CMSC691C Multimedia Networking A Course Overview

Download Report

Transcript CMSC691C Multimedia Networking A Course Overview

CMSC691C Multimedia
Networking
A Course Overview
Padma Mundur
CSEE, UMBC
[email protected]
List of Topics
 Multimedia Networking: Source Representations,
Networks, and Applications
 Multimedia Compression Fundamentals & Coding
Standards
 Scalable Video Coding for Heterogeneous Networks
 Fundamentals of IP Routing
 IETF QoS Efforts
 Existing Solutions for Scalable Multimedia QoS
The
Telephone
(Voice)
Network
• Circuit switched network:
–
–
–
–
Analog (since1890): manually switching
Digital: voice  bit stream (64 Kbps)
Better channel utilization by time-division multiplexing:
Reservation fixed for the whole transmission
A
C
B
•
The
Internet
(Data)
Network
Packet-switched network:
–
–
–
–
–
packets share resources (buffers, links)
reservation not fixed, but on-demand
multiple links (connectivity, reliability)
buffers (store, process, forward)
control information in packets (s,d,seq#)
Internet Users Growth
Source: www.isc.org
1B
mobile users by 2005 and 1B Internet users by
2005
90%
of all new mobile phones will have internet
access by 2003 (Morgan Stanley Dean Witter, May 2000)
Multimedia over IP Networks
VoIP
Video
Conference
Internet
Wireless
Browsing
Video Clip
Attachment
E-mail
Music
Streaming
Movies
Streaming
Information
Search
Finance,
Brokerage
Digital Photos
Multimedia Networking Applications
• Media Broadcast: simultaneous pushing of content
to multiple recipients
– Network IP Multicast – Multicast enabled routers and
switches
• Hosted Streaming: content users initiate requests
and content networks/providers push content
through network
• Interactive Conferencing: no centralized source of
contents
Multimedia Broadcast over IP
Internet
content provider
clients
IP (internet protocol) makes it
possible to link all (global) nodes
together independent of
applications and terminal devices
Hosted Multimedia Streaming
•Multicast capable
•More Robust
•Access to Storage
•Relieves Web Server
Audio
Video
Animation
Proprietary
Format
Media
Encoding
Media
Server
•Standalone player
•Java based player
•Browser plug-in player
•Appliance
Send Stream
To Clients
Clients
Send Request to
Media Server
To hear or view a
media file without
downloading it
Web
Server
Send Request
To Servers
•Decode
•Buffer
•Sync.
Interactive Conferencing and
Meeting Server
Internet
Bandwidth concern for
multipoint interaction
How A Server Distributes the Data
Meeting Token Holder
Meeting Client
Token
Dynamic Token Passing
New Token Holder
Meeting Client
Token
Old Token Holder
Multimedia Signals and Bitrates
Source
Bandwidth
(Hz)
Sampling
Rate
Bits per
Sample
Bit Rate
Telephone Voice
Wideband speech
Wideband audio
(2 channels)
B/W documents
Color Image
CCIR-601 (NTSC)
CCIR-601 (PAL)
SIF (standard)
CIF (common)
QCIF (quarter)
200—3400
50—7000
20—20,000
8000 samples/sec
16,000
44.1 Ks/sec
12
14
16/channel
300 dpi (dots/inch)
512x512
720x480x30
720x576x25
360x240x30
352x288x30
176x144x7.5
1
24
24
24
12
12
12
96 Kbps
224 Kbps
1.412 Mbps
(2 channels)
90 Kb/inch2
6.3 Mb/image
248.8 Mbps
248.8 Mbps
31 Mbps
37 Mbps
2.3 Mbps
Audio & Video Quality Requirements
IP Networks
• IP uses packet switching
– Suitable for unexpected burst of data without establishing an
explicit connection.
– Bandwidth is shared statistically so data can be sent at any time.
• IP is not reliable nor delay-bounded.
–
–
–

Best effort
Queuing delay, especially when congested.
Network failures can cause temporary packet loss.
