IT Hybrid Model

2016-11-09 19:03:55   0  举报





仅支持查看

AI智能生成

IT Hybrid Model是一种结合了公有云和私有云优势的云计算模型。在这种模型中，企业可以根据业务需求灵活地使用公有云和私有云资源。公有云提供了高度可扩展性和低成本的优势，而私有云则提供了更高的安全性和控制权。通过将这两种云环境结合起来，企业可以在需要时快速扩展资源，同时保持对数据和应用程序的安全控制。此外，IT Hybrid Model还可以帮助企业实现灾难恢复和业务连续性目标，因为它允许企业在两个不同的云环境中备份和恢复数据。总之，IT Hybrid Model为企业提供了一种灵活、安全且经济高效的云计算解决方案。

Internet

作者其他创作

大纲/内容

5. Application

1. Physical

Units

Speed of Light 3*10^8

Speed of electrical signal in copper wire: 2 * 10^8

Transmission Rate bit/sec or byte/sec

Propagation Delay = Length of Channel/Speed of Signals

Transmission Delay = Bits of per frame/Speed of Signals

Latency = P-Delay + T-Delay

NO NEED TO MULTIPLY BY 2 FOR P-Delay!!!!!

Medium

Wireless

Mechanism

Radio wave propagation

Naturally supports broadcasting over a region

Naturally supports mobility enabled network

Easy and inexpensive to deploy

Potential collision, need to regulate

Data rates vary greatly as signals are unstable

Wire

Fiber Optics

3 components: light source, transmission medium, detector

enormous bandwidth (THz) and tiny signal loss – hence high data rates over long distances

much thinner and lighter than metal wires, installation and maintenence can be more expensive/difficult

Optical fibre/fibre optics is fragile and requires more protection around the cable compared to copper.

Wavelength * Frequency = Speed of Light, lambda * miu = c

Example: SONET

Copper Wire

Coaxial Cable

Better shielding than twisted pair = higher speeds over greater distances

Carry either analogue or digital signals(携带模拟或数字信号)

Still widely used for cable TV/Internet

Twisted Pairs

Bandwidth dependent on distance, wire quality/density

No specialist skills required

Easy to engineer a fixed data rate over point-to-point links

Can be expensive to deploy, esp. over distances

Doesn't readily support mobility or broadcast

Enormous bandwidth over long distances

Satellite (belongs to wireless)

Can rapidly set up anywhere/anytime communications (after satellites have been launched)

Can broadcast to large regions

Drawback: Limited bandwidth and interference to manage, high delay

Examples

high bandwidth but also high latency

Satellite links

Trans-continent optic fibres

low bandwidth and low latency

56-kbps modem/telephony modem

low-end personal area wireless network(e.g. Zigbee)

Link Model

Symbol Rate (Baud Rate)

Number of signal changes per second

one signal = logV bits, V is the level divided

Baud Rate = signal rate = bit rate / logV = B, the unit is times/second, namely, kHertz, DIFFERENT FROM NYQUIST THEOREM!!!

Link Terminology

Full Duplex

Half Duplex

Simplex

Multiplexing

Multiple digital signals are combined into one signal over a shared medium.

The aim is to share an expensive resource, make use of the channel at the same time with multiple signals.

Time/Space/Frequency

Data Communication Theory

Transform continuous signals into digital values: Sampling the amplitude values of the signal

Nyquist Theorem

Max. data rate = 2B log2V bits/sec (here max means frequency is fixed, if we fix B and V, then it becomes minimum data rate needed)

B for bandwidth(Hertz), Hz stands for number of samples we collect per second, V for levels divided. Bandwidth can be used to represent various units, including samples/sec, baud/sec or bits/sec, here we use samples/sec, which is equal to Hertz

If a signal has bandwidth B, then the signal can be fully reconstructed by sampling with 2B rate

namely, f[sampling] > 2B.

If signal has V levels, each symbol can be represented by log2V bits.

