Internet Protocols

Two Protocols

Open Systems Interconnection (OSI)

• Overview
  • Based on strictly controlled layers
  • ISO standard
  • If you follow it it works, but not much room for expansion
  • Layers
    1. Physical
       • Provides electrical connections to a transmission medium
       • Uses bits
       • One device to another
    2. Data link
       • Packet framing. Hardware addressing
       • Uses MAC addresses and frames
       • One device to another
    3. Network
       • Addresses packets and routes them to provide end-to-end communication
       • One network to another
    4. Transport
       • Transfers data reliably or unreliably
       • Reliable transfer involves a connection
       • One process to another (TCP/UDP)
    5. Session
       • Opens/Closes session between applications
       • One application to another (socket port) (and the rest below)
    6. Presentation
       • Formats data (compression and encryption)
    7. Application
       • Provides the user interface

Internet Protocol (IP)

• Overview
  • Incremental implementation and model
  • Based on loosely layered model
  • Created for ARPAnet (US military)
  • Not a strict standard, and fully extensible to allow for rapid innovation
• Layer
  1. Access (ethernet and electrical connections)
  2. Network (IPv4 and IPv6)
  3. Transport (TCP, UDP)
  4. Application Protocols (FTP, TFTP, HTTP, SMTP, …)
• Within OS
  1. Access: Network interface and driver
  2. Network/Transport: Protocol stack
  3. Application: Sockets API
• Packet and Frames
  • Packet switching network protocol
  • The size of data sent over the network is more than just the data due to the overhead of each layer
  • Packets are for transmission between networks (contains IP)
  • Frames are for transmission through the access (contains MAC address)
  • Steps
    1. Data
    2. Addressing
    3. Routing
    4. Error Detection
• Delivery Mechanisms
  • Unicast: Single destination
  • Multicast: Group destination
  • Broadcast: All network hosts

Transport Protocols

Transmission Control Protocol (TCP)

• Connection-oriented protocol (impression of a dedicated channel)
• Data is delivered in-order
• Lost data (dropped packets) is re-transmitted
  • All data is sent with a transmission number and a checksum
  • Acknowledge sent for each package
  • Will resend if not acknowledged fast enough
• Only unicast
• Three step handshake
  1. A to B: sync
  2. B to A: acknowledge sync
  3. A to B: acknowledge
• Three step handshake used for closing as well (just with finish instead of sync)
• Features
  • The client sends a confirmation of data delivery to the server and data arrives
  • If a confirmation does not arrive after a timeout period, the server attempts retransmission for the data that was not delivered
  • In a congested network, TCP can delay data transmission
• Flow control
  • Prevents a sender form overwhelming a receiver
  • 
  • Receive window is the remaining space in the receive buffer. It can be used for simple load balancing
  • When the window size is 0, it will stop transmitting data
  • With every ack message, the receiver advertises its current receive window

User Datagram Protocol (UDP)

• Connectionless (like a mailbox)
• Best effort delivery
• Data may arrive out of order
• No handshake
• Unicast, multicast, or broadcast
• Features
  • Adapts to bandwidth-intensive applications that tolerate a loss of packets (ex: video streaming)
  • There are fewer delays in data transmission
  • It is used to send a large number of packets at a time
  • There is a possibility that you may lose some data