You are on page 1of 53

A Study of Dynamic Bandwidth Allocation Algorithms for Passive Optical Networks and Simulation of 10G-EPON Using OPNET Modeler

By Christos Panagiotou

Supervisors Professor Ioannis Tomkos Konstantinos Kanonakis


1 / 53

1.

Introduction in PON networks ..................................................................................................... 3 1.1 PON History ..................................................................................................................................... 3 1.2 PON Standards, limitations and advantages ...................................................................... 3 1.3 PON operation ................................................................................................................................ 4

2.

Past/Current Research Directions in PONs............................................................................. 6 2.1 TDM-WDM PONs ............................................................................................................................ 6 2.2 Split-ratio ......................................................................................................................................... 6 2.3 Long Reach PONs .......................................................................................................................... 7 2.4 Energy-efficient PONs ................................................................................................................. 7

3.

Bandwidth Allocation algorithms in PONs .............................................................................. 8 3.1 Need for DBA .................................................................................................................................. 8 3.2 MAC Operation ............................................................................................................................... 8 3.3 Classification of DBA schemes ................................................................................................. 9

4.

Detailed literature review of DBA schemes for PONs ....................................................... 11 4.1 Classification of Papers............................................................................................................. 29

5. 6.

Conclusion ......................................................................................................................................... 30 OPNET ................................................................................................................................................. 31 6.1 What is OPNET and how it is used ........................................................................................ 31 6.2 Closer look into OPNET and how it works.......................................................................... 32

7.

10G EPON simulation .................................................................................................................... 37 7.1 ONU .................................................................................................................................................. 37 7.2 OLT.................................................................................................................................................... 39 7.3 Bandwidth allocation scheme simulation .......................................................................... 42 7.4 Simulation output and conclusions ...................................................................................... 43 7.5 Current condition and future work and research............................................................ 47

8.

Conclusion ......................................................................................................................................... 48

Appendix: Source Code......................................................................................................................... 49 References ................................................................................................................................................ 53

2 / 53

1. Introduction in PON networks

1.1 PON History


In the last 10 years, the telecoms system has transitioned from a copper-based infrastructure to a fiber-based infrastructure. The changeover started with the wide area networks (WANs) that offer connection among urban centers and advanced with the metropolitan area networks (MANs) that offer connection among service providers areas inside a city region. Simultaneously, local area networks (LANs) that interconnect nodes inside a private site have experienced common bit rates migrate from 10 Mb/s to 1 Gb/s through copper wiring. Even though important bandwidth enhancements happened in the service provider networking sites (i. e., WANs and MANs), along with at the client premises (i. e., LANs), the connection amongst the individual client sites and the public service provider sites failed to have the similar degree of improvement. This so-called access network that delivers the connection between individual and public networks continue to depend on an ageing copper system. The xDSL and cable modem technology improvements created minor enhancements in bandwidth efficiency but didn't wide open the bottleneck that is present within access networks. A fiber system becomes necessary in the access networks to deliver increased bit rates, in addition to more versatility. From the provider perception, access network connections have diverse income dynamics compared to connections in the WAN and MAN. Although WAN and MAN links hold the bit channels of several profit making clients, access network connections hold just one or few profit making bit channels. That is why, access networks are extremely vulnerable to cost. Cost concerns are delaying the deployment of a innovative physical plant in the access networks. Implementing a passive optical network (PON) between internet service providers and client premises can offer an economical as well as adaptable infrastructure that will supply the needed bandwidth to clients for several years ahead. PONs are a system where a shared fiber channel is created by employing a passive optical splitter/combiner in the physical layer. Sharing the fiber channel signifies lower cost in the physical fiber deployment, and making use of passive elements in the physical plant signifies lowered regular expenses simply by not sustaining remote control services with electric power. These lowered expenses make PONs a competitive option for access networks, which are inherently expense sensitive.

1.2 PON Standards, limitations and advantages


At a top level, PONs are categorized through the utilized link-layer protocol. Whereas an asynchronous transfer mode (ATM) PON (APON) uses ATM, an Ethernet PON (EPON) uses Ethernet, and a gigabit PON (GPON) utilizes the GPON encapsulation method (GEM) along with ATM cells to back up Ethernet. The International Telecommunication Union (ITU) has created specifications for APONs: G. 983 broadband PON (BPON), along with GPONs: G. 984 gigabit-capable PON (GPON). The IEEE has created a standard for EPONs: IEEE 802.3ah Ethernet in the first mile. BPON is the earliest PON standard, described in the mid-1990s, even though there's an installed platform of BPON, the majority of the new industry deployment concentration is currently on EPON/GE-PON. GE-PON, also known as EPON, is implemented extensively in Japan and offers for a symmetrical 1.0 Gbps data rate in both directions, upstream and downstream. It's the very first gigabit PON technology to accomplish excessive volume deployment.

3 / 53

Probably the most essential benefits of GE-PON are the usage of native Ethernet transport protocols. Low-cost asynchronous Ethernet has been used in the extensive data network for decades and there are important economies of scale related to gigabit Ethernet elements like optical interfaces. The asynchronous design of Ethernet allows Ethernet-based network devices to become lower cost compared to equivalent clock-synchronous ATM or SONET-based devices. GPON has acquired lots of interest ever since the ITU released the ITU-T G.984 recommendation in 2003. However, the recommendations remain in flux with specifications even now being modified. Consequently, GPON is not yet extensively implemented. GPON is dependent on the prior ITU BPON standard but has several characteristics to GE-PON. Launched in 2003 as ITU-T G.984, GPON relies on a brand new native Generic Encapsulation Method (GEM) transport layer which sustains several "non-native" transport protocols such as ATM, Ethernet, as well as TDM. The initial purpose was to expand support and industry popularity through supporting a number of protocols, but the impact has been to increase complexity to those techniques not needing extra protocol support. A vital feature is the 2.5 Gbps downstream data rate and also the 1.25 Gbps upstream data rate. GPON functions in an equivalent manner to GE-PON when supporting Ethernet since it's main transport protocol. However due to the fact Ethernet, gigabit Ethernet and 10Gb Ethernet don't support a 2.5 GHz clock rate, in contrast to GE-PON, GPON won't take advantage of the option of low-cost Ethernet optical elements. Below, a table that summarizes different standards of PON.

Fig.1.2

1.3 PON operation


A PON typically contains a natural tree topology, where one optical line terminal (OLT) located in the central office of the provider links a number of optical network units (ONUs) in the area. The OLT is linked to the ONUs using a feeder fiber that is then divided employing a 1: N optical splitter/combiner make it possible for the ONUs to share the optical fiber. The transmission route out of OLT to ONU is known as downstream and works as a broadcast carrier. The transmission route through the ONUs towards the OLT is known as upstream. The upstream impulses propagate from ONU to OLT nevertheless aren't returned back to every ONU; consequently, the PON isn't a broadcast carrier in the upstream route. The EPON is a multipoint- to-point carrier, in which the ONUs can't identify each others transmission since the upstream optical transmission isn't received from the ONUs. Nevertheless, ONUs share exactly the same
4 / 53

fiber; therefore, their transmissions may conflict, and contention solution should be carried out. To prevent collisions in the upstream route, time division multiplexing (TDM) or wavelength division multiplexing (WDM) may be used. WDM offers a great deal of bandwidth to every end user, but demands that every ONU make use of a distinctive wavelength, that offers inventory problems for providers that have to stock numerous ONU kinds. TDM permits all of ONUs to share just one wavelength, therefore, decreasing the amount of transceivers in the OLT and permitting an individual ONU kind. First era PONs utilize wavelengths to split up the upstream and downstream channels but make use of TDM to prevent upstream transmission collisions among ONUs. As a result of topology of the PON, MAC protocols which depend on connection among all nodes can't be applied. A PON provides connection from the OLT to all ONUs within the downstream and through every ONU towards the OLT in the upstream (i.e., only the OLT offers connection for all nodes). This connection design dictates the usage of a centralized MAC protocol located in the OLT. This results in a polling-based MAC, in which the OLT polls ONUs and allows them having access to the shared PON carrier. This issue is discussed in later chapter.

Fig.1.3

5 / 53

2. Past/Current Research Directions in PONs

2.1 TDM-WDM PONs


The broadband access network, also called the last mile network as discussed above, links the provider central offices to the enterprise and home customers. Due to the intense boost in bandwidth requirement, the way to provide the higher capacity of the optical fiber nearer to the home homes and small enterprises is a crucial solution in the future broadband access networks. It is through this perception that the FTTx models-fiber to the home (FTTH), fiber to the curb (FTTC), fiber to the premises (FTTP), etc. -were launched. Providing the capability for approximately 100 Mbps or even more bandwidth to every user, the FTTx designs intent at offering fiber straight away to home, or perhaps quite close to the home, through where solutions like DSL or wi-fi may take over. In the backbone networks, capacities more than 1 Tb/s over a single fiber are already accomplished through wavelength division- multiplexing technology using many channels. The introduction of low-cost and high-speed Ethernet based local area networks is speeding up the needs for high-speed connections at the same time. In order to break the bottleneck between the LAN and the ultrahigh-capacity backbone sites, cost-effective as well as high-performance access options are desired. Passive optical networks have been discovered among the appealing access solutions, as proven through the numerous study and standardization attempts emphasizing this particular area. Presently, the commonly approved optical access solution is the time-division multiplexed PON. TDM-PONs merge the higher capacity available through optical fiber while using reduced set up and maintenance expense of a passive system. The broadcast transmission in the downstream as well as the time-sharing transmissions in the upstream reduces the bandwidth of unique customers, however the producing reduced transceiver count reduces the cost good enough in order to explain the tradeoff. One simple method to raise the capacity is to allocate some wavelengths to every customer for down/upstream transmissions, leading to the WDM-PON. WDM-PONs develop point-to-point connections between the central office and every customer, therefore no sharing is necessary. Nevertheless, good performance is not without having increased cost-whenever the user turns off his / her connection, the equivalent transceiver in the CO is nonproductive and is not able to service different customers, in which case the system resource is actually left unused. WDM-PONs are high-performance but high quality access options nowadays. On the other hand, since bandwidth needs raise and optical element costs reduce, WDM-PONs may become far more useful and substantial. Regrettably, the migration plan from TDM to WDM is not simple. A complete WDM-PON has specific transmitters/receivers for every user and wavelength-routing equipment in the system. The price needed to include these brand new wavelengths and adjust the structure to back up WDM, amount to a significant obstacle to use. Recommended options offer cost-effective and easy service enhancements with small effect on the current TDM facilities.