Nyquist Rate is the sampling rate needed to prevent aliasing, Nyquist frequency is the minimum frequency we have given a sampling rate to reconstruct the signal

Shannon Theorem

Max. data rate = B log2(1 + S/N) bits/sec, S for signal strength, N for noise stength

Digital Modulation

Modulation Types

Frequency Modulation

Interval of X axis between any two points

Amplitude Modulation

Interval of Y axis between any two points

Phase Modulation (Passband Transmission)

Distance of any point from itself to origin

QAM 16 and QAM 64

Shape is like Lattice

Time Modulation

Blocks of Band and Guard Band

Transmission Calculation

Week 2 Slides 8-11

2. Data Link

Mechanism

Data Link layer accepts packets from the network layer, and encapsulates them into frames that it sends using the physical layer; reception is the opposite process

Week 3 Slides 5

Focus on local delivery, addressing, and media arbitration.

Type

Connection-oriented/Connectionless

Acknowledged/Unacknowledged

Whether to notify the sender with msg of delivery

Example

Con-less/UnACK

LAN

For example, it cannot determine the status of the destination adapter to ensure that it can receive a frame.

It cannot ensure that bandwidth is available throughout the transmission.

Con-less/ACK

Wireless

No logical connections used, but each frame sent is individually acknowledged.

If it has not arrived within a specified time interval, it can be sent again.

Con-Ori/ACK

WAN, VPN

It guarantees that each frame is received exactly once and that all frames are received in the right order.

Checksum

A checksum is a bit string that is computed from the contents of a frame and attached to the end of the frame as a parity

CRC

Cyclical Redundancy Checksum - polynomial code used to compute a checksum for a block and attached to block in transmission

Provide some level of reliability over the unreliable physical layer, Transmission can be reliable or unreliable

Why there still needs error checking provided that there's already checksum in datalink layer?

Many data-link protocols do not have acknowledgments of successful frame reception and acceptance

Some data-link protocols might not even have any form of checksum to check for transmission errors.

In those cases, higher-level protocols must provide flow control, error checking, and acknowledgments and retransmission.

Why put on trailer rather than header?

The CRC is computed during transmission and appended to the output stream as soon as the last bit goes out onto the wire. If the CRC were in the header, it would be necessary to make a pass over the frame to compute the CRC before transmitting. This would require each byte to be handled twice – once for checksumming and once for transmitting. Using the trailer cuts the work in half.

Hamming Distance

Basically checksum is a heap, wherein it's like using binary search to target the actual position of error bits

Framing Method

Character count

one !BYTE! (value ranges from 0 to 127) acts as counter placing at the head of frame

Flag bytes with byte stuffing

start and end with FLAG on head/trail, for each FLAG data add an ESC, for each ESC data add and ESC

Start and end flags with bit stuffing

every time got 5 consecutive 1s stuff with a 0, once 6 consecutive 1s are detected, stop receiving and start validation

Each frame begins and ends with a special bit pattern e.g. 01111110

Physical layer coding violations

The encoding of bits as signals often includes redundancy to help the receiver.

This redundancy means that some signals will not occur in regular data. For example, in the 4B/5B line code 4 data bits are mapped to 5 signal bits to ensure sufficient bit transitions. This means that 16 out of the 32 signal possibilities are not used.

We can use some reserved signals to indicate the start and end of frames. In effect, we are using 'coding violations' to delimit frames.

Framing is the method used by data link layer to break raw bit stream into discrete unit and generate checksum.The primary purpose therefore, of framing, is to provide some level of redundancy over the unreliable physical layer.

Error Control

How can we ensure all frames are delivered to the network layer and delivered in canonical order ?

Control Frames

Positive and Negative acknowledgments sent by receiver

Timers

Timeout values independent of acknowledgments.

Sequencing

Sequence number helps to resolve sort orders in the case of multiple retransmissions.