Time critical applications cannot operate well due to large e-mail
attachments and Web surfing
 Delay and jitter degrade voice and video performance
Multimedia Signals
•
•
•
•
•
•
•
Text
Speech
Audio
Image (B/W and color)
Video
Graphics & Animation
Documents (various formats)
Image & Video Coding Standards
•
•
Combination of lossy (transform coding) and lossless (run-length,
Huffman, Arithmetic coding, LZW, etc) coding techniques along
space and time.
JPEG - Joint Photographic Experts Group
•Still image compression, intraframe picture technology
•Motion JPEG (MJPEG) is sequence of images coded with JPEG
•
MPEG - Moving Picture Experts Group
•Defined by ISO/IEC, several standards MPEG1, MPEG2, and now MPEG4
•
H.263/H.263+/H.26L - Videophone/Conferencing
•Low to medium bit rate, quality, and computational cost defined by ITU
•Used in H.320 and H.323 video conferencing standards
A Complete JPEG Encoding
DCT
Quantize
Zig-zag
011010001011101...
Run-length
Code
Huffman
Code
From Image to Video Coding
• Intra-frame compression (similar to JPEG)
•Remove redundancy within frame (spatial)
• Inter-frame compression (motion compensation)
•Remove redundancy between frames (temporal)
• Rate Control (constant bit-rate or constant SNR)
Video Coding Standards
 MPEG1 – VHS quality, VCD (1992)
•CIF images, 4:2:0 sampling, 1.5 Mbs, Frame encoding
 MPEG2 - broadcast quality, HDTV and DVD (1994)
•CCIR 601 images, 4:2:2 sampling, 4-15 Mbs
•Interlaced and progressive scanning, Frame and field
 H.261 for videotelephony (p=1,2) & videoconferencing (p>= 6) (1992)
•Improve JPEG through temporal redundancy
 H.263 – low bitrate video coding (1995)
•Half pixel motion compensation, 4 (optional) modes
•Optimized VLC tables & better motion vector prediction
 H.26L(H.264) – flexible, high quality video applications (2002)
•1/4 pixel accuracy for MC, 7 different block sizes for ME/MC
•Residual coding uses 4x4 blocks & an integer transform
MPEG-4: An Emerging Standard
• For multimedia applications
– Interactive natural & synthetic contents
– Various access conditions: low bit-rate, error prone, heterogeneous
(scalable)
– Management and protection of media contents
• Standard
– 1st generation (1998-2000): 1st+2nd versions, frame based content
creation & communication, 64-384 Kbps, mobile videophone (3G
and IP) and digital camcorder
– next generation (2001-): upto 2Mbps, frame/object based, scalable
streaming, interactive set-top box
Heterogeneous IP Networks
 Adaptive Rate Control
 Scalable Coding
 Real-time bandwidth
estimation
 Receiver feedback
 Adaptive Multicast control
Video Scalable Coding
• Why a scalable video codec?
– Compression efficiency
– Robustness with respect to packet loss
– Adaptation to the changing bandwidth
• Techniques of scalable video coding
–
–
–
–
–
–
Temporal
Spatial
Signal-to-Noise Ratio (SNR)
Data Partition
Wavelet
Fine Granularity Scalability (FGS)
IP Stack: A Layered Architecture
users
network
Application
Web (HTTP), E-mail (SMTP),
File transfer (FTP), Name resolution (DNS),
Remote terminal (TELNET), …
Transport
Reliable multi-connection bit-stream (TCP),
unreliable multi-connection (UDP).
Network
Unreliable end-to-end delivery of
packets up to 64 KB.
