Computer Networks

Voice Over Internet Protocol (VoIP) 

Introduction to Computer Networks Direct link networks  Internetworking  End-to-end protocols 



Voice over Internet Protocol (VoIP) 

Voice quality Transport



Network QoS





Multimedia Streaming EE493Q: Digital Speech Processing

Layered Architecture Application

Network

FTP

HTTP

TCP

SMTP

UDP

IP

Data link

Physical

TELNET

NET1

NET2

……

EE493Q: Digital Speech Processing

NETK

Local Area Networks (LAN) 

Ethernet (IEEE 802.3) 

 

Rooted in an early packet radio network called ALOHA developed at the University of Hawaii and developed by Xerox Palo Alto Research Center in 1970s The most successful LAN technology in the last twenty years Carrier Sense Multiple Access with Collision Detection (CSMA/CD)

FDDI (IEEE 802.5)  Wireless (IEEE 802.11) 

EE493Q: Digital Speech Processing

What is an Internetwork (Internet)? Internet refers to an arbitrary collection of networks interconnected to provide some sort of host-to-host packet delivery service  How is internetwork different from networks? 





Network often refers to either a directly connected or a switched network (e.g., 802.5, Ethernet or ATM) → “physical network” An internetwork is an interconnected collection of networks → “logical network”

EE493Q: Digital Speech Processing

Example Network 2 (Ethernet)

Network 1 (Ethernet)

R2

R1 Network 3 (FDDI) R3 Rn=Router FDDI=Fiber Distributed Data Interface EE493Q: Digital Speech Processing

Network 4 (wireless)

Internet Protocol 

Invented by Kahn and Cerf in 1970s 





The best case study of a scalable and heterogeneous networking protocol

A datagram is a type of packet that is sent in a connectionless manner over a network IP service model is sometimes called “Best Effort” – i.e., although IP makes every effort to deliver datagrams, it makes no guarantees  

Addressing scheme: identify all hosts in the internet Datagram (connectionless) model of data delivery

EE493Q: Digital Speech Processing

Why Best Effort? It is the simplest service you could ask from an internetwork and it keeps the routers as simple as possible  The ability of IP to “run over anything”: many technologies IP runs today did not exist when IP was invented  Best effort delivery does not just mean that packets can get lost 

 

They can be delivered out of order The same packet can be delivered more than once EE493Q: Digital Speech Processing

Packet Format in IPv4

The maximum size of an IP datagram is 65535 bytes (fragmentation and reassembly might be needed when physical networks do not support such long packets) EE493Q: Digital Speech Processing

Fragmentation and Reassembly 

Why are they needed? 

Heterogeneous collection of networks have different packet-size limits – e.g., up to 1500 bytes for Ethernet and up to 4500 bytes for FDDI

All fragments after fragmentation will be assigned a unique IDENT field in the second word, which enables them to be reassembled at the host  IP does not attempt to recover from missing fragments 

EE493Q: Digital Speech Processing

Global Addresses 

IP addresses are hierarchical  



Network part: identifies the network to which the host is attached Host part: identifies each host uniquely on that particular network

IPv4 addresses (32bits)     

Class Class Class Class Class

A: 0 + network(7 bits) + host (24 bits) B: 10+ network(14 bits)+host (16 bits) C: 110 + network(21 bits)+host (8 bits) D: specify a multicast group E: unused EE493Q: Digital Speech Processing

Examples

EE493Q: Digital Speech Processing

Datagram Forwarding in IP  





Every IP datagram contains the IP address of the destination host The network part of an IP address uniquely identifies a single physical network that is part of the larger internet All hosts and routers that share the same network part of their address are connected to the same physical network and can communicate with each other by sending frames over that network Every physical network that is part of the Internet has at least one router that is also connected to at least one other physical network EE493Q: Digital Speech Processing

Bridges, Switches and Routers 

They all forward messages from one to another. The differences are   



Bridges are link-level nodes: forward frames within an extended local area network (LAN) Switches are network-level nodes: forward packets within a packet-switched network Routers are internetwork-level nodes: forward datagrams to implement internet

However, the distinction could become vague in some cases (e.g., multiport bridge) EE493Q: Digital Speech Processing

