Dakshi Agrawal, Mandis S. Beigi, Chatschik Bisdikian, Kang-Won Lee IBM T. J. Watson Research Center, Hawthorne, NY, USA. 10th IFIP/IEEE International Symposium on Integrated Network Management, 2007 Chen Bin Kuo (20077202) Young J. Won (20063292)
Dakshi Agrawal, Mandis S. Beigi, Chatschik Bisdikian, Kang-Won Lee
IBM T. J. Watson Research Center, Hawthorne, NY, USA.
10th IFIP/IEEE International Symposium on Integrated Network Management, 2007
Chen Bin Kuo (20077202)
Young J. Won (20063292)
DPNM Lab.
Introduction The IPTV Distribution Model Problem Formulation Solution Design Design of the Planning Tool Concluding Remarks
04/19/23 2
DPNM Lab. 04/19/23 3
Integration of services over converged networks• Providing the opportunity for legacy players• Emergence of triple-play service offerings
Telephony services companies (TelCos) • Providing services based on the DSLs• Upgrading their network to be able to provide triple-play
services
DPNM Lab.
This paper focuses on the emerging deployment of TV and video-on-demand services by TelCos
IPTV can utilize network resources efficiently and facilitate new service features such as: Multiple views on the same event Integrated video-on-demand (VoD) - listings for live and
VoD programming Program navigation and search VCR-like commands
04/19/23 4
DPNM Lab.
This paper presents: A model for IPTV service distribution and key parameters
be used to analyze the performance
A general framework for planning an IPTV service deployment and management
A solution design for a deployment management tool based on the framework proposed in this paper
Overall issue of IPTV service provisioning
04/19/23 5
a) Client Domainb) Network Provider Domainc) Service Provider Domaind) Quality of Experience (QoE)
04/19/23 6
DPNM Lab. 04/19/23 7
• Residential gateways• Set-top-box (STB)
• Residential gateways• Set-top-box (STB)
• Distributing various services• Based on the FTTN• Last-mile and second-mile network• (a) DSLAM, (b) routers
• Distributing various services• Based on the FTTN• Last-mile and second-mile network• (a) DSLAM, (b) routers
DPNM Lab.
a) Client Domainb) Network Provider Domainc) Service Provider Domain
1. Super Headend (SHE) Manages and processes all incoming broadcast video feeds and to
the downstream
2. Video Headend Office (VHO) Typically serves a region or a metropolitan area Inserts local TV channels and advertisements into the IPTV streams
3. Video Switching Office (VSO) Multiplexing video service with other services (VoIP, broadband
Internet access)
04/19/23 8
DPNM Lab.
d) Quality of Experience (QoE) Representing a collection of metrics to reflect the
subscribers’ satisfaction QoE metrics
Video quality Channel change time (channel zapping time) Blocking probability for VoD requests
Additional metrics can be supported by the framework
04/19/23 9
a) A Model of the IPTV Infrastructureb) Optimization Problem Formulation
04/19/23 10
DPNM Lab.
Modeling an IPTV network using a graph consisting of nodes and edges Link has propagation delay and packet loss rate parameters
Modeling sites and servers as queueing systems One may substitute more sophisticate models when they
become available Capture a macroscopic behavior of viewers
For example, by the Nielsen ratings [4] Deriving the channel viewing preference for each
community
04/19/23 11
DPNM Lab.
Given an IPTV infrastructure currently serving a set of existing communities The problem is to fine the way to maximize the number of
new subscribers without adding new resources
Observing that the problem can be formulated as a combinatorial optimization problem such as knapsack problem or a bin packing problem NP-hard Efficient algorithms exist
04/19/23 12
a) Community Modelb) Channel Zapping Delayc) Data Server Modeld) Video Quality Models
04/19/23 13
DPNM Lab.
Assuming viewing profile of viewers are available to service provider
Define a viewer community to be a collection of viewers Residing in a geographical proximity and treated as uniform
For each community: Channel viewing preference: The VoD content duration statistics: The viewer request rate vector:
04/19/23 14
DPNM Lab. 04/19/23 15
DPNM Lab.
A viewer in community j switches to channel i
The zapping delay for community j
The overall zapping delay
04/19/23 16
DPNM Lab.
Adopting the M/M/c/(c+K) queueing model
04/19/23 17
DPNM Lab.
Blocking probability can be solved in queueing system [7] [8]
One may choose to use a more elaborate model – VoD server infrastructure
In [9] [10] for VoD system design also use Markovian queueing models or extensions of these models
04/19/23 18
DPNM Lab.
Adopting the moving pictures quality metirc (MPQM) [11] [12] Representing a numeric score denoting a viewing
experience from bad (1) to excellent (5) A basic human vision model which takes into account the
viewers perception of the video
MPQM model:
04/19/23 19
a) Software Architectureb) Algorithmic Structurec) Case Study – Adding New Markets
04/19/23 20
DPNM Lab.
Developed as a proof of concept of the proposed framework
Functional diagram
04/19/23 21
DPNM Lab.
Using a knapsack algorithm to solve the problem Multiple knapsack problem (MKP):
NP-hard problem [5] already presented an efficient algorithm for MKP
Relationship Each community is an item, each IPTV node is a knapsack
with certain capacity Connecting a new community has some value Cannot directly apply
04/19/23 22
DPNM Lab.
Fitting model to MKP: Server capacity:
A server typically has a fixed bound for the rate of request Treating like the weight of the item in MKP
Channel zapping delay: Using the iterative calculation in (5), we can efficiently test this
condition Service blocking probability:
Easily tested for each sites because it depends on the site parameters Under Poisson assumption, we can simply update it
Network parameters: For this parameter, we just need to consider the new community
04/19/23 23
DPNM Lab.
A service provider has two VHOs near mid size cities that are currently over-provisioned
The service provider tries to serve ten new emerging communities out of these two VHOs
04/19/23 24
DPNM Lab. 04/19/23 25
DPNM Lab. 04/19/23 26
This paper focused on a framework to aid planning and managing the deployment of IPTV services
The models are used to map a set of external parameters Service support resources, network nodes and topology, and
communities of viewers Depending on the complexity of the deployment
options either exhaustive scans or intelligent scans can be used
Different deployment objectives can be studied through the framework
04/19/23 27