Physical
Point-to-point links (PPP, SONET, …),
LANs (Ethernet, FDDI, wireless, …)
IP Packet Routing: Delay and Loss
Router
Hello
Router
Router
Hello
Router
Router
Router
Router
transmission
A
Router
IP
Network
Router
propagation
B
nodal
processing
queue management
Queuing and Scheduling (1)
 FIFO - First In First Out queuing, definitely not
compatible with QoS since high priority packets can get
stuck behind low priority packets
Queuing and Scheduling (2)
 Priority Scheduling - services higher priority queue
whenever there are packets present, can lead to starvation
of lower priority queues
Queuing and Scheduling (3)
 Custom Queuing (or Weighted Round Robin Scheduling) - services all
queues (with different service time) within a traffic class, round robin
assuring that all queues get appropriate treatment
Queuing and Scheduling (4)
 Weighted Fair Queuing (WFQ) - queue is serviced based on a weight
proportional to the bandwidth dynamically allocated to it
Congestion Control & Queue Discard
 Tail Drop
– Drops arriving packets when buffers in queue are full, can
lead to network meltdown due to TCP global
synchronization
 RED = Random Early Detection
– Queuing algorithm for congestion avoidance that randomly
discards packets from queues in an attempt to prevent TCP
retransmits simultaneously on all flows
Congestion Control & Queue Discard
 WRED = Weighted Random Early Discard
– A variant of RED that attempts to weight queues for random
early discard
 Tri-Color Marking (deterministic)
Queue
Limit
Yellow
Drop
Threshold
Red
Drop
Threshold
IP QoS and Multimedia
• Quality of Service (QoS) methods aim at trading quality
vs. resources to meet the constraints dictated by the user,
the functionality and the platform.
• QoS originally developed in network communication, and
recently extended to the domain of multimedia
communication.
• QoS relevant in multimedia scalable systems, where the
resources and the functionality can be controlled by a set
of parameters.
IP Quality of Service (QoS)
 Techniques to intelligently match the
performance needs of applications to available
network resources
 QoS Metrics
•availability
•delay (latency)
•delay variation (jitter)
•throughput (average and peak rates)
•packet loss
IETF IP QoS Efforts
• Policy based IP QoS Solutions
– Integrated Services (RSVP protocol): flow based
– Differentiated Services (DiffServ byte settings): packet based
– Multi-Protocol Label Switching (MPLS): flow+packet based
• IP Multicast and Anycast
• IPv6 QoS Support
Connection Oriented QoS
 Int-Serv (Integrated Services): IETF RFC 1633
 Defined by RSVP requires resource reservation at each
hop end-to-end for each IP packet flow, and end-to-end
signaling along nodes in the path
 Reserve resources at the routers so as to provide QoS for
specific user packet stream
 This architecture does not scale well (large amount of
states)
 Many Internet flows are short lived, not worth setting up
VC
Integrated Services / RSVP
 Sender sends a “PATH” message to the receiver
specifying characteristics of traffic
• every intermediate router along the path forwards the “PATH”
message to the next hop determined by the routing protocol
 Receiver responds with “RESV” message after receiving
“PATH”. “RESV” requests resources for flow
Connectionless QoS: IP Diff Serv
 Mark IP packet to specify treatment: IETF RFC 2474, e.g., first
class, business class, coach, standby
 Per Hop Behaviors (PHBs) based on network-wide traffic classes
 Flows are classified at the edge router based on rules, and are
aggregated into traffic classes, allowing scalability
 Diff Serv uses the IP header TOS byte (first 6 bits), which is
renamed the DS field
 Diff Serv defines code points (DSCP) for the DS field, DE (default)
= 000000 = best effort, and EF (Expedited Forwarding) = 101110 =
low latency, etc.
DiffServ Operation
Each ISP configures its own routers to match the service that it
offers, and each ISP has its own DiffServ Domain.
Customer Site
ASP
email
Voice Video
PHB
PHB
PHB
DS Domain
DS Domain
SLA
(service level
agreement)
SLA
SLA
Edge Router
Interior Nodes
MPLS Fundamentals
 MPLS is a forwarding scheme that tags packets with labels
(independent of layers 2,3) that specify routing and priority
(IETF RFC 3031)
 Enables scalability by alleviating IP over ATM problems
• Defines a homogeneous network based upon label-switching
• Requires all devices (i.e., ATM switches) to be capable of routing
 Enables differentiated services via QoS-aware label switched
paths (LSPs)
 Designed to run over a wide range of media
• ATM, frame relay, and Ethernet
Unicast/Multicast
Unicast
Host
Router
Multicast
Host
Router
Multimedia IP Multicast
 Why multicast?
