CS422: Computer Networks
Notes for Dec 8th – Dec 12th
By Melroy Saldanha
DNS (Domain Name
System):
- Humans prefer host
names instead of IP addresses
www.yahoo.com --> 204.71.200.68
- DNS does this mapping
- Pre DNS there was a
file, /etc/hosts, with the host-to-IP mapping
- This file was
maintained manually. It was ok to
do it this way when Internet was small, but impractical with 1000’s of
computers.
- DNS uses a hierarchy
of servers that handle the domains of the name. Each domain of the name
can correspond to a different name server.
- Hostnames are divided
in to domains
- e.g.
ector.cs.purdue.edu
- edu - domain 0
- purdue - domain 1
- cs - domain 2
- ector - host name
- One registers a
desired name with a central authority and it is placed in a top level
domain.
- Within the
“organization” (pets.com) you can:
- Subdivide the
domain: dog.pets.com; cat.pets.com
- arbitrary levels
are allowed
- these arbitrary
levels are not standardized
- these arbitrary
levels are controlled by the local organization
- Name resolution is hierarchical: (DNS
Server)
- Domain servers
correspond to naming hierarchy
- Individual groups
within an organization can choose their own hierarchy
DNS Name Server
Resolution (lookup)
- DNS uses client/server
paradigm
- Client known as “resolver”
- Multiple DNS servers
maybe used during a resolution
- Each DNS server
corresponds to a contiguous part of the naming hierarchy
- All domain servers are
linked together and they know the IP address of their neighbors.
- DNS lookup can be
done using either TCP or UDP
- An application that
needs to get a IP address from a host name will
become a client.
- There are 2 kinds of
resolutions:
- Recursive:
- If the client
chooses recursive resolution, the host asks the immediate DNS server to
resolve the hostname.
- If the immediate
DNS server does not have an answer (can’t find the host on its list),
this DNS server asks its parent
- This goes ‘up’ the
hierarchy until a DNS server has an answer
- Iterative:
- Green – Do You
know who “UCB” is
- Red – No, but I know a DNS server that
might
- Blue – Yes I know “UCB”
- If the host asks
the immediate DNS server and it cannot answer it will return the address
of a DNS server that may know the answer, and so on.
- To reduce DNS
traffic, the name to IP address mappings are
cached.
- The time they are
cached depends on the entry and this info is stored in the DNS servers
Nslookup
- A DNS client useful
for debugging
- A DNS query may return
more than one IP address if the host has multiple interfaces
- Also you specify the
type of host in the DNS query:
- Type:
- any
- mx, mail exchange
- Example of nslookup use
- > /usr/bin/nslookup
- > set querytype=A (any host)
- > yahoo.com
- Name: yahoo.com
- Address:
216.115.108.245 216.115.108.243
- set querytype=MX (mail exchange)
- > yahoo.com
- Name: yahoo.com
- mail
exchanger=mx2_mail.yahoo.com (might not be ‘2’)
- > set querytype = CNAME (the DNS server)
- > yahoo.com
- ns0.corp.yahoo.com
- DNS caching is
important otherwise the internet would be flooded with DNS requests
The Evolution of
Programming Network Applications
- Sockets is the ‘hard
way’ to do network programming
- Sockets:
- Equivalent to
programming in assembly language
- You have to take
care of every single byte that goes through the network
- With sockets your
program uses the network channel as if it where writing to a file or pipe
- RPC (Remote Procedure
Calls)
- A server exports
procedures to the clients
- The client server
interaction if modeled as a procedure call
- The programmer
writes in a special language called IDL (interface definition language)
the header definitions of the procedures that the server will export
- This interface
description is preprocessed by a special IDL pre-processor and generates
code for the client and server that does the entire communication
- The programmer doesn’t
need to write any networking code.
To program the server it will add code to code already generated,
to implement the procedures
- The client will use
the procedures in the server as if they where local
- The Client Stubs:
- provide a function
with the name of the Remote Procedure Call
- they put the arguments into a network packet(s). This
is called serialization, or marshalling. The format of the data in the
network can be
- XDR – External
Data Representation. It was created by SUN to represent data in the
network
- XML – textual
representation of data that may also include semantics
- The client sends
the data arguments to the server and waits for the answer
- When the answer
comes back from the server, it translates back the results from the
network representation to the machine representation. This process is called deserialization or unmarshalling.
- Then the results
are returned to the caller
- This process is
transparent to the program
- The Server Stubs:
- Wait for an RPC
invocation
- It deserializes the arguments
- It invokes the
procedure indicated by the client
- It serializes the
results and sends the results back to the client
- The programmer
doesn’t need to know about sockets
- Implementations of
RPC - Sun RPC
Evolution of Programming Networked Apps
- Networking Tools
- Sockets
- RPCs (Remote Procedure Calls)
- RMI (Remote
Method/Object Invocation)
- Programming Language
equivalents to Networking tools
- Assembly Language like
Network Sockets
- Procedural
Languages (e.g. C, Pascal) like RPC
- Object Oriented
Languages (e.g. JAVA, C++) like RMI
- RMI (Remote
Method/Object Invocation)
o
The client
server interaction is modeled as a object method
invocation.
o
The stub or
proxy is a local representation of the remote object
o
The proxy
forwards the call to the remote object
o
The client
program interacts with the remote object as if it where local
o
The proxy code
and the code in the server that exports the remote object is generated by a
special preprocessor from the description on a interface description language
o
RMI
Implementations
Summary
2nd half of semester
- Protocols/layering
- ISO-7 layer
reference model
- Internet Reference Model
- Internet
- What is the
Internet?
- Why the internet is
necessary
- Routers and Hosts
- IP (Internet
Protocol)
- IP addresses (4
bytes in length)
- Network/Host
portions of the address
- Classes of IP
addresses (A, B, C, D, E)
- Dotted decimal
notation
- Special addresses
(all 0’s, all 1’s, loopback, etc.)
- IP packet format
- All fields *TTL,
SRC, DST
- IP Routing
- IP routing table
and default router
- IP semantics
- IP is
connectionless
- IP packets can be
delayed, come out of order, or even be lost
- Best Effort
Delivery
- ARP (Address
Resolution Protocol)
- Translates IP
addresses to local hardware addresses
- ARP look-up
algorithm
- ARP cache mechanics
- IP fragmentation and
reassembly
- Why fragmentation
is necessary
- MTU – maximum
transfer unit
- How Fragmentation
works
- Fields used for
reassembly (source
- Reassembly timeout
- ICMP (Internet
Control Message Protocol)
- ICMP types: Source
quench, time exceeded, destination unreachable, echo reply (ping)
- Uses (applications
discussed): Traceroute, path MTU
- Transport protocols
(UDP, TCP)
- Characteristics
- Advantages/disadvantages
of both
- TCP
- Characteristics:
- Reliable
- Connection
oriented (app-2-app)
- Full duplex
- How the
reliability is achieved
- Retransmission
- Acknowledge
(positive ACK when packet received)
- Features:
- Adaptive
retransmission
- Flow control
(window size)
- Slow down the
sender if buffer space is not available
- Cumulative ACK
- Congestion Control
- Slow Start
- Reliable
connection start/shut-down
- 3-way handshake
- TCP header fields
- DNS (domain name
service)
- Translates host
names to IP addresses
- Domain Names
- Domain Name Server
hierarchy
- DNS lookup
(resolution)
- Recursive
- Iterative
- DNS caching
- DNS types
- Evolution of Network
Programming
- Sockets
- RPC
- RMI
- Final on Friday December
19th