Transcript QoS models for Traffic Mapping v3x

Quality of Service (QoS) models
for Traffic Mapping
Kashi Basu
kbasu@
Main QoS Parameters
• Bandwidth
• End-to-End Delay
• Jitter
• Error and Loss
Media Types and Their QoS
Requirements
• Video
• Audio
– Speech-Quality Audio
– Music-Quality Audio
• Text and Images
The Quality-of-Service Architecture
(QoS-A)
• Made of horizontal layers and vertical planes
• The upper layer, the distributed systems
platform provides multimedia
communications and QoS specification in an
object-based environment.
• The orchestration layer provides inter-flow
synchronisation and jitter correction across
related application flows.
• The transport layer which contains a range of
QoS configurable services and mechanisms
for intra-flow QoS management.
The Quality-of-Service Architecture
(QoS-A)
• The network layer is responsible
for end-to-end QoS routing of
the flows across heterogeneous
underlying network
infrastructures. This layer along
with the lower layers form the
basis for end-to-end QoS
support.
.
The Quality-of-Service Architecture
(QoS-A)
• Control functions are carried out
in three vertical planes.
• The protocol plane consists of
protocol profiles for media
transfer and control.
• Control sub-plane
• User sub-plane
• The QoS maintenance plane is
responsible for dynamic
resource monitoring and
maintenance of their associated
protocols.
The Quality-of-Service Architecture
(QoS-A)
• The flow management plane operates on a
longer time scale and is responsible for QoS
management. Its functions includes:
– flow establishment that involves end-to-end
admission test and resource reservation.
QoS Support in ATM Networks
ATM Adaptation Layer
Convergence
sub-layer (CS)
SAR sub-layer
ATM Layer
Physical
sub-layers
ATM Service Class
Acronym
CBR
Meaning
Constant bit rate
Rate and Traffic Characteristics
Example Applications
Constant, isochronous
Uncompressed voice and
video; fixed telephony,
leased lines, ISDN
VBR-rt
Real-time variable
bit rate
Varies predictably, synchronous
Compressed voice and video;
cellular telephony
VBR-nrt
Non-real-time
variable bit rate
Varies predictably, asynchronous
ABR
Available bit rate
Bursty, but can vary on demand,
asynchronous
LAN interconnection
UBR
Unspecified bit rate
No specifiable requirements
Email
Transaction processing
Web browsing
ATM Adaptation Type
Adaptation
Type
Stream type for which designed
Constant bit-rate, isochronous
1
e.g. leased lines, ISDN, PSTN, uncompressed voice and
video
Variable bit-rate synchronous
2
e.g. MPEG video, mobile telephony services (GPRS,
HSPDA, UMTS), voice over xDSL
3/4
Data streams carried over public carrier networks
Comment
Traditional (G.711)
telephony
Compressed voice
and video
Defunct
All data streams
5
e.g. LAN backbones and wide area LAN interconnection
services
IP
An example ATM call
Set up call to F
(variable bit rate,
AAL 5,
PCRb, SCRb, MBSb)
ATM source
end system (ES)
1. Source end system signals dest ES address and call
category/AAL/QoS to ATM switch
2. Ingress switch forwards call setup info towards dest ES
3. Each switch matches QoS to appropriate outgoing link and
forwards signalling message to next switch
4. Last ATM switch in path signals QoS to dest ES
5. Dest ES accepts (or rejects) call
6. Signalling passed back to source ES gives
outgoing VPI/VCI to use (or reason in case
of call set-up failure)
Call
OK,
VPI/VC
I=
34/127
ATM edge switch
relays call
requirements to
network
ATM Cell Structure
Cell header (5 bytes)
bits 4
8
VPI
GFC (trunk
ID)
16
VCI
(call identifier)
Cell payload (48 bytes)
3 1
PTI
8
HEC
QoS in IP Network
• Integrated Services
• Differentiated Services
• Multiprotocol Label Switching (MPLS)
Integrated Services
• Service Classes:
Sender
PATH
• Guaranteed Service (GS)
• Controlled Load Sharing
Service (CLS)
R
RESV
PATH
R
R
RESV
• Best effort service (BE)
Receiver
GS Class
Service Class
Guaranteed Service
Traffic Descriptors(Tspec)
AdSpec parameters
RSpec parameters
peak rate (p),
token rate (r),
bucket depth (b)
Ctot, Dtot,
min path latency (MPL), available path
bandwidth
service rate (R), slack term (S)
Differentiated Services
• Scalable solution
• Based on marking:
– Diffserv codepoint (DSCP)
– IPv4 type of service (ToS)
– IPv6 traffic class field
• Characterised by per-hop behaviour (PHB)
– Expedited forwarding (EF)
– Assured forwarding (AF)
MPLS
• Fixed-length label based forwarding model that combines
functionalities from layer-2 switching and layer-3
forwarding
• At the ingress node packets are aggregated into forwarding
equivalence classes (FEC) based on their destination and
QoS requirements.