•
•
•
•
When sending same data to multiple receivers
Better bandwidth utilization
Lesser host/router processing
Receivers’ addresses unknown
 Applications
• Video/audio conferencing
• Resource discovery/service advertisement
• Media streaming and distribution
IP Multicast Service Model
IETF RFC 1112, each multicast group is identified by a
class D IP address
•Range from 224.0.0.0 through 239.255.255.255
 Well known addresses designated by Internet Assigned
Number Authority (IANA)
•Reserved use: 224.0.0.0 through 224.0.0.255
Members join and leave the group and indicate this to
the routers
Multicast routers listen to all multicast addresses and use
multicast routing protocols to manage groups
What IPv6 can Offer?
• Global Addressing (128 bits):
– 1 million networks per human
– 20 hosts per m2 of Earth
n bits
who you are
128-n bits
where you are connected to
• Plug and play:
• Efficient mobility (instant-on ad-hoc networking)
IPv6: Key Features and Advantages
•
•
•
•
•
•
•
•
Increased Address Space (128 bits)
Efficient and extensible IP datagram
Improved host and router discovery
Plug and Play
Enhancements for Quality of Service (QoS)
Improved Mobile IP support
Coexistence with IPv4
Built in security (authentication and encryption) in IP layer
IPv6 Support of QoS
• IPv6 Flow Labels provide support for Data Flows
– Packet Prioritizing-- sure that high priority traffic is not
interrupted by less critical data
• IPv6 supports Multicast & Anycast
– Multicast delivers data simultaneously to all hosts that sign
up to receive it
– Anycast allows one host initiate the efficient updating of
routing tables for a group of hosts.
Existing Scalable Multicast Solutions
•
•
•
•
•
•
Content Distribution Networks
Receiver-Driven Layer Multicast (RDLM)
Source Adaptive Multi-Layer Multicast (SAMM)
Filtering Method
Destination Set Grouping (DSG)
Multiple Description Coding (MDS) of Multimedia
Distributing Content to the Edges
 Adding backbone bandwidth is not the best solution, the last
mile (edge) connection is even more critical.
 How to direct traffic to the site (routing) and resolve the
appropriate server (load balancing) that will perform best for
a particular query (front-end content delivery).
 How to keep content updated efficiently (back-end content
delivery)
Getting Contents to the Edge
 Caching (on-demand “pull”)
– Contents may be “pulled” from another proxy cache in the hierarchy or
the origin of the contents.
– Problems: stale content delivery, hit statistics loss, dynamic contents
 Replication (changes made on the origin server)
– Updates are “pushed” to replica using a “back-end” content delivery
system.
– The origin server is in total control (database keep track of content
changes), with scalable architecture (multicast or packet-relay).
Getting Contents to the Edge
 Resolution Problem
– The best site: geographic vs. network proximity (quickest
service)
– Domain Name Service (DNS) criteria: # of authoritative servers
(domain) hops, # of router hops, health/load of site, round trip
latency, packet loss rate, etc.
 Hybrids of Caching/Replication
 Reverse Proxy Cache: all queries directed to proxy caches by
load balancer for front-end delivery.
 Pre-Filling of Proxy Caches: parts of the Web site are pre-filled
to the cache – i.e., replica in proxy cache form
Content Distribution Network (CDN)
 Service providers using proprietary caching/replication
technologies to build overlay networks (internet or satellite) to
deliver contents – application level multicast
Receiver-driven Layer Multicast (RLM)
• RLM Protocol Concepts (McCanne 1996):
Source: No active role in the protocol.
Receivers: On congestion, drop a layer.
On spare capacity, add a layer.
 When to drop a layer:
Whenever congestion happens. Congestion is expressed
explicitly in the data stream through lost packets.
• When to add a layer:
Join-experiment: To carry out active experiments by
spontaneously adding layers at “well chosen” time.
RLM Characteristics
 A fixed number of multicast groups.
 Lack of granularity adaptation
 Severe quality degradation when pack loss
on base layer.
 Slow adaptation to changes of varying
network bandwidth (Liu, Hwang 2002)
 Synchronization is crucial
 Well-developed protocol is crucial
Source Adaptive Multi-layer Multicast
 SAMM Protocol Concepts (Suda 1998):
 Video is encoded into several layers and
each layer has an unique discarding
priority.
 When a network link experiences
congestion, packets from the lowest priority
layer are discarded.
 The video source obtains backward
feedbacks from receivers to adjust the
number of video layers and also the
encoding rate for each layer.