Network Layer Overview - Data Plane

Network layer services and protocols

Network layer services and protocols are focused on these two tasks:

1. Forwarding
   • moves packets from a routher’s input link to appropriate router output link.
2. Routing Algorithm
   • determines route for packets from source to destination.

Network Layer Service Model

Network layer is a best-effort model so it has no guarantees on:

• successful datagram delivery to destination.
• timing or order of delivery.
• bandwidth available through end-to-end flow.
• Upper layer protocols (e.g. TCP) provide reliable data transfer instead.

Best effort model is adopted because:

• Simplicity of machanism allows Internet to be widely deployed and adopted.
• Sufficient provisioning of bandwidth allows performance of real-time applications to be “good enough” for “most of the time”.
• Application layer provides replicated and distributed services (e.g. Datacenter or CDN), placing copies of services closer to clients.
• 05-TCP Congestion Control adapts sending rates to current network conditions.

Data plane and Control plane

Data Plane

• is local and per-rounter function.
• determines how datagram arriving from router input port is forwared to a router output port.
• essentially handles forwarding.

Control plane

• is network-wide logic.
• has two approaches:
  • traditional distributed routing algorithms are implemented in a set of routers.
  • SDN (Software-defined networking) is implemented in remote servers as a centralized service.
• essentially handles routing.

Router Architecture Overview

• Router Input and output ports are composed of 3 layers:
  • Physical layer (green)
    • recieves electrical, or wireless signals and convert them into bits.
    • sends signals converted from bits.
  • Link-layer (blue)
    • processes Ethernet frame.
  • Network layer (red) provides lookup and forwarding:
    • examines header field values and forward packets to appropriate output ports using forwarding table stored in each input port’s memory (decentralized).
    • entails queueing when arrival of packets is faster than transmission of packets.
• Switching fabric
  • is the internal and high-speed architecture within a switch or router that connects input ports to output ports, facilitating data packet forwarding.
• Routing processor
  • is the core control plane components in network routers, responsible for
    • running routing protocls (like BGP, OSPF).
    • providing forwarding tables.
    • managing system functions.

Buffer Management

Packets can be dropped when buffers are full. Therefore, a careful buffer management is required for efficiency:

• drop or add: which packet to drop or add.
• priority: which packet is to be privilieged.
• packet scheduling:
  • FCFS
  • Priority Scheduling
  • Round Robin Scheduling
  • Weighted Fair Queueing

IP Address

• IP address
  • 32-bit identifer associated with each host or router interface.
• Interface
  • Connection between host/router and phsycal link (interface is covered in link layer)

IP Datagram Structure

Subnet

• A subnet is a logical subdivision of networks by dividing IP addresses.
• The practice of dividing a network into two or more networks is called subnetting.
• The higher significant octet in IP address is subnetted, the higher hierarchy (larger regions) the network spans.

For example, a network can be divided into two networks (192.168.1.0/25 and 192.168.1.128/25):

Subnet Mask

Subnet mask splits an IP address into two parts:

• Network portion which identifies a network.
  • indicated by a sequence of 1s.
• Host portion which identifies a specific interface (device) within that network.
  • indicated by a sequence of 0s.

For example, the following subnet mask reveals which octet belongs to the network and the host portion:

CIDR (Classless Inter-Domain Routing)

CIDR facilitates the notation of an IP address with the subnet mask:

Host portion conventional interpretation:

• Network address: 192.168.100.0
  • is usually assigned to the router itself.
• Ordinary host range: 192.168.100.1–192.168.100.254
• Broadcast address: 192.168.100.255
  • is used to broadcast packets to local hosts simultanesouly (e.g., to find DHCP server).

How Forwarding Works?

Match and Forward based on Longest-prefix Matching

When deciding which output port a packet should be forwarded, compare the packet’s destination IP address with forwarding table entries, then find longest address prefix.

Given a forwarding table as below:

Two packets arrived with destination IP addresses as shown above, then:

• First packet is forwarded to interface 2.
• Second packet is forwarded to interface 1.

This fast match and forward is enabled by TCAM (ternary content-addressable memory)
- TCAM compares an arriving address against many table entries in parallel, allowing an extermely quick match.