This document outlines the thesis project of Satya Prakash Rout, a student pursuing an M.Tech in Applied Optics. The project involves emulating different Dynamic Bandwidth Allocation algorithms to monitor the performance of a 10G-EPON network implementing Triple Play. Specific aspects that will be studied include implementing Triple Play with a new scheduling algorithm called TD-Sense, generating different traffic models, comparing the performance of DBA-Gated, DBA-Linear and DBA-Max algorithms, and potential future work involving long-reach PONs and green networking techniques. The document concludes by acknowledging the contributions of the student's guide, advisors and classmates to the successful completion of the project.

1. SATYA PRAKASH ROUT
REG NO-11706
M.TECH APPLIED OPTICS

3.  Problem Definition
 Introduction to Access Network
 Active optical Network vs. Passive Optical network
 Passive Optical Network Architectures
 Types of Passive Optical Network
 Types of Bandwidth Allocation for PON
 Access Network Planned for Emulation
 Dynamic Bandwidth Allocation Algorithms
 Introduction to Triple Play
 TD-Sense Algorithm: A newest DBA approach to Triple Play
 Results and Comparison between DBA Algorithms
 Future Work
 Conclusion
 Acknowledgment

4.  Studying and emulating different DBA
algorithms to monitor the performance of
10G-EPON network.
 Implementation of Triple Play with a new
scheduling algorithm and different traffic
generation models

5.  The internet users in the world is
increasing every year whereas the
users perchantage of Asia is more than
the users of different geographical
region.
 The growth of bandwidth demand for
accessing differentiated services is
exploding day by day.
 Need an efficient Access Network that
can operate at high functional speed
 Access Network is sensitive to cost and
geographical region of coverage
 Trade off between CAPEX, OPEX and
geographical region of coverage

6.  AON needs a router or switch to
route the signals whereas PON
needs an passive optical splitter
 AON needs Electrical-Optical-
Electrical conversion whereas PON
needs no E-O-E conversion
 There is no use of multiplexer and
de-multiplexer in case of PON
 Bandwidth promised per user is
higher in PON compared to AON
 PON is electrically efficient than
AON

7.  PON has mainly three elements in the architecture.
 Optical Line Terminal (OLT)
 Optical Network Unit (ONU)
 Passive Optical Splitter
Optical Line Terminal
 It is the key element of the access network, which is placed
at the central offices
 It acts an interface between the core and access network
Optical Network Unit
 It is deployed near to the customer premises and provides
services to the end users fetched from the OLT in the
downstream direction.
 In the upstream it sends the requests of the user in
upstream direction.

8. Passive Optical Splitter
 It’s a device with single input in upstream and multiple
outputs in downstream.
 The input signal power is equally divided among all the
ONUs.
 PON uses 1310nm -1490 nm for upstream direction and
1550 nm for downstream direction
 The data from both upstream and downstream propagates
in the same fiber.

9. Features APON /BPON GPON EPON 10G-EPON
Defined by FSAN and ITU-
T,G.983
ITU-T G.984
series
IEEE 802.3 ah IEEE 802.3 av
Upstream rate 155 Mbps 622 Mbps 1 Gbps/1.25
Gbps
1 Gbps/10
Gbps
Downstream
rate
155 Mbps/622
Mbps
1.25 Gbps Gbps/1.25
Gbps
10 Gbps/
10Gbps
Payload
encapsulation
ATM Payload GPON
encapsulation
Ethernet
Payload
Ethernet
Payload
Control unit T-CONT T-CONT Logical Link
Identifier
Logical Linked
Identifier
Link speed Average speed Faster than
EPON
Slower than
GPON
Very Fast
Splitting ratio 1:32 1:64 1:32 1: 64 / 1:128

10.  Static Bandwidth Allocation
 It allocates same bandwidth to all the users irrespective of
their priorities of services.
 Not suitable for high Quality of Services like Video On
Demand, Video Conferencing, High Definition TV etc..
 Made the network performance poor
 Dynamic Bandwidth Allocation
 It allocates bandwidth dynamically to all the users as per
their priorities of services.
 Highly suitable for differentiated services including voice ,
video and data.
 Increased the efficiency of the network.

11. LINUX OS PC1
OLT
LINUX OS PC 2
ONU
1
ONU
2
ONU
3
Random
Traffic
Generator
Random
Traffic
Generator
Random
Traffic
Generator
PROCESSESTHREADS

12.  10G SFP+ ER NIC cards are installed in two systems
which is well capable for pumping data at 10Gbps
rate.
 The network emulation is a semi real access network.
 One machine considered as OLT and other machine
considered as ONU.
 Multiple processes created in the system that
behaves as multiple ONUs.
 Each process has multiple threads to behave as end
users.
 Each thread created calls the traffic generator
module to send their packets to the next machine

13.  Traffic generation module used two type of
probability distribution function for generating
packets from 64 bytes -9000 bytes.
 Exponential traffic distribution is used to
generate packets from 64-1500 bytes.
 Pareto Distribution is used to produce self similar
traffic called ON-OFF codes.
 Packets generated are basically from 5000 bytes-
9000 bytes in case of Pareto traffic Distribution.
 Voice, Video and Data packets are generated
using this traffic distribution functions

14.  In high speed network, the channel
idle time is waste of resources.
 To avoid packet collision it is
essential to consider idle time.
 Trade-off should be maintained in
channel time to improve throughput
and efficiency.
 In our access network, two cases
arise related to channel idle time
Case 1:
If the downstream channel is idle,
when the report message is arrived
Case 2:
If the downstream channel is busy,
when report message is arrived

