Networking Concepts Quiz

Chapter 4: Network Layer: Data Plane

• This chapter details the data plane of the network layer.
• The provided slides are freely available for use, modification, and distribution under specific conditions
• Users should reference the source (the textbook) and the copyright information if using the slides.

Network Layer: Our Goals

• Understanding principles behind network layer services (focusing on the data plane).
• Network layer service models.
• Routing and forwarding.
• Router architecture.
• Internet architecture.
• Implementation aspects of the Internet: including IP protocol, NAT (Network Address Translation), and middleboxes (e.g., firewalls).

Network Layer: "Data Plane" Roadmap

• Network Layer Overview: encompassing data plane and control plane.
• Router features: including input ports, switching, output ports, buffer management, and scheduling mechanisms.
• IP (Internet Protocol) details: its datagram format, addressing schemes, and network address translation (NAT).
• IPv6 protocols: addressing translation
• Generalized Forwarding and SDN overview
• Middleboxes (network devices performing additional functions)

Network-layer Services and Protocols

• Network layer protocols function in all Internet devices.
• Data segments are encapsulated into datagrams
• Datagrams are passed to the link layer
• Segment delivery is completed by the transport layer
• Routers examine IP datagram header fields
• Datagrams are transferred from input ports to output ports along the end-to-end path.

Two Key Network-Layer Functions

• Forwarding: the process of moving packets from one router link to another based on look-up tables
• Routing: the process that the router uses to plan the packet's trip from source to destination, involves routing algorithms

Network Layer: Data Plane, Control Plane

• Data plane: local functions within a router determining how arriving datagrams are forwarded to router output port
• Control plane: network-wide logic determining the path that a datagram takes from source to destination host. Approaches include traditional routing algorithms (in routers) and software-defined networking (SDN) implementations (remote servers)

Per-router Control Plane

• Individual routing algorithm components in each router interact within the control plane.
• Values from the arriving packet header are used to determine actions.

Software-Defined Networking (SDN) Control Plane

• Remote controller (in a separate server) computes and installs the forwarding tables inside routers.

Network Service Model

• Service models for network layers; for individual datagrams; example services.
• Examples of services in flow of datagram: in-order datagram delivery, guaranteed minimum bandwidth to flow, restrictions on changes in inter-packet spacing

Network-layer Service Model QoS Guarantees

• QoS guarantees depend on the network architecture and service model. The Internet's best-effort service model doesn't guarantee bandwidth, loss, or order of delivery.

Reflections on Best-Effort Service

• Internet’s wide deployment is due to its simplicity.
• Real-time applications’ performance is acceptable with sufficient bandwidth provisioning.
• Replicated application-layer distributed services (e.g., data centers) contribute to the performance of real-time applications.
• Congestion control mechanisms improve the efficiency of elastic services.

Router Architecture Overview

• High-level view of a generic router architecture.
• Routing processor: responsible for routing/management in millisecond time frame.
• High-speed switching fabric: is the forwarding data plane (nanosecond time frame)
• Input/output ports: used for packet transmission

Input Port Functions

• Decentralized switching uses header field values to look up the appropriate output port using the forwarding table in input port memory.
• The goal is to process input packets at line speed.

Destination-based forwarding, Longest Prefix Matching

• Forwarding packets based on destination IP
• Use longest prefix matching for look-up tables

Longest Prefix Matching

• Use the longest matching address prefix when looking up entries in a routing table.
• Ternary Content Addressable Memories (TCAMs) are used to enable efficient lookup, regardless of table size.
• Cisco Catalyst routers use TCAMs for ~1 million routing entries.

Switching Fabrics

• Transferring packets from input links to output links.
• Switching rate = rate at which packets can be transferred from input to output ports
• Measured as multiples of input/output line rate
• Switching rate is desirable to be N multiplied by line rate in N inputs.

Switching via Memory

• First-generation routers used traditional computers with switching under CPU control.
• Packets were copied to system memory.
• Speed was limited by memory bandwidth.

Switching via a Bus

• Datagrams are transferred from input port memory to output port memory.
• Switching speed is limited by bus bandwidth.
• Example: 32 Gbps bus (Cisco 5600) is sufficient for access routers.

Switching via Interconnection Network

• Crossbar, Clos Networks, and other interconnection networks were initially designed for multiprocessors.
• They involve multiple stages of smaller switches, and exploit parallelism by fragmenting datagrams into cells, switching the cells, and reassembling datagrams at the exit.
• Cisco CRS routers use multiple switching planes for high switching capacity, using a 3-stage interconnection network, with up to hundreds of Tb/s switching capacity.

Input Port Queuing

• Queuing may occur if the switch fabric is slower than combined input ports. Delay may result from input buffer overflow.
• Head-of-the-line (HOL) blocking occurs when a datagram at the front of queue blocks others from moving forward.

Output Port Queuing

• Buffering required when datagrams arrive from fabric faster than link transmission rate
• Drop policy is important in this context: which datagrams are dropped if no buffers are free.
• Scheduling discipline decides queue order for transmission.