2.2 Split-ratio
Based on the existing 2.5Gb/s GPON standard, the highest reach is restricted to 20km and the most frequent setup to accomplish a bandwidth of around 150Mb/s per residence demands using a rather small splitting ratio 1: 16. Attempts to improve the possible PON reach to 100km (i. e. Long Reach PONs) and the splitting ratio, whilst sustaining an increased access bandwidth per end user, are several years underway, although much more sophisticated PON technologies (e. g. WDM-TDMA hybrid PONs or OFDMA PONs) guarantee to support a higher splitting ratio and/or lengthier distances.

6 / 53

According to [1] the raise of the splitting ratio results in an expense reduction because of the reduced amount of fibers utilized but this isn't so important as the ducts and ditches dimensions which are employed in existing FTTH deployments do not let for the maximum benefit to be reached. A raise of the splitting ratio isn't growing linearly the expense shavings as opposed to older PONs. Nevertheless, the outcomes for 1: 512 splitting ratio declare that over 25% outside plant financial savings should be anticipated when compared to PONs with a splitting ration of 1: 16.

2.3 Long Reach PONs


Even though PON infrastructures can offer sufficient bandwidth for present purposes, both the continuous raise of the amount of customers and the bandwidth needs of the innovative emerging services, require an enhance of this kind of access networks. As a result, it exists a growing interest in the growth of optically amplified PONs to be able to expand the reach and the split ratio of these networks, therefore resulting in the so-called long-reach PONs. These network architectures can be cost-effective because they make easier the infrastructure for the reason that access and the metro networks can be merged into a individual network. As mentioned, Long-reach Passive Optical Networks try to mix the capacity of metro and access networks by increasing the reach and split ratio of the traditional PONs. LR-PONs seem as effective options having feeder distances close to 100km as well as high split ratios up to 1000-way. However, transmission of the signals in long distances as much as 100km results in higher propagation delay while large split ratio can result in lengthy cycle times leading to big queue occupancies and extensive packet delays. For this reason, dynamic bandwidth allocation algorithms are proposed in the next chapters to decrease average packet delays in LR-PONs.

2.4 Energy-efficient PONs


Although EPONs have quickly acquired great recognition in broadband access networks, nevertheless the present standard doesn't have any administration protocols attempting at decreasing energy consumption. The primary thought to improve power efficiency is to swap ONUs to sleep mode and decide an appropriate wakeup time schedule at the OLT. Of course, there is a trade-off between increasing the energy saving and ensuring the network efficiency. Another method is through taking into consideration the distinct network usage habits of various type of end users (i. e., day-to-day bandwidth demand profiles). End users (e.g., company and home users) can be allocated effectively to various wavelengths in LR-PONs with tree-and-branch or ring-and-spur topologies at networking planning phase and accomplish high network utilization all the time. The behavior-aware end user task approach accomplishes an important enhancement over the conventional task approach with regards to utilized wavelengths and could therefore conserves the power absorbed by the LR-PON.

7 / 53

3. Bandwidth Allocation algorithms in PONs

3.1 Need for DBA


To understand the significance of DBA in PONs, be aware that the traffic on private links in the access network is pretty bursty. This is contrary to MANS or WANs, in which the bandwidth needs are rather steady because of the aggregation of several traffic sources. In an access network every link symbolizes just one or small group of clients with really bursty traffic situations as a result of a small amount of ON/OFF resources. Because of these bursty dynamics, the bandwidth demands differ extensively as time passes. Consequently, static allocation of bandwidth to unique subscribers (or sets of subscribers) in a PON is extremely ineffective. Making use of a DBA algorithm which adapts to immediate bandwidth needs is a lot more effective, taking advantage of the advantages of statistical multiplexing. Therefore, DBA is a essential function for EPON structure. The DBA which runs at the OLT is responsible for offering statistical multiplexing. The OLT demands immediate bandwidth requirement details from every ONU to generate access decisions. Getting this exact info isn't feasible because of the non-zero propagation delays on a PON, generally approximately 100 sec, which is considerably bigger than the transmission period of the maximum size Ethernet frame: 12.3 sec. The ONUs need to record their instant queue capacities in a control frame and pass on this throughout the PON to the OLT.

3.2 MAC Operation


To effectively make use of the small upstream bandwidth, an PON system need to use a medium access control (MAC) procedure to arbitrate the usage of the shared channel to prevent data collisions in the upstream route. Bandwidth negotiation is to exchange information between the OLT and each ONU in order for each ONU to report its bandwidth demand to the OLT and for the OLT to send its bandwidth allocation decision to each ONU. For this purpose, IEEE 802.3ah defines a multipoint control protocol (MPCP) to support bandwidth negotiation between the OLT and ONUs in EPON, including two 64-bytes MAC control messages: REPORT and GATE. The REPORT message is generated by each ONU to report its queue status to the OLT. The OLT allocates bandwidth for each ONU based on the queue status information contained in the received REPORT message, and uses the GATE message to deliver its bandwidth allocation decision to each ONU. One way to deliver the bandwidth request or REPORT message is to dedicate a very short timeslot in the upstream channel. This requires twice laser on/off for one upstream transmission from each ONU. Another way is to piggyback the bandwidth request or REPORT message at the end of a data timeslot, which reduces laser on/off times into one per transmission for each ONU, and thus reduces the physical-layer power overhead and the inter- frame guard. MPCP is a signaling method for assisting dynamic bandwidth allocation and arbitrating the transmissions of several ONUs in an EPON structure. It exists at the MAC control layer and has two function methods: normal method and auto-discovery method. In the normal method, MPCP depends on two Ethernet control messages, GATE and REPORT, to assign bandwidth to every ONU. The GATE message is employed by the OLT to assign a transmission window to an ONU. The REPORT message is employed by an ONU to report its regional conditions to the OLT. In the auto-discovery method, the process depends on three control messages, REGISTER, REG- ISTER_REQUEST, as well as REGISTER_ACK, that are utilized to discover and enroll a recently connected ONU, and to gather appropriate details about that ONU, like the round-trip delay and MAC address. In its normal operation, MPCP in the OLT gets a request from the higher MAC client layer to transmit a GATE message to a specific ONU. After receiving this type of request,
8 / 53

MPCP will timestamp the GATE message using its local time and then send the message to the ONU. The GATE message usually includes a granted start time, a granted transmission window, and a 4-byte timestamp, that is helpful to determine the round-trip time amongst the OLT and the ONU. When the ONU gets the GATE message, it programs its local register using the figures included in the GATE message. At the same time, it also changes its local clock to that of the timestamp taken out from the received GATE message to be able to preserve synchronization with the OLT. At the granted start time, the ONU will begin to transfer information for as much as the window size. The transmission might include several data packets, based on the window size and the queue size in the ONU. No packet fragmentation is permitted throughout the transmission. If the following packet cannot be transmitted in the present window, it will be deferred to another window. A REPORT message is sent by an ONU in the designated transmission window along with a data packet. It can be sent instantly or upon request either in the beginning or by the end of a window. A REPORT is produced in the MAC client layer and is time stamped at the MAC layer. It usually provides the bandwidth need of an ONU using the instant queue length of that ONU. The ONU also needs to take into account extra over head in its request, together with a 64-bit frame preamble and a 96-bit inter-frame space related to each Ethernet packet. As soon as a REPORT message is obtained by the OLT, it is handed to MAC client layer, which is in charge of bandwidth allocation and recalculation of the round- trip delay to the source ONU. It must be noticed that MPCP is not related to any kind of specific bandwidth allocation scheme and transmission scheduling protocol, and makes it possible for them to be vendor-specific. To develop a competent polling protocol depending on MPCP, a number of issues should be taken into consideration, such as highest bandwidth limit, channel utilization, and packet scheduling.

Fig.3.2

3.3 Classification of DBA schemes


Journals and papers that refer to novel DBA algorithms for Passive Optical Networks can be categorized as follows:

9 / 53

Fig.3.3: Classification

10 / 53

4. Detailed literature review of DBA schemes for PONs


A lot of papers refer to novel DBA algorithms in different types of PONs. In this research, out of many, 62 papers were selected and studied more extensively. Summaries of these papers are following. 1.Paper: An efficient solution for mitigating light-load penalty in EPONs Authors: Yongqing Zhu, Maode Ma, Tee Hiang Cheng Summary: In this paper, a MAC protocol is proposed for EPONs upstream transmission. The advantage of this BGP-based MAC protocol is that it can greatly mitigate the light-load penalty. In the BGP scheme, the OLT decides the transmission window size granted to the next ONU according to the REPLY message from the previous ONU and the maximum transmission window size. Extensive simulation experiments have been conducted to evaluate the performance of the proposed solution combining the BGP scheme with the SPQ mechanism that prove that this solution is effective and efficient to diminish the light-load penalty, while it can provide diffserv.

2.Paper: Bandwidth allocation with a particle swarm meta-heuristic for Ethernet passive optical networks Authors: Un Gi Joo, Alice E. Smith Summary: This paper sets up an optimal bandwidth allocation problem that maximizes throughput and weighted fairness simultaneously. Then 2 algorithms are proposed. The first algorithm is a very fast heuristic one, which is optimal under certain restricted conditions, and the second one is a PSO meta-heuristic. The results of tests performed show that the heuristic algorithms perform well over a range of problem instances and types (symmetric versus asymmetric and different load values and varying numbers of ONUs). As another avenue of future research, for a specific amount of allocated bandwidth to each ONU, the sequencing rule of Weighted Shortest Processing Time (WSPT) could be used to reduce delay time.

