Chapter 2, Protocol Architecture, TCP/IP and Internet-Based Applications.

You should be able to
- Describe protocol architectures and identify why such architectures are employed.
- Identify the key components of the TCP/IP architecture and briefly describe each.
- Identify the key components of the OSI model and briefly describe each.
- Describe data encapsulation and identify how, why, when this occurs.
A protocol is a set of rules governing the exchange of data.
- Protocols are normally concerned with three items
- Syntax, or how the data is formatted
- Semantics, or what happens. This includes error handling and control information.
- Timing, or what happens next.
Architecture is
- look at Websters for a definition.
- What is an ISA from CSCI 312?
A protocol architecture is a layered structure of hardware and software that supports the exchange of data between systems. Furthermore this architecture should support distributed applications.
He starts with an example: File transfer
- Contact must be established
  - Something must initiate the communications
  - something must be listening to establish this contact with.
  - Credentials must be established
  - Data about what is to be transfered and how to transfer it must be exchanged.
- The source system must ascertain that the destination system is prepared to receive data.
- Potential data manipulation may need to occur, such as little endian to big endian transfer.
- Something needs to assure that the transfer is complete and complete the session.
At this level, there is a need to define how to transfer files between machines, that does not involve
- How those machines are connected
- How to break the message into manageable pieces
- How to make sure the pieces get there.
- How to make sure that there are no errors in the pieces
- ....
A bad programming approach is to put everything into the application, and rewrite it every time you have a new application.
A good approach is to build a modular system, or protocol architecture
These modules are arranged in a stack or layered system.
- Each layer provides a set of services to the layer above.
- Each requests services from the layer below.
- The interface between layers is well defined so that the internals of a layer may change, but the network continues to work.
The TCP/IP architecture is just such an architecture
- Named for two of the layer protocols that make it up
- TCP - Transmission Control Protocol.
- IP - Internet Protocol
- It came from DARPA
- In general three agents:
  - Applications -
    - such as mail, ftp
    - Concerned with how distributed systems interact and how they exchange data.
    - For example, much of the example information is the concern of the application.
  - Computers
  - Networks
- Originally
  - Designed to provide communications across arbitrary, multiple packet switched networks.
    - A bad land network
    - Packet Radio
    - A satellite network.
  - It has many configurable parameters because of this.
- The TCP/IP suite has the following layers:
  - Physical:
    - the actual network hardware
    - Characteristics of the media
    - What a high signal is
    - What the clock rate is.
  - Network Access layer
    - how the network hardware is used.
    - Information such as address of the machine to send the data to, but on the same physical network.
    - Quality of service, priority, other network level information.
    - This is the local area network, Ethernet, frame relay
  - Internet layer
    - IP layer
    - A router connects two (possibly) different networks.
    - The Internet layer provides services for the router to determine things such as
      - Where should this packet be sent to next.
      - Do I need to change the format of the data?
      - Should I allow this packet through?
    - Routers "speak" IP, they must be able to examine the packets to determine their destination, source and possibly other details.
  - Transport Layer
    - TCP layer
    - Among other things, is responsible for reliable exchange of data.
    - Decides which application to deliver the data to
  - Application Layer
- A quick trip through a TCP/IP session
  - Let us assume the following configuration:
    - A node on my cluster needs to transfer data to a node on my Jim Kirk's machine in physics.
    - My cluster is connected by a 1GB Ethernet switch via cat 5 cable.
    - The head node is also connected to a 100MB Ethernet switch in the department via cat 5 cable.
    - The department's switch is connected to the University's main router via fiber optic cables via 1GB Ethernet.
    - The router is also connected to a switch in Hendricks hall via 1GB over fiber.
    - The switch in Hendricks connects to Dr. Kirk's machine via cat 5.
    - Notice that there are 3 subnetworks employed in this connection.
      - The local network on the cluster
      - The LAN for Math and CS
      - The LAN for the Hendricks Hall (I assume)
      - Each is possibly operating on different media, and employing a different network access protocol.
  - In the case of the cluster network
    - This is gigabyte Ethernet.
    - Each machine has a MAC address (Medium Access Control)
    - node 2 is 00:11:11:46:FB:0B, node1 is 00:11:11:4B:31:21
    - This address is for the local Ethernet network, is should be unique to each network card.
    - Each machine has a unique IP address as well.
    - At the Application level, a connection needs to be made between node2 and hess.
      - This application will employ TCP ports.
      - A port is somewhat akin to a mail box.
      - There can be many ports on a single machine.
      - Each application needs to be connected to at least one port (assume 54321).
      - In the end the source port/ip address destination port/ip address make up a connection.
      - The application hands the message, along with the destination ip address and port number to TCP
      - It has already ask TCP to assign it a port on the local machine (12345)
    - TCP then deals with the data
      - It turns the message into one or more packets.
      - It adds TCP header information.
        
        This includes the source and destination ports
        A sequence number
        checksums plus other information.
        (see page 39)
      - This process of adding control information to a packet is called encapsulation
      - TCP then hands the data to IP
    - Ip then:
      - Adds another layer of information to the packet
        
        Source and destination IP addresses
        Other book keeping information.
      - The new datagram is handed to the Network access layer
    - The next step is to determine the physical layer address for the packet(s)
      - The netmask is or-ed with the destination address and compared to the network address. (ALL IP)
      - If they match, it is for local delivery, so the network access address of the destination is determined.
      - If they do not match, then the network access address of the router is determined.
      - This uses one or more routing protocols
      - As well as a network access layer protocol such as ARP - address resolution protocol
    - The Network access layer
      - Once attaches some information such as the MAC address of the router.
      - This actually uses both a preamble and a postabmle to generate an Ethernet packet.
      - In this case, the mac address will be that of node1
    - The packet(s) is/are then placed on the network.
    - The switch receives this packet
      - Node 1 receives the packet
        
        In this case, node 1 is acting as a router.
        It is connecting two dis-similar networks.
        It determines the next step destination via routing protocols in this case, a table lookup.
        It strips off all of the 1GB Ethernet information
        Perhaps it breaks up the packet into more packets (depending on the destination network)
        It then forwards it to the (real I assume) router.
      - And the process continues until it reaches Hess.
      - On Hess the opposite occurs.