mirror of
https://github.com/jackyzha0/quartz.git
synced 2025-12-28 07:14:05 -06:00
182 lines
5.4 KiB
Markdown
182 lines
5.4 KiB
Markdown
---
|
|
title: "19-network-layer-data-plane"
|
|
aliases:
|
|
tags:
|
|
- cosc203
|
|
- lecture
|
|
sr-due: 2022-11-20
|
|
sr-interval: 19
|
|
sr-ease: 250
|
|
---
|
|
|
|
- transport layer messages are divided into segments
|
|
- passed to network layer
|
|
- sender: encapsulates segments into IP packets, passes to link layer
|
|
- reciever: deliver segments to tranport layer protocol
|
|
- sender and reciever are connected by multple routers
|
|
- routers
|
|
- examine header fields in all IP packets passing through it
|
|
- moves IP packers from input ports to output ports to transfer packets along end-end path
|
|
- network layer protocols
|
|
- IP
|
|
- internet control message protocol (ICMP)
|
|
- use for troubleshooting type things
|
|
|
|
|
|
|
|
# Forwarding
|
|
forwarding
|
|
- move packets from a routers input link to appropriate output link
|
|
- local, per-router function
|
|
- function in data plane
|
|
|
|
routing
|
|
- determine route taken by packets from source to destination
|
|
- network-wide logic
|
|
- function in control plane
|
|
- chooses shortest path through routers
|
|
|
|
# network layer service model
|
|
best effort
|
|
- try its best to deliver packet
|
|
- does not guarantee successful delivery
|
|
- transport layer provides reliability
|
|
- no big deal if packet is lost
|
|
- does not guarantee timing or order of delivery
|
|
- each packet sent individually, possibly through different paths
|
|
- does not guarantee bandwidth available
|
|
- network layer uses packet switching which unlike socket switching does not reserve resources
|
|
|
|
# Routers
|
|
|
|
|
|
destination-based forwarding: foward based only on destination IP address
|
|
|
|
input port queueing: if datagrams arrive faster than fowarding rate into switch fabric
|
|
|
|
# IP addressing
|
|
- unreliable host-to-host communication protocol
|
|
- datagram formatting, IP addressing, packet switching
|
|
|
|
|
|
- ver: IP protocol versin number
|
|
- head len: header length in bytes
|
|
- type of service: diffserv(0:5) or ECN (6:7)
|
|
- length: total datagram length (bytes)
|
|
- time to live: remaining max hops (decremented at each router)
|
|
- upper layer: which protocol use at transport layer
|
|
- 16-bit id, flgs, fragment offset: fragmentation/reassembly
|
|
- header checksum
|
|
- source IP
|
|
- destination IP
|
|
- options: e.g., timestamp, record route taken
|
|
|
|
overhead:
|
|
- 20 bytes TCP
|
|
- 20 bytes IP
|
|
- = 40 bytes + app layer overhead for TCP+IP
|
|
|
|
|
|
interface: connection between host/router and physical link
|
|
- router have multiple interfaces
|
|
- host has one or two interfaces :e.g., ethernet and wireless
|
|
|
|
IPv4 address: 32-bit id associated with each host or router interface
|
|
|
|
- two main components
|
|
- network ID
|
|
- addresses in the same network have the same network ID
|
|
- host ID
|
|
- 5 classes
|
|
- 
|
|
|
|
## subnetting
|
|
- partitioning an IP network into multiple smaller network segments
|
|
- designate some high-order bits from host part as subnet ID
|
|
- 
|
|
- netmask
|
|
- a 32-bit number with all 1s for network part and all 0s for host part
|
|
|
|
|
|
- use bitwise and to find network part from address and mask
|
|
|
|
|
|
## fragmentation
|
|
- network links have an MTU (maximum transmission unit)
|
|
- different linktypes have different MTUs
|
|
- large UP datagrams are fragmented at routers
|
|
- one datagram become several
|
|
- reassemble only at destination
|
|
- IP header bits are used to identify fragments
|
|
|
|
|
|
header
|
|
- identifier
|
|
- same for all fragments
|
|
- flag: 3 bits
|
|
- 1st not used
|
|
- 2nd - do fragment
|
|
- 3rd - more fragment (0 for the last fragment)
|
|
- offset
|
|
- offset of the fragment in the packets data field (units of 8 bytes)
|
|
|
|
|
|
|
|
what needs to change in the header of fragment
|
|
- flag
|
|
- offset
|
|
- length
|
|
- ttl
|
|
- checksum
|
|
|
|
## special IPv4 Adresses and NAT
|
|
special addresses
|
|
- 127.0.0.1 localhost/loopback
|
|
- 10.0.0.0/8 private use
|
|
- 192.168.0.0/16 private use
|
|
|
|
NAT (network address translation): all devices in local network share just one IPv4 address
|
|
- allow multiple devices to share the same adresses
|
|
- help preserve the limited amount of IP address
|
|
- translate from private IP to public IP
|
|
- 
|
|
|
|
# IPv6
|
|
- inital motivation: IPv4 mostly used up
|
|
- additional motivation
|
|
- speed prcessing/forwarding: 40 byte fixed length header
|
|
- enable different network-layer treatment of "flows"
|
|
- better support for host mobility
|
|
- better support for security
|
|
|
|
datagram format
|
|
|
|
adressses
|
|
- 128 bits
|
|
- hexadecimal colon notation
|
|
- 12AB:0000:0000:CD30:0000:0000:0000:0000
|
|
- adress compression
|
|
- consecutive sections of seros are replaced with ::
|
|
- can be used only once per address
|
|
- e.g., 12AB:0000:0000:CD30::
|
|
|
|
CIDR
|
|
- classless interdomain routing
|
|
- method for allocating IP addresses and IP routing
|
|
- notation
|
|
- based of variable length subnet mask (VLSM)
|
|
- address format: IP address/x, where x is number of bits from left to right as network part
|
|
- 
|
|
|
|
extension headers: 
|
|
differencees: 
|
|
|
|
# transition from v4 to v6
|
|
- cant change all router at once
|
|
- will need mixed v4 and v6 routers
|
|
|
|
tunneling
|
|
- v6 datagram is carried as the payload in v4 datagram among v4 routers ("packet with a packet")
|
|
- used alot with 4G/5G
|
|
- 
|