3.Paper: Dynamic Bandwidth Allocation for Quality-of-Service Over Ethernet PONs Authors: Chadi M. Assi, Student Member, IEEE, Yinghua Ye, Member, IEEE, Sudhir Dixit, Member, IEEE, and Mohamed A. Ali Summary: This paper proposes that MPCP should be used as the basis for upstream transmission DBA algorithms. It differentiates DBA algorithms that support QoS and evaluate them using an event-driven packet-based simulation model that is developed using C++. It is shown that strict priority-based bandwidth allocation, under specified assumptions for traffic behavior, will result in an unexpected behavior for certain traffic classes and it is suggested the use of queue management with priority scheduling to alleviate this problem. Moreover, it is showed that algorithms that perform early bandwidth allocation for lightly loaded ONUs result in better performance in terms of average and maximum packet delay, as well as network throughput.

11 / 53

4.Paper: IPACT: A Dynamic Protocol for an Ethernet PON (EPON) Authors: Glen Kramer and Biswanath Mukherjee, University of California, Davis Gerry Pesavento, Alloptic, Inc. Livermore Summary: This paper introduces IPACT(interleaved polling scheme with an adaptive cycle time) for supporting dynamic bandwidth distribution. The advantage of IPACT is the use of a single downstream and a single upstream wavelength, and the ability to provision a fractional wavelength capacity to each user. It is also scalable with the number of users in the system. Since each ONU uses the window size required at the moment of the transmission, the polling cycle time adapts to the instantaneous queue loads, leading to an adaptive cycle time. Reduced cycle time leads to an increase in the amount of best-effort bandwidth available to busy ONUs. This increase depends on their bandwidth needs. Network performance under varying offered loads is analyzed in evaluation section.

5.Paper: Just-in-Time Scheduling for Multichannel EPONs Authors: Michael P. McGarry, Member, IEEE, Martin Reisslein, Senior Member, IEEE, Charles J. Colbourn, Martin Maier, Member, IEEE, Frank Aurzada, and Michael Scheutzow Summary: This paper proposes a two-layer structure of scheduling (framework and policy). In online JIT scheduling that is proposed, channel availability and not ONU reports determine the scheduling process. It is also more work conserving and therefore is more efficient than the non work conserving offline scheduling framework where the OLT waits for all ONU REPORT messages to make access decisions. In the simulation taken using CSIM software, it is shown that the online JIT scheduling framework has some utility in lowering the average queuing delay as well as provide differentiate ONU service. As a trade-off, on-line JIT is more complex. Future research can study how this online JIT framework affects grant sizing techniques.

6.Paper: Modified Smallest Available Report First: New Dynamic Bandwidth Allocation schemes in QoS-capable EPONs Authors: Mustafa Vahabzadeh Dolama, Akbar Ghaffarpour Rahbar Summary: In this paper, the authors provide two variations of SARF (Smallest Available Report First) algorithm, the Modified Smallest Available Report First and the Modified Smallest Available Report First with Hybrid Grant Protocol to support QoS in EPON. MSARF and MSARF_HGP distribute the bandwidth based on the waiting time of packets at the head of buffers and the smallest reported buffer length. The schemes are designed to decrease the average packet delay and packet loss. the channel utilization in MSARF is high because it does not need more guard time. A comparison between the two schemes shows us that MSARF_HGP is more suitable than MSARF at low traffic loads and MSARF is more suitable than MSARF_HGP at high traffic loads. OLT can use a hybrid scheme of MSARF and MSARF_HGP to achieve a better performance. Thus, OLT can use the MSARF_HGP at low traffic loads and MSARF at high traffic loads.

7.Paper: On the fairness of dynamic bandwidth allocation schemes in Ethernet passive optical networks
12 / 53

Authors: Xiaofeng Bai a, Abdallah Shami a,*, Chadi Assi Summary: In this paper, a robust weight-based scheme (W-DBA) is presented. It constantly maintains a maximum fairness index throughout the algorithm operation. An OLT can perform W-DBA scheme as follows: Step 1: According to the total available bandwidth, set a bandwidth threshold for each ONU based on weighted MaxMin fairness principle. Step 2: Iterate the bandwidth request of each ONU and grant as much as requested bandwidth to ONUs that are requesting no more resource than their corresponding threshold. These ONUs are termed as satisfiable ONUs. Step 3: If one or more satisfiable ONUs were found in step 2, repeat step 1. Otherwise grant each remaining ONU its corresponding threshold and end the DBA process. Finally, an evaluation is performed between M-DBA1 and W-DBA in terms of average packet delay, upstream link utilization and correlation coefficient.

8.Paper: Quality of Service Support over SUCCESS-DWA: A Highly Evolutional and Cost-Effective Optical Access Network Authors: Yu-Li Hsueh, Matthew S. Rogge, Wei-Tao Shaw, Shu Yamamoto, Leonid G. Kazovsky Summary: In this paper scheduling algorithms for the SUCCESS-DWA PON are proposed, in which WDM techniques are employed. In the downstream, the VOQ and TL-buffering schemes are considered. VOQ, unlike TL, cannot distinguish between the classes. Moreover,TL is less complex. In the upstream, a second laser in high-SLA ONU is considered. It is shown that the HP traffics from high-SLA ONUs suffer from less delay when compared to the BE traffics from low-SLA ONUs. The SUCCESS-DWA PON allows an upgrade path from the TDM PONS to the WDM PONs.

9.Paper: Resources allocation in an Intserv/Diffserv integrated EPON system Authors: LIU Qiong , LI Hui, JI Yue-feng, QIAO Yao-jun Summary: This paper proposes to apply integrated services model (Intserv) in Diffserv-based EPON, which uses per-flow processing to guarantee QoS. A DBA algorithm is presented is proposed that supports hybrid QoS scheme of the architecture. The on-the-fly method for the bandwidth request of GS queues is used to remove the idle time of upstream channel. Finally, through the simulation, it is proved that this system can provide better per-flow quality guarantee for reservation services without any negative effect on QoS of Diffserv services. Further study can be made on the increased load of network management due to SMB protocol, edge routers capabilities on supporting both Diffserv and Intserv and hardware implementation of the experinment.

10.Paper: Bandwidth Distribution Solutions for Performance Enhancement in Long-Reach Passive Optical Networks Authors: Burak Kantarci, Member, IEEE, and Hussein T. Mouftah, Fellow, IEEE, Summary: This article presents a comprehensive survey on the dynamic bandwidth allocation schemes for LR PONs. In the article, a comparative classification of the proposed schemes based on their quality-of-service awareness is proposed. A detailed comparison and summary of the schemes are also
13 / 53

presented with respect to the base PON type, maximum distance, delay/PDV/packet loss performance and the runtime overheads. Open issues for future study are schemes meeting a wide range of QoS metrics need to be explored as well as DBA schemes for hybrid TDM-WDM LR-PONs and energy-efficient schemes.

11.Paper: A traffic-class burst-polling based delta DBA scheme for QoS in distributed EPONs Authors: Yeon-Mo Yanga, Ji-Myong Nhob, N.P. Mahalika, Kiseon Kima, Byung-Ha Ahn Summary: In this paper a TCBP-DDBA (traffic-class burst-polling based delta dynamic bandwidth allocation) scheme is proposed. This scheme can effectively allow all ONUs to fairly share the upstream according to their bandwidth demands. The network resources are efficiently utilized and adaptively allocated to three traffic classes by guaranteeing the requested QoS. The new scheme has better performance in terms of packet delay, PDV and utilization compared with other DBA algorithms. It will also provide better downstream link utilization. However, a more sophisticated and rigorous algorithm could be developed to satisfy the SLAs.

12.Paper: Hierarchical scheduling to support differentiated services in Ethernet passive optical networks Authors: Yongqing Zhu, Maode Ma *, Tee Hiang Cheng Summary: In this paper a scheme that applies hierarchical scheduling for the MAC layer in EPONs is deployed. It consists of the inter-ONU scheduling and intra-ONU scheduling, to provide DiffServ as well as share the upstream among several users. The inter-ONU scheduling among the ONUs uses the efficient MAC scheme. It provides services to different ONUs based on SLAs. The intra-ONU scheduling within the ONU adopts the two-stage queuing (TSQ) mechanism to provide DiffServ to different traffics in the same ONU. It can provide better services to the lower priority traffics to achieve better fairness than the strict priority queuing (SPQ) mechanism. The combined scheme greatly mitigates the light-load penalty compared to other schemes. The simulation results show that the scheme for the hierarchical scheduling can effectively provide differentiated services for different types of traffics, and also achieve the fairness to some extent.

13.Paper: User-oriented hierarchical bandwidth scheduling for Ethernet passive optical networks Authors: Yongning Yin, Gee-Swee Poo Summary: in this paper new User-oriented Hierarchical bandwidth Scheduling Algorithms (UHSAs) are proposed that support differentiated services and guaranteed fairness among end users. For inter-ONU scheduling, an improved hybrid cycle approach that separates a frame into a static part for high priority traffic and an adaptive dynamic part for low priority traffic are adopted. For intra-ONU scheduling, credit-based scheduling approach is followed to guarantee fairness among end users. To improve scheduling efficiency and lower queue management complexity, a novel credit-based common queue (CCQ) for each traffic class is designed to enhance scheduling architecture and minimize average number of queues in the ONU. Moreover, a transmission priority scheme is proposed for different queue groups, which together with CCQ mechanism serves the objective of improving delay and delay variation
14 / 53

performance of high priority traffic, guaranteeing throughput for bandwidth sensitive medium priority traffic, as well as providing fairness among different users. The UHSAs exhibit a feature of multiple transmission opportunities (M-opportunities) per-cycle for high priority traffic that is significant in improving delay and delay variation performance for high priority traffic as compared with previous solutions of single transmission opportunity per-cycle. Detailed simulation experiments show that (1) the UHSA algorithms with M-opportunities per-cycle solution significantly improve delay and delay variation performance by reducing AD over 27.2% and DV over 19.6% as compared with traditional S-opportunity per-cycle solutions (2) the CCQ mechanism substantially enhances scheduling efficiency with a large percent of packets served from CCQ groups and lowers queue management complexity with a reduction over 22.4% of AQ at all network loads.