How Much Buffering?

• RFC 3439 suggests that average buffering should equal the typical round-trip time (RTT) times the link capacity (C).
• Increased buffering can increase delays in home routers.
• Long RTTs lead to poor real-time application performance and sluggish TCP response.

Buffer Management

• Drop: determining which packet to add when buffers are full.
• Tail drop: drops arriving packets when buffers are full.
• Priority: drop (or move) packets based on priority.
• Marking: marking packets to indicate congestion (e.g., ECN and RED).

Packet Scheduling: FCFS

• First-Come, First-Served. Packets are transmitted in the order they arrive at the output port. (FIFO- First-In, First-Out)
• Priority: packets are classified into priority queues and transmitted from highest priority queue
• Round Robin: server sequentially scans queues, servicing one packet from each class

Scheduling Policies: Weighted Fair Queuing (WFQ)

• Generalized Round Robin : each class, i , has weight, w, gets a weighted amount of service during each cycle (∑WI).
• Minimum bandwidth guarantee per class.

Sidebar: Network Neutrality

• Technical: How an ISP should share/allocate its resources (Packet Scheduling and buffer management).
• Social, Economic: Protecting free speech, encouraging innovation, and competition.
• Enforced legal rules and policies.
• Different countries have different "takes" on network neutrality.

Sidebar: Network Neutrality (FCC Order 2015)

• No Blocking
• No Throttling
• No Paid Prioritization

ISP: Telecommunications or Information Service?

• Title II: imposes "common carrier duties" on telecommunications services.
• Title I: applies to information services

IP Datagram Format

• Version number
• Header length
• Type of service
• IP identifier, flags, fragment offset
• Time to live, upper layer protocol, checksum
• Source/destination IP addresses
• Options
• Payload (variable length)

IP Addressing: Introduction

• IP address: 32-bit identifier.
• Interface: connection between host (or router), and physical link.

Subnets

• Subnet: devices in the same subnet have the common high-order bits in their IP addresses.
• Host part: the remaining low-order bits of the IP address

Subnets (Recipe, Addressing Structure, CIDR)

• Recipe: detach each interface from its host, and router, creating "islands" of isolated networks. Every island is a subnet.
• Structure: Subnet + host.
• CIDR (Classless Inter-Domain Routing): subnet portion of an address of arbitrary length, allowing more flexible use of address space

IP Addresses: How to Get One?

• Hard-coded (manual input by sysadmin)
• DHCP (Dynamic Host Configuration Protocol): from a server during system boot and network access

DHCP: Dynamic Host Configuration Protocol

• Host dynamically obtains IP address from a network server when it joins.
• Renewing leases, address reuse, mobile user support (joining/leaving network).
• DHCP Discover / Offer / Request / ACK messages exchanged.

DHCP: More Than IP Addresses

• Additional information DHCP may return:
• First-hop router's address
• DNS sever's address
• Network mask

DHCP: Example

• DHCP is used dynamically obtain IP address for devices connecting to the network (laptop).
• Messages are encapsulated in UDP, IP, and Ethernet. -DHCP server sends address: DHCP ack

IP Addresses: How to Get ISP Block of IP Address Space?

• ICANN (Internet Corporation for Assigned Names and Numbers) allocates blocks of addresses to ISPs via regional registries
• ISPs may distribute further across their network using subnet blocks, creating more addresses for individual organizations and devices.

Hierarchical Addressing: (Route Aggregation)

• Hierarchical addressing simplifies routing advertisement.

Hierarchical Addressing: (More Specific Routes)

• When an organization moves ISPs, more specific route aggregation is used.

IP Addressing: Last Words

• ICANN allocates IP blocks through regional registries (RRs) to ISPs. Management includes delegation of TLDs (.edu, .com).
• NAT and address space exhaustion are handled with NAT, which can be controversial
• IPv6 is used for address space extensibility.

IPv6 Motivation

• Motivation: 32-bit IPv4 space is fully allocated.
• 40-byte header.
• Enables different network-layer treatment of "flows."

IPv6 Datagram Format

• Priority/Flow Label
• Fixed length 40 byte header
• Source/destination addresses
• Payload

Transition From IPv4 to IPv6

• Gradual transition due to incompatibility between IPv4 and IPv6 routers
• Tunneling (carrying IPv6 datagrams inside IPv4 datagrams) enables interoperability

Tunneling and Encapsulation

• Tunneling is an important mechanism for transitioning from IPv4 to IPv6.

IPv6 Adoption

• Google and NIST are using IPv6 with increasing numbers of clients and domains respectively.
• 25 years and counting! Significant application-level advancements have occurred in this time frame

NAT: Network Address Translation

• Local network devices share a single network address (as seen by the outside world)
• The NAT router translates source and destination addresses.

Test your knowledge on various networking concepts, including addressing schemes, SDN, NAT, and the differences between forwarding and routing. This quiz covers essential topics related to the network layer and control plane functions. Perfect for those studying computer networking and information systems.