Error Handling

Codeword/Hamming Code

TextBook page 205

Hamming Distance - the degree of similarity between two codewords

On channels that are highly reliable, such as fiber, it is cheaper to use an error-detecting code and just retransmit the occasional block found to be faulty. However, on channels such as wireless links that make many errors, it is better to add redundancy to each block so that the receiver is able to figure out what the originally transmitted block was.

The receiver actually has no idea how to verify the received code, but only compute the distance between each possible code(it has a code book) and the received code, and choose the one mostly similar to it as correction code.

Include enough redundancy to allow receiver to determine an error occurred and request retransmission (error-detecting)

To reliably detect d errors, you need a distance d + 1 average distance of codes because with such a code there is no way that d single-bit errors can change a valid codeword into another valid codeword. This means that all part of the error could be detected, but actually if you just want to know if there is an error, the distance needed could be even smaller, but in that case you just detected PART of the errors (no guarantee of finding out all errors, so could have missed some).

Include enough information in frames to allow reconstruction/deduction of original content (error-correcting)

Similarly, to correct d errors, you need a distance 2d + 1 code because that way the legal codewords are so far apart that even with d changes the original codeword is still closer than any other codeword.

Example

Consider a code with only 4 valid codewords

0000000000, 0000011111, 1111100000, and 1111111111

This code has a distance of 5, which means that it can correct double errors or detect quadruple errors.

If the codeword 0000000111 arrives and we expect only single- or double-bit errors, the receiver will know that the original must have been 0000011111.

If, however, a triple error changes 0000000000 into 0000000111, the error will not be corrected properly.

Flow Control

Regulating the flow of data so as to resolve fast senders and slow receivers problem (slow receiver not swamped by fast sender).

Design Principles

1. Provide Service to network layer

2. Handling transmission errors (only transmission, cannot handle loss caused by mediator or validate them/ resolve them)

3.1. Regulating the flow of data so that slow receivers are not swamped
by fast senders.(Data flow regulation)

Principles to control when sender can send next frame (Fast sender vs low receiver problem resolved)

When the sender sends too much data the receiver will respond a msg to tell it send less next time()

3.2. ARQ (Automatic Repeat reQuest) Protocols (Flow Control)

Concept of ARQ

Receiver acks frames that are correctly delivered

Sender sets timer and resends frame if no ack

Stop and Wait Protocol

Mechanism

Uses ACK and timeout to achieve a reliable transmission over an unreliable service

Fast sender, slow receiver not appropriate, need buffer space on receiver end

Timeout, resend, duplicate, ignore, corrupted, ignore, success, accept

Application

In Noisy Channel

Frames can be lost either entirely or partially

Every time a frame arrives, check if there's any duplicate/error (check its sequence number, e.g. if the received seq is m, it implies that m-1 has already been successfully received by receiver and acknowledgement has also been received by sender, other wise the seq would be m-1 instead of m), if not, it is accepted and passed to network layer when an ACK is generated and sent to sender, if there is, ignore such frame and send back an NegativeACK ===> this is how ARQ works

If there isn't, mark this frame and send back an ACK

Requires timeout function to determine arrival or non-arrival of complete frames

In Error-free Channel

SEND - ACK - SEND - ACK

Link Utilization = Tf / (Tf + 2Tp) = (L/B ) / (L/B + 2Tp) = L/ (L+ 2Tp B).

Tf: transmission time, Tp: propagation time, L: packet length in bits, B: bandwidth of channel

Sliding Window Protocol

Overview

Senders don't need to wait for acknowledgement for each frame before sending next frame

Data is commonly transmitted in both directions simultaneously

Send acknowledgments by piggy-backing

Utilisation usually 100%.

One-bit Sliding Window Protocol

Problems can arise if both sides send an initial packet-synchronization issues

Week 3 Slides 34

Mechanism is almost same as stop and wait protocol

Selective Repeat

NAK (negative ack) causes sender retransmission of a missing frame before a timeout re-sends window

And for sure there's also positive ACK, which simply asks for next frame

Buffers good frames after an error point, and relies on sender to resend oldest unacknowledged frames

Go-Back-N

Discards all subsequent frames from error point, sending no acknowledgment, until the next frame in sequence.