• Packets in each FEC traffic trunk are assigned with a 32-bit
shim header (on top of their layer-3 header) which consists
of a 20 bit fixed label that serves as an index to the routing
table.
• Route lookup based on this fixed length label is faster
compared to the traditional longest prefix match of IP
address.
Traffic Mapping- Intserv to ATM
Traffic Mapping- Diffserv to ATM
QoS Parameters for Traffic
Characterisation
• Dual bucket parameters are not ideal for
streaming traffic
Token bucket
Token
Data queue
Leaky bucket algorithm
Token bucket algorithm
QoS Parameters for streaming media
Time
Length
Average packet inter-arrival
time{ a i },
Squared Coefficient of Variation
(SCV) of the inter-arrival time
{ C a2 },
Average packet length { l }.
SCV of the packet length { C l2 }
Traffic mapping: n-to-n model
IPv6
IP
Flows
Flows
AAL
-5
Encapsulation
Cell
Fragmentation
Fragmentation
f1
vc1
vc1
q1
AAL
AAL
f2
AAL
q2
AALAAL
Virtual
Circuit
AALAAL
AAL
vc2
vc2
ATM
Link
ATM Link
f3
q3
AAL
AAL
f4
AAL
AAL
AAL
vc3
vc3
q4
AAL
AAL
AAL
AAL
AAL
Mapping individual IP flows onto separate ATM VCs (n-to-n model)
vc4
ATM Cell slot
Fixed synchronous cell slots (E.g.
2.72μs over SONET OC-3)
ASYNCH
STAT
MUX
Asynchronous transmission rates
between the circuits (E.g.
telephony, compressed video and
LAN traffic)
Priority based n-to-n scheduler
Cell-slot conflict between VCs over the same ATM link
Traffic mapping: n-to-1 model
Class-based queuing
• Class-based queuing (CBQ)
– Used for this purpose to provide bandwidth isolation between the
classes.
– Flows are aggregated into classes with one queue for each class
– Packets may be classified based on media type (such as video and
audio class), destination, protocol or more specific traffic
characteristics such as inter-arrival time and packet length or a
combination of these parameters.
– Either work-conserving or non work-conserving scheduler could be
used depending
– Resources are allocated to a class and not to individual flows,
therefore adequate admission control and resource management
techniques are also essential for providing QoS guarantee to the class.
– Classes may be further classified into different priority levels and
served accordingly using priority queuing.
Weighted Fair Queuing
• Weighted Fair Queuing (WFQ)
– Uses a work conserving scheduler; therefore each
active flow also gets a weighted share of any
unused bandwidth.
– each incoming packet is tagged with a virtual
finishing time which is the time the packet
– The scheduler then services packet from the
queues based on the ascending order of their
finishing time.