Ethernet Address and IP Address Ethernet addresses (48 bits) are configured into the network adaptor by the manufacturer  However, IP addresses need to be reconfigurable because a host might be connected to different networks (e.g., travel with your laptop to a conference)  Dynamic Host Configuration Protocol (DHCP) improves the manageability of a network 

EE493Q: Digital Speech Processing

Routing and Forwarding Forwarding table is used when a packet is being forwarded and so must contain enough information to accomplish the forwarding function  Routing table is the table that is built up by the routing algorithms as a precursor to building the forward table  Routing is among the most difficult technical issues in computer networks. 

EE493Q: Digital Speech Processing

Routing as a Problem of Graph Theory

B 5

6

D

4

1

2

E

A 1

9 C

1 3

G

1 7 F

EE493Q: Digital Speech Processing

Transport (End-to-End) Protocols 

Objectives of transport layers       

Guarantee message delivery Deliver the message in the same order they are sent Deliver at most one copy of the message Support arbitrarily large messages Support synchronization between sender and receiver Allow receiver to apply flow control to the sender Support multiple application processes at each host EE493Q: Digital Speech Processing

Recall: Limitations of networks 

Networks may     



Drop messages Reorder messages Deliver duplicate copies of a given message Limit messages to some finite size Deliver messages after an arbitrarily long delay

Two classes of protocols are most popular  

User Datagram Protocol (UDP) Transmission Control Protocol (TCP) EE493Q: Digital Speech Processing

Simple Demultiplexer: UDP

UDP Header Format UDP does not provide the reliability and ordering guarantees that TCP does Without the overhead of checking if every packet actually arrived, UDP is faster and more efficient for many lightweight or time-sensitive purposes. EE493Q: Digital Speech Processing

Reliable Byte Stream: TCP 



A virtual circuit (VC) is a communications arrangement in which data from a source user may be passed to a destination user over more than one real communications circuit during a single period of communication, but the switching is hidden from the users The Transmission Control Protocol (TCP) is a virtual circuit protocol that is one of the core protocols of the Internet protocol suite. Using TCP, applications on networked hosts can create connections to one another, over which they can exchange data in packets. The protocol guarantees reliable and in-order delivery of data from sender to receiver. EE493Q: Digital Speech Processing

Voice Over Internet Protocol (VoIP) 

Introduction to Computer Networks Direct link networks  Internetworking  End-to-end protocols 



Voice over Internet Protocol (VoIP) 

Voice quality Transport



Network QoS





Multimedia Streaming EE493Q: Digital Speech Processing

VoIP and Its History   



VoIP refers to the routing of voice conversations over the Internet or through any other IP-based network. The Network Voice Protocol (NVP) research sponsored by DARPA in 1973 was the ancestor of VoIP It was in 1995 that the first Internet Phone Software appeared; The breakthrough in VoIP history came when hardware manufacturers such as Cisco Systems and Nortel started producing VoIP equipment that was capable of switching A major development starting in 2004 has been the introduction of mass-market VoIP services over broadband Internet access services, in which subscribers make and receive calls as they would over the PSTN

EE493Q: Digital Speech Processing

PSTN: Public switched telephone network

VoIP Diagram

PBX: Public branch exchange SIP: Session initiation protocol

EE493Q: Digital Speech Processing

Business Model Some cost savings are due to utilizing a single network to carry voice and data, especially where users have existing underutilized network capacity they can use for VoIP at no additional cost.  VoIP to VoIP phone calls on any provider are typically free, whilst VoIP to PSTN calls generally costs the VoIP user.  Why eBay Is Buying Skype? 