14.Paper: Flexible scheduling ensuring inter-ONU-group fairness in 1G/10G EPON coexistence system Authors: LI Qi , LI Hui, JI Yue-feng Summary: In this paper a novel bandwidth allocation algorithm called weight-optimized dynamic bandwidth allocation for coexistence of 1G/10G EPON (WOCE-DBA) for the system is proposed. The proposed WOCE-DBA can make effective adjustment to mitigate the unfairness bandwidth allocation between different ONUs groups observed under CE-DBA, which does not take intra-ONU-group unfairness into consideration. From the simulation results, it can be seen that WOCE-DBA outperforms CE-DBA in that it reduces the packet delay difference between the 1G-US and 10G-US ONU group, thus eliminating the unfairness in bandwidth sharing between the two groups. Also, it increases the average system throughput

15.Paper: OSLG: A new granting scheme in WDM Ethernet passive optical networks Authors: Ali Razmkhah, Akbar Ghaffarpour Rahbar Summary: In this paper a new granting scheme for WDM EPONs, called ONU Side Limited Granting (OSLG) that saves upstream bandwidth, thus resulting in decreasing queuing delay and packet drop ratio is proposed. In OSLG instead of OLT, each ONU determines its transmission window. Two OSLG algorithms are proposed in this paper: the OSLG_GA algorithm that determines the size of its transmission window in such a way that the bandwidth wastage problem is relieved, and the OSLG_SC algorithm that saves unused bandwidth for more bandwidth utilization in the future. The OSLG can be used as granting scheme of any DBA to provide better performance in the terms of packet drop ratio and queuing delay. Both algorithms are quite complex. Evaluation results show that OSLG_SC has the best performance in terms of queuing delay and packet drop ratio.

16.Paper: Performance Evaluation of Standard IPACT for Future Long Reach Passive Optical Networks (LR-PON) Authors: A. M. Ragheb, M. Elnamaky, H. Fathallah, and H. T. Mouftah Summary: In this paper the authors propose a new IPACT algorithm for EPON using Matlab Simulink tool. The IPACT algorithm has been tested under small coverage area and large coverage area(LR-PONs).
15 / 53

Results also show that the implementation of the IPACT algorithm for Long-Reach PON is not preferable. For future study, it is proposed propose for investigation a new polling protocol for LR-PON to simulate the multi-thread polling exiting algorithms.

17.Paper: Multi-Thread Polling: A Dynamic Bandwidth Distribution Scheme in Long-Reach PON Authors: Huan Song, Byoung-Whi Kim, and Biswanath Mukherjee Summary: This paper proposes a multi-thread polling algorithm to effectively and fairly distribute the upstream bandwidth dynamically in a LR-PON. This algorithm exploits the benefits of having multiple threads running simultaneously and enabling users to send bandwidth requests before receiving acknowledgement from the central office. Compared to other algorithms, the proposed algorithm has an advantage on average packet delay. Numerical results show that, by setting the proper initial thread interval and tuning threshold, the average upstream packet delay is decreased, especially at high traffic load; and the network throughput is increased.

18.Paper: A bandwidth assignment polling algorithm to enhance the efficiency in QoS long-reach EPONs Authors: N. Merayo, T. Jim enez, P. Fern andez, R.J. Dur an, R.M. Lorenzo, I. de Miguel and E.J. Abril Summary: A novel polling algorithm called LOHEDA is proposed to provide subscriber differentiation in the upstream in LR-PONs. As SPs prefer to operate at loads with no congestion and losses, the developed algorithm has been designed to improve the efficiency for those loads at which the network is likely to work. This new scheme permits to anticipate the transmission of some packets in order to take advantage of the wasted bandwidth between consecutive transmissions of ONUs. As a result, not only does this algorithm behave alike other polling algorithms for high loads, but it also improves the efficiency at low and medium loads, leading to a reduction in the average packet delay. The new algorithm compared to LIPSA, another algorithm for LR-PONs, achieves the same throughput efficiency and less average packet delay. This improvement in mean delay is getting bigger as the distance increases.

19.Paper: A New Dynamic Bandwidth Allocation Protocol with Quality of Service in Ethernet-based Passive Optical Networks Authors: Fu-Tai An, Yu-Li Hsueh, Kyeong Soo Kim, Ian M. White, and Leonid G. Kazovsky Summary: This paper proposes an algorithm that minimizes packet delay and delay variation for high priority traffic to ensure QoS. It is a hybrid slot-size/rate DBA MAC protocol which assigns a large slot size for low priority, such as best effort, traffic. This maximizes network throughput when offered load is high. The fairness for low priority is guaranteed by using queue length as weighting factor. The protocol also provides quasi-non-intrusive ranging. Ranging process will not disturb the high priority traffic. Simulation results show excellent network throughput. Simulation results also show low packet delay and packet delay variation for high priority traffic compare to conventional slot-size based DBA algorithm.

20.Paper: A New efficient dynamic MAC protocol for the delivery of multiple services over GPON
16 / 53

Authors: Jin Jiang John M. Senior Summary: In this paper, a new approach for bandwidth reporting is developed to achieve dynamic awareness of both the changes of ONU queue status and the consequences of previous allocations. In addition, a novel balance transferring mechanism that enables a register and transfer process for successive allocation across multiple frames is presented. By combining these two into a framebased scheduling procedure, an efficient dynamic bandwidth assignment MAC protocol is developed to support differentiated services over GPON. the new protocol compared to a conventional protocol, not only decreases delays, lowers delay variation, fairness, and supports for various QoS requirements in GPON, but it also contributes to a significant improvement in network performance with an average of 22% increase in the GPON upstream throughput.

21.Paper: Adaptive polling algorithm to provide subscriber and service differentiation in a Long-Reach EPON Authors: Noem Merayo Tamara Jimnez Ramn J. Durn Patricia Fernndez Ignacio de Miguel Rubn M. Lorenzo Evaristo J. Abril Summary: A novel interleaved polling algorithm for Long- Reach EPONs is proposed in this paper, in order to provide subscriber and class of service differentiation. It is demonstrated that the new polling algorithm applied to a 100km Long-Reach EPON performs better than centralized methods, where bandwidth prediction is needed to overcome the higher RTT in which ONUs cannot transmit. As polling methods in Long-Reach EPONs do not require prediction, they are much simpler and show less computational complexity than centralized schemes, avoiding the inaccuracy of bandwidth prediction. Simulation results show that the new algorithm decreases delay and the packet loss rate reduction compared to other algorithms that do not support IPACT.

22.Paper: An SLA-Based Energy-Efficient Scheduling Scheme for EPON with Sleep-Mode ONU Authors: Lei Shi I, Sang-Soo Lee II, and Biswanath Mukherjee Summary: In this paper, an SLA-based scheduling scheme for EPON in which OLT can adjust sleep time and ONU can quit sleep mode for sending expedited frames. Simulation results show that the ONU could save 60% of the energy consumed.

23.Paper: Bandwidth Allocation over EPONs:A Controllability Perspective Authors: Si Yin1, Yuanqiu Luo, Nirwan Ansari, and Ting Wang Summary: In this paper, a state space model has been proposed to analyze the controllability of different bandwidth allocation schemes over EPON. 3 schemes have been analyzed, the predictor-based DBA, the IPACT DBA, and SBA. The analysis shows that the predictor-based DBA is able to drive the upstream bandwidth allocation into the optimum state, while neither IPACT nor SBA can achieve the optimal state. Future analysis will focus on a more detailed analysis of the predictor-based DBA system.

17 / 53

24.Paper: Bandwidth Distribution with Adaptive Threshold-based Optical Burst Assembly in Long-Reach EPON Authors: Burak Kantarci and Hussein T. Mouftah Summary: It is well known that LR-PONs suffer from increased propagation delay. In this paper, a new bandwidth distribution approach which adopts adaptive-threshold based optical burst assembly of OBS networks and multi-server polling of LR-EPON is proposed. Numerical results show that the proposed algorithm decreases delay and improves buffer management.

25.Paper: Bandwidth management for WDM EPONs Authors: Michael P. McGarry and Martin Reisslein, Martin Maier, Ahmet Keha Summary: This paper discusses the problem of bandwidth management for WDM EPONs and utilized scheduling theory to find the best solution for the scheduling part of the problem. It is proved that a multidimensional scheduling approach using results from scheduling theory can provide much better bandwidth management by means of better wavelength utilization than a static wavelength assignment. Also, an online scheduling approach can provide lower queuing delays than a cyclical offline scheduling approach. However, a hybrid online-offline scheduler provides lower queuing delays as the network approaches full capacity.

26.Paper: DYNAMIC BANDWIDTH ALLOCATION ALGORITHM FOR DIFFERENTIATED SERVICES OVER WDM EPONS Authors: Kae Hsiang Kwong, David Harle, Ivan Andonovic Summary: In this paper, a new WDM-PON system is proposed where multiple wavelength channels are established in both upstream and downstream directions. The simulation results illustrate significant degradation of a single channel IPACT as the number of ONUs is increased. In this paper, the WDM IPACT-ST is proposed as one future upgrade solution for EPON network. In the proposed scheme, limited changes are required in both ONU and OLT to support multiple wavelength communications, representing a very cost-efficient model when compared to laying a new ONUs-OLT tree. The integration of differentiated services into WDM IPACT-ST is also demonstrated in this paper.

27.Paper: Dynamic bandwidth allocation algorithm for Next Generation Access Network Authors: Joo, Jung Min, Ban, Yun Ji Summary: In this paper a DBA algorithm is proposed that provides higher bandwidth efficiency and lower queuing delay of ONU. In simulation results it is shown that packet delay and queue length are lower compared to other DBA, SBA or IPACT algorithms.

18 / 53

28.Paper: Dynamic bandwidth allocation algorithm in Ethernet passive optical networks Authors: Hee-Jung Byun, Ji-Myung Nho and Jong-Tae Lim Summary: in this paper, an estimation-based dynamic bandwidth allocation algorithm to keep the queue length of each ONU low and to improve the packet delay on the EPON. The proposed algorithm is compared to SBA and IPACT and simulation shows it is more efficient in terms of queuing delay and high throughput

29.Paper: Dynamic Bandwidth Allocation for Long-Reach PON: Overcoming Performance Degradation Authors: Bjrn Skubic, Ericsson Research, Jiajia Chen and Jawwad Ahmed, Royal Institute of Technology KTH, Biao Chen, Zhejiang University, Lena Wosinska, Royal Institute of Technology KTH, Biswanath Mukherjee, University of California, Davis Summary: This article addresses the problem of high propagation delay in LR-PONs. A potential solution to the performance degradation is the introduction of a multi-threaded DBA. In this article, the authors examine for both GPON and EPON, the extent to which DBA performance degradation can be reduced by exploiting multi-threading. It is found that for both standards, multi-threading, if done properly, can be used to mitigate the performance degradation due to the increased reach. To make bandwidth allocation efficient, new schemes for coordinating the multiple threads are required in LR-PONs. The NA+ scheme is introduced and adapted to both GPON and EPON in order to eliminate the overgranting associated with multi-threading. NA+ coordinates work between the DBA processes. As a result, the advantages of multithreading can be fully exploited without the conventional side effects. Simulation results favor previous conclusion.

