Software Protocol Layers
LESSON 5

Concepts:

• "protocols are to computer communication as what programming languages are to computation."
• Layers are conceptually stacked modules of hardware-dependent or independent protocol software.

• Complex data communication systems require a set of cooperative protocols (called a protocol family or protocol suite) which solve the following communication problems:
  • Hardware failure.
  • Congestion.
  • Packet delay or loss.
  • Error detection and/or correction.
  • Flow control.
  • Redundant packet arrival.

Functionality of Layers:

• A set of goals and constraints governing a particular communication problem is usually set by a standards organization (organizations optimize the functionality)
• Choice of different approaches
• The design of an internet architecture and the organization of protocol software are interrelated.

Protocol Type by Layer:

• Hardware/Firmware protocols usually reside at the Physical and Link Layers.
• Software protocols conceptually reside above hardware protocols and can be located within:
  • Device drivers - lowest software layers, hardware dependent.
  • The operating system - middle software layers, hardware independent.
  • Application programs - applications and highest software layers.

ISO - Open Systems Interconnection (OSI) model:
(standards bodies)

TCP/IP model:
(researchers)

Three conceptual layers of software protocols that build upon a fourth layer of hardware/firmware protocols.

OSI - TCP/IP Differences:

• OSI - provide reliability at all layers (including point-to-point at link layers).
• TCP/IP - provide reliability only at hosts (end-to-end problem).
• These reflect the argument over where to put the complexity in communication systems:
  • Subnet
  • Hosts

Focus of Reliability:

• TCP/IP software focuses most of it’s reliability control in the Transport layer, including:
  • Detection / Recovery of datagram loss.
  • Duplicate packets.
  • Time-outs
  • Acknowledgments
  • Flow control.
  • Checksums

Protocol Layering Principles:

• Layered protocols are designed so layer N at the destination receives exactly the same object sent by layer N at the source.
• Developers only need focus on one layer at a time and focus on the messages needed to be passed to get it’s job done.

• Application and Transport layers - deal with end-to-end issues and are designed so the software at the source communicates with it’s peer at the final destination.
• Internet and Network-Interface layers - view an end-to-end connection as the each pair of links in the network.

Layering within a Network Substructure:

• Two methods for accommodating point-to-point connections:
  • Each physical link is treated like any other network in the internet - unique addresses - both hosts share the same link, but each have unique IP addresses.
    • Disadvantage - proliferates the number of networks and increases routing table sizes.
  • Accommodating point-to-point connections to avoid assigning multiple IP addresses - to the same physical link.
    • This treats all connections as a single independent IP network with it’s own frame format, hardware addressing, and data link protocols.
    • This approach forces extending the protocol layering scheme to add: Intranetwork routing layer.
    • Intranetwork routing layer is used to accommodate SLIP and PPP protocols.
    • The point-to-point intranetwork module resides between interface device drivers and the underlying hardware.

• Intranet routing distinguishes among physical connections (routing them to their correct link) on a single point-to-point network.

Boundaries in the TCP/IP model:

• Protocol Address Boundary - separates high-level and low-level addressing.

• Operating System Boundary - separates the system from application programs.

Disadvantages of Layering:

• Software that results may be extremely inefficient.
• The smaller the packets are, the larger the percentage of overhead and control information.

• Route selection and choosing optimal maximum transfer unit (MTU) sizes can help.

Multiplexing, Demultiplexing, and Inverse Multiplexing:

• Multiplexing/Demultiplexing - computers include extra bits that encode the frame or packet type, originating program, and the protocols used. This allows all frames or packets to share the same channel by being combined into a single stream and then separated at their destinations.
• Depending upon the flow of information, multiplexing and/or demultiplexing occurs at most protocol layers.
• When demultiplexed, frames or packets are delivered to their appropriate protocols for further processing.

• Inverse multiplexing is an extension of multiplexing/demultiplexing that is not used by protocols, but hardware.
  • It involves dividing a single stream of data over many out-going connections, to increase the available bandwidth.
  • The multiple out-going streams are then recombined at the destination.

Hiding underlying network details:

• Goal of network protocol design - to hide the technological differences between networks, making interconnection independent of the underlying hardware. This directly applies to higher-layer software protocols.
• Two approaches to hiding network details include:
  • Application-level - using application programs to handle heterogeneity.
  • Network-level - hiding details in the operating system.

Application-Level Approach:

• Services that the end user of the Internet sees; application programs that use the network to carry out useful tasks.
• Same applications on each machine understand the details of connection for it’s machine and handles the interconnection.
• Cumbersome approach, must build new applications to add new functionality
• Step-at-a-time communication is poor over many machines

Network-Level Approach:

• Interconnection provides a mechanism that delivers packets from their original source to the destination in real- time.

Interconnection at Network-level:
Advantages

• Advantages:
  • Efficiently maps onto underlying network hardware.
  • Detaches communication chores from applications (application independent and physical connection details are transparent).
  • Flexible (general-purpose protocols designed to perform a specific task).
  • Modular (can interchange protocols.
  • Users need not understand how networks connect or the traffic they carry.

Interconnection of Networks:

• Internetworking - detaches the notions of communication from details of network technologies and hides low-level details from the user.
  • It drives all software design decisions and explains how to handle addressing, routing, etc.

• Network Independence:
  • Universal service.
  • Hide low-level details.
  • Universal set of machine identifiers (naming and addressing).
  • Common user interface and command-sets (standard application services).
• Gateways - software in routers pass packets between connected networks, routing them to their appropriate destinations and supporting the interconnection of networks:
  • Larger the interconnection network, the more complex routing decisions need to be made by the gateway.
  • Gateways route packets based on the destination network, not on destination host.
  • The amount of routing information needed by the gateway is proportional to the number networks.

TCP/IP Protocols:

• TCP/IP internet protocols treat all networks equally.
• Any communication system capable of transferring packets counts as a single network, independent of its delay and throughput characteristics, maximum packet size, or geographical scale.
• TCP/IP defines an abstraction of "network" that hides the physical network’s details.

All work was written, produced, and is copyrighted by Daniel Z. Tabor Jr.
Page created by Daniel Z. Tabor Jr. Copyright ©1995 Illusion Industries Inc.