There's no need to keep a buffer as the sending is moving sequentially, once an error gets detected, receiver simply goes back N frames and prompt sender to resend.

PiggyBagging In This Context

Now consider building duplex link using two simplex links

We can intermix "forward" (for data between A and B) and "reverse" (acks between A and B) using the
same channel.

Thus instead of four channels, we can manage duplex communication with two channels.
This idea is piggybacking.

piggybacked data when returning acks

Sub-Layers

MAC (lower layer) - Medium Access Control

1. Framing/De-Framing

2. Handle transmission errors of frames

3. Decide when medium is used

Protocols

Point to Point Protocol (PPP)

PPP (Point-to-Point Protocol) is a general method for delivering packets across links

Framing uses a flag (0x7E) and byte stuffing

Connectionless un-acknowledged service

Errors are detected with a checksum

Week 3 Slides 43

Differentiation

On point to point networks, there are only singular sender and receiver pairs, eliminating transmission contention.

On broadcast networks, determining right to transmit is a complex problem.

Medium Access Control (MAC) sub-layer is used to assist in resolving transmission conflicts

ADSL

Week 3 Slides 46

Channel Allocation

Static Channel Allocation

Frequency Division Multiplexing (FDM)

Used in TV and Radios

Arbitrary division of a channel into segments and each user allocated a dedicated segment for transmission

Inefficient when number of senders > allocated segments, number of senders is not static or when traffic is busy.

Dynamic Channel Allocation

Time Division Multiplexing (TDM)

2G network

Channel segmentation is dynamic, segment allocation is dynamic.

Contention

ALOHA

Users transmit frames whenever they have data; users retry after a random period of time for any collisions came across

When traffic is low, ALOHA can get packets sent as soon as the RANDOM WAIT(distinguish from 1-persistent CSMA, which keeps listening so there's no random wait) terminates

When traffic is high, ALOHA leads to more collisions

Unlike the ARPANET where each node could only talk directly to a node at the other end of a wire or satellite circuit, in ALOHAnet all client nodes communicate with the hub on the same frequency.

Slotted ALOHA

Allows the users to start sending only at the beginning of defined slots. Increase efficiency of pure ALOHA by reducing possibility of collisions

When traffic is low, Slotted ALOHA still need to wait till its time slot even if random waiting time terminates

When traffic is high, Slotted ALOHA reduces chances of collision

CSMA/CD

1-persistent

Wait (polling) until channel idle; transmit one frame and check collisions

if collision, wait for a random time and repeat

non-persistent

If channel busy, wait random period and check again; if not, start transmitting = ALOHA

p-persistent

with a probability p it chooses to either 1-persistent or non-persistent

The lower the load is, the higher the efficiency is with 1-persistent CSMA cuz it keeps listening to the channel.

(Less likely to be examined) Binary Exponential Backoff Algorithm (typically used in Ethernet)

Randomly choose one time slot to send data if detects any collision, double the window size in next retransmission

Each station randomly pick a time slot within the range of its own window size every time.

Collision Free

BitMap

1 bit per station overhead

N stations, each contention period consists of N time slots, first round each station insert 1 bit if it is intended to send data

this votes will be shown to all stations and they know in which order to send data later(numerical order, from small to large)

Efficiency d/(d+N) at low load, d/(d+1) at high load.

Binary Countdown

Each station has an address associated, e.g. 1010 is an ADDRESS of such station

If this station wants to send data, it broadcast bit by bit with its address

whenever the first digit it wants to take possession of but finds to be non-0, it stop broadcasting its address (by OR operation, not XOR, 1 OR 1 is 1, 1 OR 0 is 1)

If multiple stations are broadcasting the same digit with 1, there's a tie, they keep moving until one of them get another 1 in advance to the other

By this way the overhead is reduced

Efficiency can reach up to 100% if the frame format has been cleverly chosen so that the sender's address is the first field in the frame, even these logN bits are not wasted, generally d/(d+logN).