Experiment Set-up and Results
Parameter
Video class
Audio class
No. of sources
3
3
Interarrival time for each
source
40 msec
6 msec
SCV of the interarrival
time
1.0
1.0
Cumulative arrival rate
6 mbps
2 mbps
99 %tile Delay (msec)
100
80
60
40
20
0
0
50
100
150
Total Background Load (Mbps)
200
Response
time fortime
thefor
video
classclass
under
varying
background
the
data
traffic
Fig. 6.3: Response
the video
under
varying
background load
load ofofthe
data
traffic
99 %tile Delay (msec)
w1
50
w2
40
w3
30
20
10
0
0
20
40
60
80
Traffic load on each data queue (Mbps)
Fig. Throughput
6.5: Throughput
the data
queues
under
varying
traffic
of theofdata
queues
under
varying
traffic
load load
Analytical results
Analytical results (exact solution of the
G/G/1 queue with G-type single vacation)
Simulation
results
Experiment
results
99 %tile delay (msec)
C a2 SCV
1 I=1
aAvg
i  I40m
= 40sec,
msec
100
80
60
40
20
0
0.25
0.375
0.5
Utilisation factor
0.625
Comparison
of the response
time
for the video
class
of consisting
homogeneous
sources
Fig. 6.6: Comparison
of the
response
time for
theconsisting
video class
of homogeneous
sources
2
a i  6m sec, C a  1
99 %tile delay (msec)
AvgI = 6 msec SCVI = 1
40
30
20
10
0
0.25
0.375
0.5
0.625
Utilisation factor
0.75
Comparison
the responseoftime
for the audio
consisting
of homogeneous
Fig. 6.7:ofComparison
the response
timeclass
for the
audio class
consisting ofsources
homogeneous
sources
Note:
ai
= Average interarrival time of packets
Ca2 = Squared coefficient of variation of the interarrival time of packets
Architecture of the IP/ATM
Internetworking Device
Architecture of the IP/ATM internetworking device
An QOS aware end-to-end network
QoS-based IP network
{Intserv, Diffserv, MPLS, etc.}
Edge
Device
ATM core
Edge
Device
QoS-based IP network
{Intserv, Diffserv, MPLS, etc.}
A QoS aware heterogeneous end-to-end path comprising IP and ATM technologies
with the IP/ATM internetworking devices acting as an interface between them
References
• R. Braden, D. Clark, and S. Shenker. Integrated Services in the Internet
Architecture: An Overiew. RFC 1633.,June 1994
• Braden, R., Zhang, L., Berson, S., Herzog, S. and Jamin, S., "Resource
Reservation Protocol (RSVP) Version 1 Functional Specification," RFC2205,
IETF, Sep. 1997.
• Shenker, S., Partridge, C. and Gu'erin, R., "Specification of Guaranteed
Quality of Service," RFC 2212, IETF, 1997.
• Garrett, M. and Borden, M., "Interoperation of Controlled-Load Service
and Guaranteed Service with ATM," RFC 2381, IETF, Aug 1998.
• Basu, K., Ball F. and Kouvatsos, D.D. "A Simulation Study of IPV6 to ATM
Flow Mapping Techniques". SCS Transaction Journal on Network Modeling
and Performance Issues, Vol. 78, Issue 7, pg. 423-430, 2002.
• Wang, Z., Internet QoS: Architectures and Mechanisms for Quality of
Service, pp. 64-69,168 -176, San Francisco, USA: Morgan Kaufmann, ISBN:
1558606084, 2000.
Presentation topic
•
•
•
•
•
•
•
•
•
Network architecture for Video Broadcast in an IP network
Duration: 10 mins
Key Terms: Video server, QoS, streaming, MPLS, Diffserv
Some references to start with:
Kuo, G.-S., Lai, C.T.: A new architecture for transmission of MPEG-4 video
on MPLS networks. In: ICC 2001 - IEEE International Conference on
Communications, June, vol. 5, pp. 1556–1560 (2001)
J. Asghar , F. Le Faucheur and I. Hood "Preserving video quality in IPTV
networks", IEEE Trans. Broadcast., vol. 55, no. 2, pp.386 -395 2009
C.D. Cranor, M. Green, C. Kalmanek, D. Shur, S. Sibal, J.E.
Van der Merwe, and C.J. Sreenan, “Enhanced Streaming Services in a
Content Distribution Network,” IEEE Internet Computing, 5(4): 66-75,
2001.
Resource: Google scholar, Safari, Brookes library resources ( both online
electronic materials and offline printed media)