches, etc. Also, some are reserved, and network classes
(class A = up to ~16.7 million nodes, class B = 65,534, class
C = 256) make assigning enough (but not too many) IPs to
each organization, company, etc difficult!
IPv6 to the rescue!

More address space! Much more—2128
addresses… more than we’ll ever need.
~5 x 1028 for each person alive today.




Greater flexibility in assigning addresses
Backwards compatible with IPv4
Will be able to tell the difference between realtime (video, voice, etc) traffic and less timesensitive (e.g. email) traffic
So what’s the hold up?
Money. And time, but time is money.
Hardware replacement, technician training, etc…
Network Organization: between the internet and
your computer…
LANs: the smallest networks


Single building, or group of
nearby buildings…
People who own the LAN are
usually the same people who own the
physical area the LAN is on
MAN: metropolitan-area networks. Cover a city

In Mountain View, anyone can hop onto Google’s
MAN for free Wi-Fi access!
WAN: Wide-Area Networks. Huge.
Once, LANs, MANs, and WANs were very
different. MANs and WANs were designed
for high-speed throughput, meant as
backbone systems for slower LANs.

Now, they’ve become very similar.
Distinguished primarily by ownership!
LANs are becoming faster, easier to
integrate with WAN technology. MANs
may disappear entirely

WiMAX, a MAN technology, seems like it will
lose out to 802.11n (“WLAN”)…
References
Our textbook
http://www.novell.com/info/primer/prim05.html
http://www.spottek.ca/Tutorials/Intro.html
http://www.wtcs.org/snmp4tpc/literature.htm
www.infocellar.com/networks/osi-model.htm