Limited Contention

Adaptive Tree Walk

We have a number of known slots to use, so we start with assigning a time slot to the root(top)

all computers start to contend, when they find the slot is in use, go down to another slot and check again, until there is no lower slot

If we cannot break them up any more, choose the leftmost one based on DFS algorithm

Wireless LAN

MACA

Sender asks receiver to transmit short control frame

Stations near receiver hear control frame

Sender can then transmit data to receiver

MACAW

Multiple Access with Collision Avoidance for Wireless

Hidden Terminal and Exposed Terminal Problems

A -> B <- C, A and C think B is free, but actually it's not, D<-A - B -> C, A and B think channel is full, but actually they can transmit data simultaneously

When a station is in the range of two transmitters or relays, interference affects signal reception

Classic Ethernet (A typical implementation of LAN)

Week 4 Slides 31

MAC Address

A 48-bit number encoded in the frame

eg 00:02:2D:66:7C:2C

LLC (upper layer)

1. Handle packets from Internet Layer/Pack frames into packets(fragmentation)

2. Decide Conncection-Oriented/Connectionless, ACK/ un-ACK

3. Regulate Flow

Sub-Layers are only divided in LAN because the medium is shared (in WAN it's point to point, no shared medium)

3. Internet

Differences between Transport layer and Network layer

1. The Internet Protocol is designed to provide an unreliable (not guaranteed) packet service across the network (must utilize an appropriate Transport Layer protocol to provide a reliable stream, e.g. delayed/ duplicated packet).

2. Services of transport layer should be independent of router technologies (running on hosts) while network layer code runs almost entirely on router

3. Transport layer should be shielded from number, type and topology of routers

What is inter-networking?

1. Multiple LAN connected through routers (if there's no router then it is not an inter-network)

2. Different packets may be routed to the same destination via different paths, resulting in out-of-order sequencing at the receiver.

Vitual Circuit & Datagram

check below, General

Internetworking

A TCP/IP datagram is packet-switched to its destination individually. Each frame's header contains a globally significant switching address, VCI/full address dependent on routing mechanism.

This address allows a routing decision to be made each time the packet is forwarded.

The packets being sent over TCP/IP are simply put on the wire and routed.

Allows flexibility in routing around obstructions.

But network performance can vary a great deal depending on conditions at the routers and on the degree of network traffic at any given time.

IP Packet

IPv4 Frame Structure

Header

Version: IPv4 or IPv6

IHL(Header Length)

Type

Total Length: maximum length 65535 bytes

Identification

DF: Don't Fragment byte

MF: More Fragment byte - are there more or is this the last one ?

Fragment offset: where in the datagram the current fragment belongs

TTL: limits packet lifetimes - hops or seconds

Protocol: TCP, UDP, others ...

Header Checksum: verifies the header only

The checksum ensures that the information in a received header is accurate.

However, IP does not attempt to detect errors that may have occurred to the data in each packet.

Source Address: IP - host/network

Destination Address: IP - host/network

Options: eg security, strict vs loose source routing, record route, timestamp

Fragmentation and Reassembly

Large packets need to be routed through a network whose maximum packet size is too small.

Reassemble could appear either at the destination or at each of the network's gateway, which is breaking a long packet or message into smaller packets. (Textbook page 434)

Solution

Transparent

1. when your packet travels to one network the gateway will fragment it
2. reassemble at another gateway
3. send to the next network until destination is reached

packets fragmented / reassembled in each network

By not allowing some fragments to follow one route to the ultimate destination and other fragments a disjoint route, some performance may be lost. More significant is the amount of work that the router may have to do. It may need to buffer the fragments as they arrive, and decide when to throw them away if not all of the fragments arrive.