16.  Throughput and Delay are
the major output
parameters of measuring the
network performance.
 The OLT grant window for
transmission to ONUs by
allocating bandwidth
dynamically
 Three DBA algorithms have
been proposed to improve
the performance of our
access network.
Throughput
Delay
NETWORK
PERFORMANCE

17. DBA-GATED
•Trails on a simple procedure of granting window with out complex
calculation
•The offer load is equal to the rate of generation of packets in the
traffic generator module
DBA-
LINEAR
•Grants a window size proportional to the request size of the
window.
•It uses a linear factor to grant the users a window size little more
than the requested one.
DBA-MAX
•OLT scheduled a GATE threshold value fixed for every user. If users
asks a Grant window more than the threshold value, then threshold
value of Gate will be assigned to user
•OLT is not being blocked by the by the other users, it creates a lot
overhead compared to other two algorithms

18. Triple
Play
Data
VideoVoice

19.  Triple play is the service that
provides voice, video and data
over IP network.
 All the internet service
providers are competing to
provide triple play with better
quality of service.
 In the project the followings
are considered……………
 voice packets (62-80 bytes),
 video packets(6000-8000 bytes)
 data packet(800-2500 bytes)

20.  This is a online scheduling algorithm which
proposed especially for Triple Play.
 The performance of the network is
monitored by varying both throughput and
delay in the algorithm
 Both delay sensitive and non delay sensitive
traffic has been considered.
 Delay sensitive traffic are basically, the voice
packets and video packets.
 Non-delay sensitive traffic are normal
packets of random size.

21.  Type 0 traffic is denoted by delay sensitive
traffic where as Type 1 traffic is denoted by non-
delay sensitive traffic.
 The channel free time is calculated which plays
a major role of eliminating packet collision.
 This algorithm deals with both scheduling of
Type 0 and Type 1 traffic.
 Offer load(β) = b1 β1 +b2 β2
 Β1 is offer load of Type 0 and β2 is the offer load
of Type 1
 It focuses on proper utilization of channel free
time.

22.  Collect the report lengths of all the users and
store inside the buffer,
 store (ONU_SIZE(Length);// report length
 calculate (channel_free_time);
 Check the following four conditions…..
 T(0) > T2 - T1, then schedule T(0) but don’t
schedule T(1)
 T(0) < T2 - T1, then schedule T(0) will be
scheduled and T(1) scheduled at T2 – T1
 T(0) + T(1) < T2 –T1, then schedule both T(0) and
T(1)
 T(0) + T (1) > T2 – T1 && T(0)< T2 –T1, then
schedule both T(0) and T(1), but cut the report
length of T(1)

23. Advantages
 Removes bandwidth starvation
 Handling both delay and non-delay sensitive
traffic present in Triple Play concept.
 Maintain a trade-off between throughput ad
delay.
Disadvantages
 When the packet size increased beyond 9000
bytes, throughput is decreasing and delay is
getting increased.

29.  Simulation of Long-Reached PON with ONUs
located at far distances beyond 100 Km.
 Real time video relaying through the network
and see the network efficiency as increasing the
number of ONUs
 Buffer reduction and ONU power saving
mechanism to promote Green Networking.
 Implementation of various security aspects to
eliminate packet spoofing by hackers
 Use of multiple 1G NIC cards to be act as ONUs
and running processes on each systems to make
it end users. The 10G NIC cards can be used
together to make it as server (OLT) for
transmissions

30.  F. Effenberger and T.S. E1-Bawab, “Passive Optical Network(PONs):
Past, Present and Future,” Optical Switching and Networking”, vol.6 ,
pp-143-150,2009
 L. J. Hasler, "Design and Evaluation of Buffering Strategies for
Network Coding in Ethernet Passive Optical Network," Master,
Electrical Engineering and Computer Sciences, Berlin University of
Technology, Berlin.
 G. Kramer, Ethernet Passive Optical Networks: McGraw-Hill, 2005
 C. Lam, Passive Optical Networks, Principles and Practices: Elsevier
Publications, 2007.
 Rajiv Ramaswami, Kumar N. Sivarajan, Galen H.Sasaki, Optical
Networks, a practical perspective: Elsevier, 2012.
 R. Roy, et al., "Performance of 10G-EPON," Communications
Magazine, IEEE, vol. 49, pp. 78-85, 2011.
 J. M. Steve Gorshe. (2009). Introduction to IEEE 802.3av 10Gbit/s
Ethernet Passive Optical Networks (10G EPON).
 F. Aurzada et al., “Delay Analysis of Ethernet Passive Optical
Networks with Gated Service,” Arizona State University Technical
Report, Mar. 2007.

31.  First of all I would like to bow at the lotus feet of my beloved
divine mother and the Founder Chancellor Bhagawan Sri Sathya
Sai Baba for his invisible grace that he has been showering all
through the project.
 I profusely thank my guide Dr. S.Siva Sankara Sai, Head of
Department of Physics, Sri Sathya Sai Institute of Higher
Learning for giving me the chance to work in the project.
 I thank Prof. Krishna Moorthy Shivalingam, IIT Madras for
guiding us all through the project. I profusely thank Sri C.M
Ganesh, IIT Madras and A.Narayanswamy, Sumitomo Electric
Inc., USA without whom this project would not have been
complete and successful.
 I thank all my classmates G.V Sivasree, Sowmendran.P, Sumukh
Nandan.R, Bhanu Prakash Gadi, Satyabrata Mohanty and Murali
Ravi