30.Paper: Dynamic Bandwidth Allocation Schemes in Hybrid TDM/WDM Passive Optical Networks Authors: Ahmad R. Dhaini, Chadi M. Assi, and Abdallah Shami Summary: In this paper, a simple upgrade architecture from EPON to WDM-PON is discussed. A new class of allocation algorithms (DWBAs) is proposed where inter- and intra-channel statistical multiplexing is exploited to improve the overall performance. The algorithms performance is evaluated through extensive simulation.

31.Paper: Dynamic Bandwidth assignment for Multi-service access in long-reach GPON Authors: C.-H. Chang, N. M. Alvarez, P. Kourtessis, J. M. Senior Summary: In this paper, a novel, two-state bandwidth assignment MAC protocol is demonstrated to provide service level differentiation in long-reach GPONs. More specifically, the dynamic TSD protocol manages to overlap the idle time slots in each packet transmission cycle with a virtual polling cycle to increase the effective transmission bandwidth by means of a prediction method to estimate the bandwidth requirement of each ONU. Network performance investigations of the TSD scheme versus a developed algorithm, have displayed significant increase in channel throughput with an improvement in packet loss rate to allow high link utilization rates over extended network traffic.
19 / 53

32.Paper:

Efficient Scheduling Disciplines for next Generation QoS-Aware GPON Networks

Authors: Konstantinos Kanonakis, Student Member, IEEE, Ioannis Tomkos, Member, IEEE Summary: This paper proposes a novel MAC scheduling algorithm for GPON networks that can offer efficient operation, combining QoS with high network utilization. It achieves improved performance by combining both fixed and dynamic bandwidth allocations along with predictive allocations and the novel concept of offset-based scheduling with flexible intervals, which means that it can effectively adapt to fluctuations of traffic and meet quality of service requirements. The system performance is evaluated using OPNET.

33.Paper: Energy Management Mechanism for Ethernet Passive Optical Networks (EPONs) Authors: Ying Yan, Shing-Wa Wong, Luca Valcarenghi, She-Hwa Yen, Divanilson R. Campelo, Shinji Yamashita, Leonid Kazovsky, and Lars Dittmann Summary: In this paper, we propose an Energy Management Mechanism (EMM) within the IEEE 802.3ah control scheme. The main idea is to switch ONUs to sleep mode and determine a suitable wakeup time schedule at OLT. Of course, in the proposed EMM there is a trade-off between maximizing the power saving and guaranteeing the network performance. Two types of downstream scheduling schemes, Upstream Centric Scheduling (UCS) and Downstream Centric Scheduling (DCS) are compared, which are different in the way they assign active and sleep states to ONUs. Simulation results in terms of energy consumption (-90%) and queuing delay (+10ms) are shown for the EMM.

34.Paper: Energy-Efficient Long-Reach Passive Optical Network: A Network Planning Approach Based on User Behaviors Authors: Lei Shi, Sang-Soo Lee, Huan Song, and Biswanath Mukherjee, Fellow, IEEE Summary: In this paper a network planning approach based on user behaviors is discussed. By considering the different network usage behaviors of different kind of users, they can be assigned efficiently to different wavelengths in tree-and-branch or ring-and-spur LR-PONs at networking planning stage and achieve high network utilization. Heuristic searches can provide such assignments and their approximate solutions are very close to the lower bound. The behavior-aware user assignment achieves significant improvement over the traditional method in terms of used wavelengths and could decrease the energy consumed by an LR-PON.

35.Paper: GATE-driven Dynamic Wavelength and Bandwidth Allocation for WDM EPONs Authors: Nelson Antunes, Christine Fricke, Philippe Robert and James Roberts Summary: In this paper a novel DWBA algorithm that is considerably more efficient than classical approaches is proposed.Traffic capacity is maximized while the mean packet delay is reduced. The algorithm is simple to implement and conforms to existing EPON standards. Closed-form traffic capacity formulas are given for a WDM-EPON where ONUs have full access to all wavelengths. A drive for future study can be how the potential for flexible scheduling can be exploited to realize effective inter-ONU
20 / 53

differentiation. 36.Paper: Improving the Efficiency of Online Upstream Scheduling and Wavelength Assignment in Hybrid WDM/TDMA EPON Networks Authors: Konstantinos Kanonakis, Member, IEEE, and Ioannis Tomkos, Member, IEEE Summary: This paper presents the online and offline scheduling algorithm for hybrid WDM/TDMA EPONs. Although online algorithm decreases frame delay in all terms, it doesnt utilize wavelengths efficiency, especially in the case that ONU is far away from OLT. Several low- and higher-complexity solutions are analyzed to overcome those inefficiencies, leading to improved utilization of network capacity and reduced frame delay. More specifically, the LFT algorithm, manages to reduce frame delay significantly, in low traffic though. EFT and LFT that adopt void filling decrease and increase wavelength utilization under all loading conditions but as a trade-off complexity of the algorithm is increased. VF is proportional to the number of ONUs in the network.

37.Paper: Interleaved Polling Algorithm with Inserted Cycles to Support Service Level Agreement in Long-Reach EPONs Authors: Tamara Jimnez, Noem Merayo*, Ramn J. Durn*, Patricia Fernndez*, Rubn M. Lorenzo*, Ignacio de Miguel*, Natalia Fernndez, Evaristo J. Abril Summary: In this paper, a novel dynamic bandwidth allocation algorithm for LR EPONs is proposed. The new algorithm differs clients as it takes into consideration the SLA contracted by each subscriber, achieving better performance for the highest SLAs in terms of the mean packet delay. Moreover, the algorithm is able to work with the same efficiency even when the LR EPON distance is extended to 225 km. Therefore, it keeps the same good values of throughput, mean packet delay and mean cycle time independently of the end to end distance.

38.Paper: Limited sharing with traffic prediction for dynamic bandwidth allocation and QoS provisioning over Ethernet passive optical networks Authors: Yuanqiu Luo and Nirwan Ansari Summary: In this paper the authors propose a dynamic bandwidth allocation scheme, i.e., limited sharing with traffic prediction (LSTP), for upstream channel sharing over EPONs. LSTP enables dynamic bandwidth negotiation between the OLT and its associated ONUs, alleviates data delay by predicting the traffic arrived during the waiting time and pre-reserving a portion of bandwidth for delivery. The performance of LSTP has been theoretically analyzed in terms of the prediction success probability, the delay reduction, and the data loss probability. The simulation results in OPNET also demonstrate that LSTP enhances the accuracy of the prediction of the incoming data during the waiting time.

39.Paper: Multi-Thread Polling: A Dynamic Bandwidth Distribution Scheme in Long-Reach PON Authors: Huan Song, Byoung-Whi Kim, and Biswanath Mukherjee
21 / 53

Summary: In this paper a multi-thread polling algorithm to effectively and fairly distribute the upstream bandwidth dynamically in LR-PONs is proposed. This algorithm exploits the benefits of having multiple polling processes running simultaneously and enabling users to send bandwidth requests before receiving acknowledgement from the CO. The proposed algorithm is compared to traditional DBA, and show its advantage on average packet delay. Key parameters of the algorithm are analyzed, such as initiating and tuning multiple threads, inter-thread scheduling, and fairness among users. Simulation results show the algorithms advantage to decrease the average packet delay and improve network throughput under varying offered loads.

40.Paper: Offset-Based Scheduling With Flexible Intervals for Evolving GPON Networks Authors: Konstantinos Kanonakis, Member, IEEE, and Ioannis Tomkos, Member, IEEE Summary: This paper introduces a novel framework for DBA in GPONs employing the Offset-Based Scheduling with Flexible Intervals (OSFI) concept in order to achieve distribution of bandwidth in such networks ensuring not only high system utilization but also clean-cut QoS differentiation based on the individual demands of user services. Moreover, ways to enhance efficiency in the case of next generation LR-GPONs are discussed. The algorithm is tested in OPNET and the results confirm its enhancements. 41.Paper: OLT Design Approach for Resilient Extended PON with OBS Dynamic Bandwidth Allocation Sharing the OLT Optical Resources Authors: Josep Segarra, Member, IEEE, Vicent Sales and Josep Prat, Member, IEEE Summary: In this paper, two designs for the OLT that shares its available optical resources in a FTTH access network, thus providing great traffic efficiency is proposed. The traffic is managed by a DBA algorithm based on OBS that suits the employed modulation format. Both OLT designs have been evaluated by simulations in Short Range Dependence and Long Range Dependence traffic behavior and Priority Queuing. The design with Tunable LDs in optical regions has been simulated with two TLDs and users distributed uniformly in several number of tPONs. The results show that the use of shared TLDs does not increase significantly the delay, while furnishes better network efficiency and protection failure. Finally, employing Priority Queuing with a Strict Priority discipline improves the DBA: the latency of prioritized packets is notably reduced.

42.Paper: On SLA Constraints in Dynamic Bandwidth Allocation for Long-Reach Passive Optical Networks Authors: Burak Kantarci, Member, IEEE, Hussein T. Mouftah, Fellow IEEE Summary: In this paper a QoS-aware bandwidth allocation scheme which is called as Periodic Gate Optimization with QoS-awareness (PGO-QoS) is proposed in order to decrease long propagation delay in LR-PONs. It is based mostly on multi-threaded MPCP. It is showed that PGO-QoS leads to shorter average delay than multi-threaded MPCP while slightly decreasing the packet loss ratio. As expected, larger buffer size increases the packet delay, especially under heavy traffic. However, changing the buffer size does not affect the relative performance of PGO-QoS to the multi-threaded MPCP. The simulation results also reveal the performance of PGO-Qos in terms of pre-specified delay bounds where each ONU runs an estimation to determine whether an incoming packet can reach the OLT within its pre-determined
22 / 53

delay bound.