Non-Transparent

1. when your packet travels to one network the gateway will fragment it
2. send THROUGH another gateway to the other end
3. resesemble at destination

fragments are reassembled at destination

The main advantage of nontransparent fragmentation is that it requires routers to do less work. IP works this way. A complete design requires that the fragments be numbered in such a way that the original data stream can be reconstructed.

IP Address

Mask of Subnet

Subnetting allows networks to be split into several parts for internal uses whilst acting like a single network for external use

Routing Table

Class

ABCDE, 1 bit, 2 bits, 3 bits, 4 bits, 5 bits reserved identifier respectively

Private IP Range

10.0.0.0/8 (16,777,216 hosts)

172.16.0.0/12 (1,048,576 hosts)

192.168.0.0 /16 (65,536 hosts)

Tunneling

Tunneling is used when the source and destination have different networks in between. Source Packets are encapsulated over the packets in the connecting network. e.g. Tunneling IPv6 through IPv4.

Routing Algorithm

Non-Adaptive Algorithm(static method)

Dijkstra

Flooding

Adaptive Algorithm(change according to the state)

Distance Vector Routing (Bellman-Ford Algorithm)

Dynamic programming, keep 'updating' the shortest path to each node by receiving the msgs from neighbours

1. your neighbour will tell you the its distances to its neighbours, and you also know your distance to such neighbour

2. based on this table and your knowledge, update your table of distances to each node (those passed from your neighbours)

Link State Routing

More computation but simpler dynamics

Widely used in practice

Nodes knows only the distance to their neighbours, not topology

1. all nodes floods to neighbours by sending each distance of neighbours, then each node has a picture of entire topology

2. use Dijistra to create a forwarding table, indicating for each destination, which node is the NEXT node.

3. done

Hierarchical Routing

Each router knows everything about other routers in its region but nothing about routers in other regions

Routers which connect to two regions act as exchange points for routing decisions

Broadcast Routing

When a broadcast packet arrives at the router, router checks if the line it came from is also the line it used to send packet to source, if so, broadcasts the packet to all other output lines it has, otherwise discards the packet.

Multicast Routing

Internetwork Routing

Protocols

IP as the primary data transport protocol in this layer, but could also be ICMP, ARP, RARP

ICMP (Internet Control Messaging Protocol)

ICMP is a companion to IP that returns error info

Required,and used in many ways, e.g., for traceroute

ARP (Address Resolution Protocol)

finds MAC address of a local IP address (pls see Extension Diagram for more info)

Glue that is needed to send any IP packets

Host queries an address and the owner replies

DHCP (Dynamic Host Configuration Protocol)

Assigns a local IP address to a host

Gets host started by automatically configuring it

Host sends request to server, which grants a lease

When you bring your laptop to school it will know the router it connects to, the subnet mask, etc
and when you bring it back home, it will dynamically change this information

4. Transport

Purpose

Integrate data and distribute them into processes

Transport Layer code runs entirely on hosts, Network Layer on routers

Transport layer in TPDU, or segment, Network layer in packets

Reliable Connection Establishment

Connection Release

Asymmetric Disconnection

Either party can issue a DISCONNECT, which results in DISCONNECT TPDU and transmission ends in both directions

may result in data loss

Symmetric Disconnection

Both parties issue DISCONNECT, closing only one direction at a time -allows flexibility to remain in receive mode

Strategies

Again use Three way handshake

Finite retry

you try N times if there is no response then the host will end this process

Timeouts

Addressing

Addressing in transport layer is typically done using Transport Service Access Points (TSAPs)

Usually this is done via port 80

Transport Primitives

SOCKET

BIND

LISTEN

CONNECT

SEND

RECEIVE

User Datagram Protocol(UDP)

Connectionless, no sequence order in packet headers, no flow control, individual packets pass through different routes

ADV

Multiplexing/de-multiplexing capabilities (N-clients sending at the mean time), thus efficient in tranmission.

Appropriate for multicasting to recipients without having to keep states at senders.