EE493Q: Digital Speech Processing

Nomenclature        

PSTN: Public Switched Telephone Network PBX: Public Branch eXchange SIP: Session Initiation Protocol TCP: Transmission Control Protocol RTP: Real-time Transport Protocol UDP: User Datagram Protocol RSVP: ReSerVation setup protocol QoS: Quality of Service EE493Q: Digital Speech Processing

Voice Quality 

Factors affecting voice quality     



Speech codec Echo control Packet loss Delay and delay variation (jitter) Network design

Speech quality assessment    

Mean opinion score (MOS) The effect of environmental noise The effect of channel degradation (e.g., packet loss) The effect of tandem encoding/decoding EE493Q: Digital Speech Processing

Speech Coding Algorithms

EE493Q: Digital Speech Processing

MOS Comparison

EE493Q: Digital Speech Processing

Concatenation and Transcoding

EE493Q: Digital Speech Processing

Transport 

Network protocols 









IP: a connectionless best effort network communication protocol TCP: Use acknowledgement and retransmission to ensure packet receipt UDP: Unreliable connectionless delivery service using IP to transport messages RTP: used in conjunction with UDP to provide end-to-end network transport functions

VoIP uses RTP/UDP/IP instead of TCP/IP EE493Q: Digital Speech Processing

Bandwidth Efficiency 

Transmission is not always free, therefore transmission efficiency has its merit 





Speech coding/compression algorithms: coding delay and packetization delay Silence/voice detection (e.g., Appendix B of ITU-T Recommendation G.729): comfort noise is preferred instead of absolute silent background Header compression: 40bytes → 2-7 bytes (header without compression consumes a significant portion of bandwidth) EE493Q: Digital Speech Processing

Delay 



Transmission delay consists of codec processing delay and propagation delay G.131 requirements   



<150ms: acceptable for most user applications 150-400ms: acceptable for international connections >400ms: unacceptable for general network planning purposes; however might be tolerable in some exceptional cases

Impact of delay on speech quality 

Some study shows that the MOS value decreases from 3.74 (±0.52) to 3.48 (±0.48) when delay increases from 100ms to 350ms EE493Q: Digital Speech Processing

Delay Budget for VoIP Using G.729

Note: Dnw could be as high as 100ms between US and China EE493Q: Digital Speech Processing

Network QoS 

Debatable issue: is QoS really necessary? 

Unnecessary for a network with small traffic (links always <30% occupied)

QoS is mostly desirable when bandwidth becomes expensive (e.g., wireless mobile VoIP or in some remote areas of the world)  QoS can be achieved by 

 

Managing router queues Routing traffic around congested parts of the network EE493Q: Digital Speech Processing

Two Approaches to QoS 

Fine-grained approaches which provide QoS to individual applications or flows 



Coarse-grained approaches which provide QoS to large classes of data or aggregated traffic 



Integrated services such as Resource Reservation Protocol (RSVP)

Differentiated services such as asynchronous transform mode (ATM)

VoIP adopts differentiated QoS approach EE493Q: Digital Speech Processing

Comparison Between RSVP and ATM RSVP

ATM

Receiver generates reservation

Sender generates connection request

Soft state (refresh/timeout)

Hard state (explicit delete)

Separate from route establishment

Concurrent with route establishment

QoS can change dynamically

QoS is static for life of connection

Receiver heterogeneity

Uniform QoS to all receivers

EE493Q: Digital Speech Processing

Voice Over Internet Protocol (VoIP) 

Introduction to Computer Networks Direct link networks  Internetworking  End-to-end protocols 



Voice over Internet Protocol (VoIP) 

Voice quality Transport



Network QoS





Multimedia Streaming EE493Q: Digital Speech Processing

What is the Difference between Downloading and Streaming? 

Downloading   



Fully download media files to your computer’s hard drive before you could begin playing them Pro: quality guaranteed Con: you have to wait a long time

Streaming   

You can listen to or view media files AS they are downloaded in Real Time. Pro: minimal waiting time, no need to store Con: quality depends EE493Q: Digital Speech Processing

Why Streaming is Cool Exactly the most seamless way to integrate video or audio into web sites  Source material for streaming can consist of either live presentations or prerecorded material.  Anytime and anyplace distribution of on demand media to people any where in the world, which makes streaming technology so different from broadcasting and so promising for web-based education and training. 

EE493Q: Digital Speech Processing

Deeper Impact The Web now becomes capable of accomplishing what no other medium has been able to do.  It can use as its content all other media formats. Text, video, audio, images and even live radio and TV broadcasts can all be integrated and delivered through a single medium.  Wait until Web2.0 (also called “semantic web”) 

EE493Q: Digital Speech Processing