43.Paper: Online excess bandwidth distribution for Ethernet passive optical networks Authors: Jason R. Ferguson, Martin Reisslein, and Michael P. McGarry Summary: In this paper online excess bandwidth distribution (OEBD) to overcome the stability problems of the existing offline excess bandwidth distribution mechanisms is proposed. It is found that OEBD with the correct parameter settings can provide lower queuing delays for LR PONs. Even for mid-range EPONs, OEBD provides a higher delay stability limit (higher maximum achievable channel utilization). The difference in stability limit is exacerbated for LR PONs. The flexibility provided by OEDB will allow network designers to selectively trade off lower queuing delays with fairness guarantees. There is still future research on OEBD. One important direction is to further comprehensively study the parameter setting for OEBD to ensure robust, good performance across a wide range of scenarios. Another avenue is to examine the compatibility of OEBD with emerging DBA strategies for long-range EPONs, such as multithread polling.

44.Paper: Packet Delay Variance and Bandwidth Allocation Algorithms for Extended-Reach GPON Authors: Timothy G. Smith, Rodney S. Tucker, Kerry Hinton and An V. Tran Summary: In this paper it is shown that for an extended reach GPON the delay in conventional report/allocate DBA algorithms increases the packet delay variance for CBR traffic. When multiple upstream packets are sent in the same upstream channel, significant delay occurs. An alternative method to perform allocations that satisfies the ITU standards and produces zero packet delay variance is proposed. By eliminating the Packet Delay Variance for CBR traffic, complicated receivers no longer require to re-time the traffic.

45.Paper: Periodic GATE Optimization (PGO): A New Service Scheme for Long-Reach Passive Optical Networks Authors: Burak Kantarci, Member, IEEE, and Hussein T. Mouftah, Fellow, IEEE Summary: In this paper, a novel scheme, Periodic GATE Optimization (PGO) for LR-PONS is presented which is based on a previously proposed service scheme where the OLT generates multiple threads to poll the bandwidth requests of the ONUs. In PGO, OLT periodically builds an ILP formulation by using the collected requests of the ONUs, and solves the model for the overloaded ONUs. Based on the outputs, it determines how to credit the ONUs until the next ILP formulation whenever they are overloaded. The performance of PGO is evaluated by simulation, and shows that it introduces further decrease in average packet delay until the heavy traffic. Average queue length at the ONUs is also shortened by PGO. The results imply that PGO achieves these enhancements without increasing the packet drop probability. A QoS-aware version of PGO, namely PGO-QoS is also presented. PGO-QoS consists of two modules such as intra-ONU scheduling where the dequeuing proportion of the QoS classes is determined and reported to OLT and DBA module where mostly PGO runs. Finally, the performance of PGO-QoS for each QoS class and for the overall network in terms of average packet delay, average queue size, and packet drop
23 / 53

probability under different scenarios is evaluated. Based on the simulation results, PGO-QoS is shown to introduce further decrease in delay, queue length, and packet loss probability for the high priority class requests that are granted by multithread polling in LR PON.

46.Paper: Quality of Service in TDM/WDM Ethernet Passive Optical Networks (EPONs) Authors: Ahmad R. Dhaini, Chadi M. Assi, and Abdallah Shami Summary: In this paper three new DBA schemes for QoS support in hybrid WDM/TDM PON networks are proposed. The first scheme showed its dependency on the adopted intra-ONU scheduler weight profile and its impact on the overall performance of the first proposed WDM-PON architecture. On the other hand, the other two schemes, which are designed to handle the bandwidth allocation in the second proposed WDM-PON architecture that allows transmission of each ONU on two channels simultaneously, demonstrate different preference in the network performance. The second scheme provides optimal high priority traffic jitter performance but with low bandwidth utilization and medium/low priority traffic increased delay. Alternatively, the third scheme provides a performance balance between these parameters.

47.Paper: Shortest Propagation Delay (SPD) First Scheduling for EPONs with Heterogeneous Propagation Delays Authors: Michael P. McGarry, Martin Reisslein, Frank Aurzada, and Michael Scheutzow Summary: In this paper, EPONs with an offline scheduling framework, which enables QoS mechanisms by collecting bandwidth requests from all ONUs before the OLT makes dynamic bandwidth allocations for transmissions on the shared ONUs-to-OLT upstream channel is examined. The Shortest Propagation Delay (SPD) first scheduling policy is proposed. It sequences the ONUs upstream transmissions in increasing order of the ONUs propagation delays, which means that the upstream transmission of the ONU with the smallest propagation delay is scheduled first. The competitiveness of SPD first scheduling is analyzed and found that it achieves very close to optimal performance. SPD scheduling through probabilistic analysis and simulations is evaluated and results show significant reductions in packet delay with SPD first scheduling in EPONs with heterogeneous propagation delays, especially when Limited grant sizing is employed. For future study, the convergence of fiber-based and wireless access networks, which will potentially cover large geographic areas and thus have highly heterogeneous propagation delays can be investigated.

48.Paper: Simultaneous and Interleaved Polling: An Upstream Protocol for WDM-PON Authors: Frederick Clarke, Suman Sarkar, and Biswanath Mukherjee Summary: In this paper, a Simultaneous and Interleaved Polling with Adaptive Cycle Time (SIPACT) for the WDM-PON architecture is proposed. In a WDM-PON, since an ONU can support multiple different channels, a scheme where the OLT could poll intra-wavelength (on the same wavelength) users in an interleaved fashion and inter-wavelength (among different wavelengths) users simultaneously is proposed. Using SIPACT, the lower and upper bounds on available bandwidth for each ONU and the
24 / 53

average packet delay are analyzed

49.Paper: SLA-Aware Dynamic Bandwidth Allocation for QoS in EPONs Authors: Su-il Choi and Jaehyung Park Summary: A new dynamic scheduling algorithm, referred to as cyclic-polling-based dynamic bandwidth allocation with service level agreements (CPBA-SLA), for service differentiation that meets the service level agreements between the OLT and ONUs is proposed in this paper. The proposed DBA scheme provides a constant and predictable average packet delay and an improved delay jitter of the expedited forwarding traffic without the influence of load variations.

50.Paper: Sleep Mode for Energy Saving PONs: Advantages and Drawbacks Authors: Shing-Wa Wong, Luca Valcarenghi, She-Hwa Yen, Divanilson R. Campelo, Shinji Yamashita, and Leonid Kazovsky Summary: This paper focuses on some feasible implementations of sleep mode in PONs. In particular, ONU sleep mode is considered. The paper first outlines the ONU wake-up process using current TDM PON protocols. Current and novel ONU architectures that selectively switch-off some elements are then described. Their advantages in terms of energy savings are analytically computed under different traffic scenarios. Using the proposed architectures, analytical results show more than 50% energy saving under realistic TDM traffic. 51.Paper: When Are Online and Offline Excess Bandwidth Distribution Useful in EPONs? Authors: Jason R. Ferguson, Michael P. McGarry, and Martin Reisslein Summary: In this paper, an Online Excess Bandwidth Distribution (OEBD) to overcome the stability problems of the existing offline excess bandwidth distribution mechanisms is proposed. It is found that OEBD generally achieves delays between offline excess bandwidth distribution and IPACT-Limited below the stability limit of offline excess bandwidth distribution and below IPACT-Limited for load levels above the stability limit. Further study can be made on OEBD. One important direction is to further comprehensively study the parameter setting for OEBD to ensure robust, good performance across a wide range of scenarios. 52.Paper: Intelligent Dynamic Bandwidth Allocation Algorithm in Upstream EPONs Authors: Nurul Asyikin Mohd Radzi, Norashidah Md. Din, Mohammed Hayder Al-Mansoori, Intan Shafinaz Mustafa, and Sajaa Kh. Sadon Summary: An upstream access scheme for EPONs, named the intelligent fuzzy-logic-based dynamic bandwidth allocation algorithm (IFLDBA) has been proposed in this paper. The algorithm provides the allocation of bandwidth between ONUs and within ONUs. Fuzzy logic is used to improve the bandwidth allocation within each ONU. the IFLDBA algorithm is compared to the existing broadcast polling algorithm. The simulation results show that IFLDBA has bandwidth utilization of up to 21% better and lower delay than the broadcast polling algorithm.

25 / 53

53.Paper: Burst-Polling-Based Dynamic Bandwidth Allocation Using Adaptive Minimum Guaranteed Bandwidth for EPONs Authors: Wansu Lim, Changho Yun, Yeon-Mo Yang, Hyunju Lee, and Kiseon Kim Summary: In this paper an enhanced burst-polling dynamic bandwidth allocation (EBDBA) method is proposed to support broadband access networks based on QoS for EPONs. This algorithm dynamically increases or decreases the minimum guaranteed bandwidth of the three traffic classesexpedited forwarding (EF), assured forwarding (AF), and best effort (BE) trafficaccording to the requested bandwidth of the ONUs. As a result, packet delays decreases and throughput increases. Simulation experiments also confirm EBDBAs superiority over conventional algorithms.

54.Paper: Dynamic Bandwidth Allocation With Adaptive Polling Cycle for Maximized TCP Throughput in 10G-EPON Authors: Hiroki Ikeda, Member, IEEE, and Kenichi Kitayama, Fellow, IEEE Summary: In this paper a DBA algorithm with an adaptive polling cycle (DBA-APC) to maximize throughput in an EPON. It works as follows: it observes the queue length for TCP packets in the ONUs for the upstream in every polling cycle. It determines the TCP throughput requirement for upstream based on the observed queue information. It determines the optimum polling cycle time and the required upstream bandwidths based on the throughput limit due to the TCP RTT and the bandwidth utilization. It allocates the upstream bandwidths based on the required bandwidth for the entire ONUs by using a typical DBA algorithm. It optimizes the polling cycle time for the allocated bandwidth if the allocated bandwidth is smaller than the required bandwidth. Then it sets the polling cycle time. The simulation results show that DBA-APC effectively utilizes the bandwidth in 10G-EPON to transmit TCP traffic.

