Lecture Notes
for
CSCI 1301: Great Ideas in Computer Science
(c) 1993 - 2007 John E. Howland, All Rights Reserved
Department of Computer Science
Trinity University
One Trinity Place
San Antonio, Texas 78212-7200
Internet: jhowland@ariel.cs.trinity.edu
A network is a collection of two or more connected machines together with
operating system software which facilitates communication between machines
(sometimes called nodes) on the network. There are several common
interconnection topologies used to network computers. The choice of network
connection architecture depends on many factors such as proximity of machines,
building features, desired network performance, etc.
Bus networks have a single network cable (plus ground wire or shield)
which attaches to all machines on the network. Different types of wiring are
widely used. One type, a coxial cable, uses a single copper wire with a
special ground shield conductor which surrounds the copper core. Another type
uses two pairs of twisted copper wire.
Each machine on the network sees every message sent on the network, but only
processes messages which pertain to that machine. Special signaling techniques
are used to detect and avoid message collisions when two machines attempt to
transmit a message at the same instant in time. The method used to avoid
future collision when two or more machines detect that their messages were
transmitted at the same time is to have each machine compute a small, but
random, delay time before trying to transmit again. This network access
discipline is known as Carrier Sense Multiple Access Collision Detect (CSMACD).
Star networks use a single network node which connects to every other
node in the net and provides communication services to other machines on the
net.
In a star network, each machine may communicate with the Hub machine without
worrying about whether or not their messages will collide with other messages.
The performance of a star network is determined largely by the performance of
the Hub machine. The Hub machine performs all message routing to the desired
destination machine.
Ring networks use a ring of connections from one machine to the next.
Messages are forwarded from one machine to the next until they arrive at the
desired destination. A special message, called the token, is sent from
one machine to the next around the ring. The message collision problem is
solved in a ring network by requiring each machine to wait until the token
arrives before transmitting a message. When a message arrives at a machine it
is forwarded to the next machine in the ring unless the message is for that
machine. If a message is circulated all the way around the ring and arrives
back at the sending node, then this condition is recoginized as an error (the
addressed machine was unavailable to receive the message).
Messages which are sent from one machine to another are subdivided into
relatively small sized blocks called packets. This is done primarily to
prevent one machine from unfairly monopolizing the network when transmitting
long messages.
An internet is a collection of interconnected networks together with
special software which facilitates communication between networks.
An internet is built from two or more networks by having a network connection
between two machines, each of which are on different networks.
The machines M5 and M3 are sometimes called bridges, routers or
gateways because these machines each exist on two networks. Special
software is required to route messages between the networks through these
machines. Sometimes, rather than use two general purpose computers to function
as a bridge or gatway connecting two networks, a special purpose computer
together with routing software is used to connect two networks.
When several different networks are connected together, a common set of
operating system routines must be provided on each machine. Such software is
referred to as a network protocol. When the machines run the Unix operating
system, an often used protocol is TCP/IP. TCP/IP stands for Transmission
Control Protocol / Internet Protocol. These protocols provides efficient and
reliable communication mechanisms for sending data and messages across an
internet.
An internet usually contains a variety of different machines which are not
easily compatible with one another. Hence, there must be different
implementations of the common network software protocols, such as TCP/IP, and
each of these implementations must provide the necessary communication
compatibility between the incompatible machines.
Other noteworthy protocols include Novel IPX, AppleTalk and DECNET. Each of
these protocols has been developed by a particular vendor, but, because they
are widely used, they are licensed by other vendors so as to provide greater
compatibility. Each of these protocols has become a defacto standard for
networking.
Machines on an internet need a naming scheme which uniquely identifies each
machine on the internet. Ther world-wide internet uses a domain naming system.
The top-level domains indentify the country in which a machine resides by a
code (uk - United Kingdom or ca - Canada) , but for historical reasons, in the
United States, top-level domains are: edu - school, gov - government, com -
commercial site, net - network administration or mil - military. The next
level domain includes a site name. trinity.edu, for example, is the domain
containing all machines on the internet which are located at Trinity
University. The next level domains specify one or more sub domains at a given
site. For example, cs.trinity.edu is the domain name for all machines in the
Trinity compter science department. However, the cs domain contains a
subdomain mars.cs.trinity.edu which contains two machines. Similarly,
engr.trinity.edu is the domain of all machines on the internet which are in the
Trinity engineering science department. Finally, each machine in a domain must
be uniquely named. For example, uranus.cs.trinity.edu is one of our laboratory
machines and is uniquely identified even though other machines named uranus
exist in other domains on the internet.
Networks and machines are numbered with 32 bit integers which are represented
usually in the form ddd.ddd.dddd.ddd where ddd is a number
ranging from 0 to 255. Each number is subdivided into two parts; the network
number and the machine network interface number. For example, the machine
uranus.cs.trinity.edu is 131.194.71.11 . A special database
system, Domain Name System (DNS) is used to associate the name
uranus.cs.trinity.edu with its number 131.194.71.11 .
Trinity's network number is 131.194 . The computer science department
uses 71 as a prefix for all of the machines in the department with
11 being the number of the network interface for uranus.
Networks are used to share system resources between several machines. For
example, a printer which is attached to one machine on a network might be
accessed by other machines on that network. Similarly, files which are stored
on one machine might be accessed by other machines on the network. So the
sharing of resources involves not only sharing devices such as printers, but
also the sharing of data.
Recent estimates of the size of the internet indicate that several million
machines are connected and new connections are being made at a rate of more
than 100,000 per month. It is also estimated that there are databases
containing tens of thousands of Giga bytes of information available on the
internet. A Giga byte is a thousand mega bytes!
Given the right permissions on unix machines which are connected to an
internet, a program running on one machine on the internet can request that
another program be run on another machine and have a resulting value returned
to the first program over the internet.
The following function may be used to evaluate a J expression on another J
equipped Unix workstation located somewhere on the Internet. This function
returns the resulting value back to the requesting J session. This function
assumes that the two involved machines share the same file system, since files
are used to send the expression to the remote workstation and retrieve the
value of the expression from the remote workstation.
remote_eval =: dyad define
y. write '../remote_eval_in'
spawn 'ssh ', x. , ' /usr/local/bin/j_rem_eval'
read_list '../remote_eval_out'
)
This
dyad uses a Unix shell script, j_rem_eval, which starts the J system
on the remote system. For example:
'janus10.cs.trinity.edu' remote_eval '10 + i. 10' ==>
10 11 12 13 14 15 16 17 18 19
#!/bin/csh
echo "((3 \!: 1) "'"'". (1 \!: 1) < 'remote_eval_in') 1 \!: 2 < 'remote_eval_out'" | /usr/local/bin/j