DSV

does not include support for flow control, error control or retransmission of bad segments

Main advantage of using UDP over raw IP packets is the ability to specify ports for source and destination pairs.

messages wrapped in a single packet, no possibility to be dissembled or ressembled during transmission

when application requires efficiency over accuracy, UDP good choice

Usage

RPC (Remote Procedure Control)

Voice over IP, online games, and many protocols built on top of the Real Time Streaming Protocol.

Transmission Control Protocol(TCP)

Features

Connection-oriented, packets travelling between destinations all use the same route

Full duplex - data in both directions simultaneously

Point to point - exact pairs of senders and receivers

Byte streams, not message streams - message boundaries are not preserved

Buffer capable - TCP entity can choose to buffer or not prior to sending depending on the context

Allows segments to be disassembled and reassembled during transmission

Requirements needed to establish reliable connection

Packets in the network have a maximum life time

Sequence numbers are used by each end of connection, such that they will not be reused during maximum packet lifetime

Use 3-way handshake to exchange initial sequence numbers between both ends of connection

Policy choices used that can affect network congestion

Retransmission policy

Out-of-order caching policy

Acknowledgment policy

Negative ack can be used to avoid timeouts

Flow control policy

Timeout determination

Approach

Three Way Handshake

Sliding Window Protocol

Congestion Window

At connection establishment, a suitable window size is chosen by the receiver based on its buffer size

If the sender is constrained to this size, then congestion problems will not occur due to buffer overflow at the receiver itself, but may still occur due to congestion within the network

Slow Start

Address fast link to slow link problem (the limit of different networks may be different)

Refer to Extension

General

Interface (between two layers)

Datagram network (packet switching)

connection-less/properties

packets injected into subnet independently

packets individually routed to destination

each packet contains full source and destination address

get routed through variable paths

routers do not hold state information about connections

difficult to implement QoS

probably faster than virtual circuit

post office model

Circuit-switched network (circuit switching)

connection-oriented

packet routed through fixed path

telephony model

Virtual-circuit network (packet switching)

connection-oriented/properties

require additional setup phase

each packet contains a VC identifier

each VC requires router table space per connection

fixed path

good QoS because resources can be reserved in advance (bandwidth, cpu cycles, memories, etc)

less overhead for storing full addresses of source and destination after connection has been set up.

a combination of Datagram network and Circuit-switched network, so there's an extra set up phase during connection(but after then the packet transmission would be faster)

the set-up phase of VC is basically sending out sentry packets until one reaches destination, and then it goes back with router port number to last router

last router then knows which router should be marked when next time a packet arrives, keep tracing back to the original source

whenever a packets arrives after then, router will append an extra VCI (Virtual circuit identifier) to its header, and the packet thereby knows where to go, states are kept for each connection

Overhead is not as heavy as CSN, but the connection is also stable

Example

Data-link layer and network layer virtual circuit protocols are based on connection-oriented packet switching.

Connection-oriented transport layer datalink protocols such as TCP may rely on a connectionless packet switching network layer protocol such as IP.

IP is connection-less but it is able to support TCP as virtual circuit because segments have numbering to identify their order.

Connection-oriented 理解而非记忆例子

Reliable message stream

Sequence of pages

Reliable byte stream

Remote login

Unreliable connection

Digitized voice

Connectionless

Unreliable datagram

Electronic junk mail

Acknowledged datagram

Registered mail

Request-reply

Database query

File Transfer (FTP), Remote Login (SSH), Web Surfing (HTTP) and Email (SMTP/POP3) should be connection-oriented, these data should not be lost.

POS terminals, video streaming and audio streaming, P2P, VPN could be connection-less, where we care performance over packet loss.

Structrue

Frame-Packet-Segment

https://www.bing.com/images/search?view=detailV2&ccid=MLJ33eH8&id=E8500A13E6BFDA773CFB4988C5C8CD9C128FC146&thid=OIP.MLJ33eH8vQU24ppjdtpgDgEsCt&q=Link+Data+Frame+Packet+Segment&simid=608042060692785129&selectedIndex=5&ajaxhist=0