55.Paper: Optical Coding for Enhanced Real-Time Dynamic Bandwidth Allocation in Passive Optical Networks Authors: Kerim Fouli, Tomaz Berisa, and Martin Maier, Senior Member, IEEE Summary: In this paper, optical pulse encoding and decoding technology is proposed to enable real-time signaling in a PON. Unique optical codes are assigned to selected ONUs equipped with the corresponding encoders. An out-of-band pulse train is broadcast from the OLT and is modulated by ONU-based switches. The encoded reflections of pulses are thus used to update the status of the Optical Coding-enabled queues at the OLT in real time. Through a performance comparison of native IPACT with its OC-enhanced counterpart, it is showed that the OC enhancement breaks the fundamental delay lower bound associated to the polling cycle. A new DBA algorithm is proposed that exploits real-time queue updates enabled through OC-enhanced polling. The proposed signaling scheme constitutes a push technology and offers the advantages of off-line scheduling in an online setting using viable low-speed OC technology, while coming at the expense of incremental optical hardware and reserved control bandwidth. Moreover, the pay-as-you-grow implementation of OC-enhanced polling is analyzed to realize QoS differentiation, elaborate on possible migration paths from conventional PONs. Future investigation must address jitter analysis, exhaustive real-time DBA algorithm design enabling absolute QoS guarantees and multiple traffic class support. Another avenue of possible research is OC architecture scalability; for
26 / 53

PONs with high splitting ratios or LR PONs,

56.Paper: A Novel Decentralized Ethernet-Based PON Access Architecture for Provisioning Differentiated QoS Authors: Sherif R. Sherif, Antonis Hadjiantonis, Georgios Ellinas, Member, IEEE, Chadi Assi, and Mohamed A. Ali Summary: This paper proposes a novel Ethernet over star coupler-based PON architecture that uses a fully distributed TDMA arbitration scheme. Supported by the decentralized scheme, this paper proposes several QoS-based DBA algorithms in which the OLT is excluded from the implementation of the time slot assignment. In contrast with the mainstream centralized EPON architectures that combine priority queuing with DBA schemes the proposed EPON architecture supports differentiated services through the integration of both scheduling mechanisms at the ONU (intra-ONU scheduling). It has been shown that the overall performance of the proposed decentralized EPON architecture and the associated bandwidth allocation algorithms are shown to be at least as efficient as their centralized counterparts.

57.Paper: Resource Management in STARGATE-Based Ethernet Passive Optical Networks (SG-EPONs) Authors: Lehan Meng, Chadi M. Assi, Martin Maier, and Ahmad R. Dhaini Summary: In this paper, an optical network unit architectures is proposed which allow STARGATE EPONs (SG-EPONs) to evolve in a pay-as-you-grow manner while providing backward compatibility with legacy infrastructure and protecting previous investment. Then, a corresponding DBA algorithm for effective resource management in these networks is proposed and its performance (delay, throughput) is investigated through simulation experiments.

58.Paper: QoS Control Schemes for Two-Stage Ethernet Passive Optical Access Networks Authors: Abdallah Shami, Member, IEEE, Xiaofeng Bai, Student Member, IEEE, Nasir Ghani, Senior Member, IEEE, Chadi M. Assi, Member, IEEE, and Hussein T. Mouftah, Fellow, IEEE Summary: This paper proposes a two-stage EPON (penetrated-EPON) architecture that allows more end-users to share an optical line terminal link, and enables longer access distances (more than 25 km distance). In addition, a new DBA algorithm is proposed which support differentiated services in a network with heterogeneous traffic. Detailed simulation experiments are conducted to study the performance and validate the effectiveness of the proposed architecture and algorithm.

59.Paper: Jitter Performance in Ethernet Passive Optical Networks Authors: Abdallah Shami, Member, IEEE, Xiaofeng Bai, Student Member, IEEE, Chadi M. Assi, Member, IEEE, and Nasir Ghani, Senior Member, IEEE Summary: This paper proposes a novel dynamic scheduling algorithm, called hybrid granting protocol

27 / 53

(HGP), to support QoS in EPON. Specifically, the proposed dynamic scheduling algorithm minimizes packet delay and jitter for delay and delay-variation sensitive traffic by allocating bandwidth in a grant-before-report (GBR) fashion. In more detail, the GATE message may include grant information for the expected EF traffic that will arrive before the next transmission start time of its parent ONU.This improves their performance without degrading QoS guarantees for other service types. the proposed algorithm is tested in terms of average packet delay, delay variation, and buffering space utilization.

60.Paper: Per-Stream QoS and Admission Control in Ethernet Passive Optical Networks (EPONs) Authors: Ahmad R. Dhaini, Student Member, IEEE, Chadi M. Assi, Member, IEEE, Martin Maier, Member, IEEE, and Abdallah Shami, Member, IEEE Summary: In this paper, the first framework for per-stream QoS protection in EPONs using a two-stage admission control (AC) system is proposed. The first stage enables the ONU to perform flow admission locally according to the bandwidth availability, and the second stage allows for global AC at the OLT. DBA algorithms which perform both bandwidth allocation and reservation simultaneously are presented as well. The AC system has shown a good performance in terms of maintaining the QoS level for already existing traffic while providing an overall acceptable minimal throughput for BE traffic even under heavy network load.

61.Paper: On the Co-Existence of 10G-EPONs and WDM-PONs: A Scheduling and Bandwidth Allocation Approach Authors: Mohammad S. Kiaei, Chadi Assi, Senior Member, IEEE, Lehan Meng, and Martin Maier, Member, IEEE Summary: in this paper the problem of optimal scheduling and bandwidth allocation in next generation 10G-EPON coexisting with 1G WDM-PONs is discussed. a network architecture for providing this coexistence and derive an ILP model for offline joint scheduling and bandwidth assignment for 10G-TDM and 1G-WDM ONUs is then proposed. To address the scalability of the ILP model, a Tabu Search based heuristic for efficiently obtaining near optimal solutions is introduced. Of course there is a tradeoff which exists in terms of delay, scheduling length, and channel utilization, when separate or the same DBA modules are used for 1G- and 10G-ONUs. upgrading WDM 1G-ONUs to TDM 10G-ONUs can improve the QoS experienced by end users, yet it would decrease the channel utilization on the existing 1G channels.

62.Paper: Toward Quality of Service Protection in Ethernet Passive Optical Networks: Challenges and Solutions Authors: Chadi Assi, Martin Maier, and Abdallah Shami Summary: This paper presents an admission control framework that is capable of supporting and protecting the QoS of real-time traffic while guaranteeing a minimum bandwidth for best effort traffic, and presents an appropriate DBA algorithm. Numerical results are presented to show the effectiveness of the proposed solution and compare the performance with solutions that rely only on traffic scheduling.

28 / 53

4.1 Classification of Papers

Category Long Reach PON (LR-PON) DWA (Dynamic Wavelength Allocation) QoS (Quality of Service)

Paper index # 10, 16, 17,18, 21, 24, 34, 35, 37, 39, 42, 43, 45 8, 30, 35 3, 6, 8, 9, 11, 12, 13, 19, 32, 38, 40, 42, 45, 46, 49, 53, 55, 56, 58, 59, 60, 62 1, 2, 3, 4, 5, 6, 7, 9, 11, 12, 13, 16, 18, 19, 22, 23, 24, 25, 27, 28, 29, 33, 35, 36, 37, 38, 43, 47, 49, 51, 52, 53, 56, 57, 58, 59, 60, 61, 62 43, 51 5, 15, 17, 25, 26, 30, 36, 46, 48, 61 30, 36, 46, 50, 56 30, 36, 41, 46 14, 20, 29, 31, 32, 40, 44, 54 14, 54, 61 22, 33, 34, 50
Table 4.1

Ethernet PON (PON)

On line algorithm Wavelength Division Multiplexing (WDM) Time Division Multiplexing (TDM) Hybrid PONs GPON 10G PON Energy efficient

29 / 53

5. Conclusion
From the papers studied, it is obvious that research on dynamic bandwidth allocation has mainly focused on Ethernet PONs and Quality of Service. An extensive study on DBA has also been made for Long Reach PONs, Wavelength Division Multiplexing and Gigabit PONs. On the other hand, not many papers cover topics of DBA such as Dynamic Wavelength Allocation, On line DBA algorithms, Time Division Multiplexing, Hybrid (TDM-WDM) PONs and Energy Efficiency in Next Generation PONs. And as cost becomes more and more significant in todays networks, energy-efficiency in PONs is a topic that needs to be studied deeper and further.

30 / 53

6. OPNET

6.1 What is OPNET and how it is used


OPNET Modeler is one of the main application products for the simulation and modeling of telecommunications protocols and Web technologies. However , OPNET's huge amounts of source code and assisting application programming interfaces (API) can be very difficult, even for skilled programmers.

Fig.6.1: OPNET Modeler

The OPNET Modeler structure contains three modeling domains: the process, the node, and the network. In the process modeling domain the programmer utilizes the behavior of numerous processes, like email clients, TCP managers, or IP interfaces. In OPNET, the modular implementations of these processes are known as process models. The entire standards of an OPNET process model is made up of finite state machine, action statements expressed in C/C++, and configurable variables. Within the node modeling domain the programmer utilizes the behavior of numerous networking appliances, such as clients, servers, switches, or routers. Node models are generally described by means of a number of functional components known as modules and by the information flow between them. The behavior of specific modules is described either by using a set of integrated variables or via a number of process models. Within the network domain the programmer utilizes complete network models including individual nodes and interconnecting communication links. A network model standard also contains the configuration of this kind of simulation model attributes as individual applications, user profiles, and network protocols. The setup of individual nodes and transmission mediums, their connectivity and geographical spots, helps to further specify a network model. The characteristic values described in the network modeling domain

31 / 53

pass on all the way down to the process models. The attribute values define either local features, applicable to individual systems, or global features, applicable to several appliances in the network. The OPNET modeling architecture is organized in a layered style with the process domain being the lowest layer, accompanied by the node domain, and lastly the network domain layer at the very top. The process models are widely used directly to build node models, which in turn are put together to construct various network models.

6.2 Closer look into OPNET and how it works


Let us now take a closer look into OPNET and its basic operations. OPNET 's hierarchy as said before is 3-tiered. It consists of Network, Node and Process model. Node model specifies objects in a network domain. Process model specifies object in node domain.

Fig.6.2.1: OPNETs hierarchy

The Project Editor specifies the network topology and configures nodes and links. One can Choose
results, run simulations and view results.

The Node Editor create models of nodes by specifying internal structure and capabilities The Process Editor Develops models of decision-making processes representing protocols,
algorithms, resource management, operating systems, etc.

The Link Model Editor Creates, edits and views link models. The Packet Format Editor specifies packet format, defining the order, data type and size of fields
contained within the packet.

Fig.6.2.2: Packet editor

32 / 53

The Simulation Tool Defines additional simulation constraints and sequences. The Analysis Tool Creates graphs and applies statistical data.
Process Domain OPNET process models consist of: State transition diagrams / Finite State Machines Blocks of C code State/Temporary variables A process is defined as an instance of a process model and can create additional child processes dynamically or can respond to interrupts They are classified as Forced(Green) & Unforced States(Red) differing significantly in execution timing.

Fig.6.2.3: Process editor

In detail Forced State (Green): Invokes enter exec, Invokes exit exec, Evaluates all conditions, If exactly 1 condition evaluates to true, the transition is traversed to next state.

33 / 53

Fig.6.2.4: Forced state

On the other hand, Unforced State (Red) Invokes enter exec, Places a marker at the middle of the state, Releases control to the simulation kernel and becomes idle, Resumes at the marker and processes the exit execs when next invoked.

Fig.6.2.5: Unforced state

34 / 53

OPNET process model OPNET allows you to attach fragments of C code to each part of a state. This code augmented by OPNET-specific functions, is called Proto-C. The three primary places to use Proto-C are: Enter Executives: This Code is executed when the module moves into a state. Exit Executives: This Code is executed when the module leaves a state. Transition Executives: This Code is executed in response to a given event. Node Models Node models consist mainly of the following:

Processor Modules Queue Modules Transmitter Modules Receiver Modules Packet Streams.

Node Domain In the node domain basic building blocks include processors, queues and transceivers/receivers. Processors are fully programmable via their process model. Queues also buffer and manage data packets Transceivers/Receivers are a node's outbound/inbound interfaces, via which the interconnection between blocks is performed. Packet streams that carry data packets from a source to a destination module. Model Components A Queue model mainly requires:

Fig.6.2.6: Queue model

A means of generating packets (source module) Queuing packets (queue module) Servicing packets (queue module) Destroying packets (sink module)

Network Domain Network models consist of nodes, links and subnets deployed in a geographical context. Nodes represent network devices and groups of devices such as servers, workstations, routers, etc. Links represent point-to-point and bus links.
35 / 53

OPNET also has many vendor support network models from Cisco, 3Com, etc. Event Driven Simulation In OPNET Simulation time advances when an event occurs.

Fig.6.2.7: Event sequence

36 / 53

7. 10G EPON simulation


In our simulation, we represent an 10-Gigabit Ethernet Passive Optical Network with 32 Optical Network Units (ONUs) and one Optical Line Termination (OLT). In this scheme, the ONUs share a 10G upstream link and transmit packets using Time Division Multiplexing. The OLT runs a bandwidth allocation algorithm and grants each ONU: - with a weighted amount of time to transmit according to the queue load of each ONU or - with a fixed amount of time (fair-queuing algorithm). Media Access Control protocol is used for exchanging control packets (REPORT and GATE) between the ONUs and the OLT.

Fig.7: 10G-EPON

7.1 ONU
The node model ONU is consisted of one transmitter, one receiver, one simple source and one queue.

37 / 53

Fig.7.1.1: ONU node model

Inside the queue, we have implemented a process model that performs the following: - As soon as a packet arrives it is checked for being either a data packet or a GATE packet. - If it is a data data packet, it is enqueued and further processed. - If it is a GATE packet, it is de-capsulated to obtain the values of start and end_timeslot values which will be used later to calculate the transmission window of the ONU. - After successful en-queuing of the packet, it is transmitted to the transmitter of the ONU and after that to the receiver of the OLT. The transmission window lasts for end_timeslot-start_timeslot*timeslot_quantum seconds, where start_timeslot = The start timeslot at which ONU begins to transmit. This value is assigned to the ONU by the GATE message sent by the OLT. end_timeslot = The end timeslot at which ONU stops transmitting. This value is assigned to the ONU by the GATE message sent by the OLT. timeslot_quantum = The time in seconds of a timeslot's duration.

38 / 53

Fig.7.1.2: Process model

By the end of the transmission window, the ONU schedules to send a REPORT packet to the OLT with the ONU_ID and number of bytes currently allocated in the queue. This number will be later used by the OLT to calculate the transmission window of the specific ONU in the next scheduling cycle.

Fig.7.1.3: REPORT packet

7.2 OLT
The node model OLT is consisted of 32 receivers, 32 transmitters, one queue and one sink.

39 / 53

Fig.7.2.1: OLT node model

Inside the queue, we have implemented a process model that performs the following: - As soon as a packet arrives, it is checked for being either a data packet or a REPORT packet. - If it is a data packet, it is sinked. - If it is a REPORT packet, it is de-capsulated to obtain the onu_id and and number of bytes waiting to be served in the queue of the respective ONU. This number will be later used to run an algorithm and create GATE packets that will grant start and end_timeslot to the ONU.

40 / 53

Fig.7.2.2: Process model

- When the OLT collects all 32 REPORT packets, it assigns timeslots to ONUs via GATE packets according to the following algorithms: bytes_requested/sum_of_bytes*total_timeslots (weighted algorithm, each ONU is assigned with a weighted number of timeslots) or total_timeslots/MAX_STREAMS (fair-queuing algorithm, each ONU is assigned timeslots) where, bytes_requested: The number of bytes each ONU has requested through the REPORT packet sum_of_bytes: The total number of bytes requested by all ONUs total_timeslots: The duration of a cycle divided by the timeslot duration (in seconds). MAX_STREAMS: The maximum number of upstream links (32 in our case). with a fixed number of

41 / 53

Fig.7.2.3: GATE packet

7.3 Bandwidth allocation scheme simulation


The OLT is responsible for all bandwidth allocations in the network. An ONU is allowed to transmit only in time slots that have been assigned to it by the OLT. The OLT has great amount of flexibility in implementing a bandwidth allocation algorithm; allocations may change dynamically over time to adapt to the different bandwidth requirements of the ONUs. The bandwidth allocation algorithm may implement anything from a strict and even division of the bandwidth among all ONUs (static allocation), to giving all bandwidth to a single ONU. This type of allocation is very inefficient, since the end-stations may not require the entire grant bandwidth all the time, and the wasted bandwidth might have been allocated to someone else. For this reason more advanced allocation algorithms have been implemented, that are dynamic in nature. These algorithms usually have access to input about the current, near real-time needs of the end stations. Dynamic algorithms may work either in a continuous timeline or as cycle based. A continuous timeline algorithm receives bandwidth requests from the end stations and allocates bandwidth according to them in a continuous mode. A cycle based algorithm divides the timeline into consecutive cycles, and calculates the bandwidth allocation for an entire cycle at a time. The grants information is sent in GATE messages, which tell each ONU its start time to transmit and the length of the allocated transmission. In order to make correct allocation decisions, the ONUs send REPORT messages, informing the OLT of their queue status. Each allocation algorithm might use this information in a different way. In this section we describe the 10GPON environment that was implemented and its components. Since there was no available OPNET module for the 10GPON standard, we constructed a 10GPON modeling environment from scratch. Network Model Assumptions In our bandwidth allocation scheme we consider two algorithms: Dynamic bandwidth allocation: Each ONU is assigned a proportional number of timeslots according to the REPORT packet the later has sent to OLT.

Fig.7.3.1: Dynamic allocation

42 / 53

As seen earlier, it is computed as follows:

Static bandwidth allocation (fair-queuing): This division implements fairness in the grant allocation. However, other divisions are also possible, as long as the sum of the static allocations does not exceed the cycles size. There is no consideration of the actual needs of each ONU, and each receives the same allocation in every cycle. Below figure depicts an example of the algorithm's allocation for 3 ONUs. Each ONU is assigned a fixed number of timeslots equal to:

Fig.7.3.2: Static allocation

This algorithm does not take into consideration the REPORT packets sent by the ONUs, it rather assigns timeslots on the spot no matter if the ONU has data to transmit or not. We consider the following numeric parameters and assumptions in the simulation:

Data rate of all links are 10 Gbps. Queue capacity of ONUs is 8 Mbits. Packets size varies between 64 and 1518 bytes. Propagation delay between ONU and OLT is 100 s. Cycle duration is 0.002 sec. Timeslot duration is 1 s. Constant number of ONUs (no ONU joins or leaves the network). Each ONU is assumed to serve a single end-user (subscriber). No packet loss has been introduced at least at this first scheme.
The interarrival time of each packet in each scenario is calculated using the following formula:

7.4 Simulation output and conclusions

43 / 53

The simulation was ran 5 times for 10 secs each for both bandwidth allocation algorithms. We consider a specific traffic load each time (0.1, 0.3, 0.5, 0.7, 0.9). We come up with the following output: For dynamic bandwidth allocation: Experiment # 1 2 3 4 5 Throughput mean 1001467776 3024281300 5002439629 7002654144 9207311265

Fig.7.4.1: Mean throughput (dynamic allocation)

Experiment # 1 2 3 4 5

End-to-end delay mean 0.000503977 0.000304072 0.000266413 0.000250414 0.000241451

44 / 53

Fig.7.4.2: Mean packet delay (dynamic allocation)

For static bandwidth allocation: Experiment # 1 2 3 4 5 Throughput mean 1001488978 3024351545 5002543427 7002770638 9207444911

Fig.7.4.3: Mean throughput (static allocation)

Experiment #

End-to-end delay mean


45 / 53

1 2 3 4 5

0.000365012 0.000221294 0.000199548 0.000193651 0.000192088

Fig.7.4.4: Mean packet delay (static allocation)

We notice that mean throughput for both algorithms is approximately the same (either for relatively small or larger traffic load). The main difference lies on packet end-to-end delay where mean delay for dynamic allocation algorithm is almost twice the end-to-end delay of static allocation algorithm. The main reason for this is the fact that in dynamic allocation algorithm each ONU does not transmit in each cycle de facto. Let us examine the following example to understand this better: ONU_1 has no data in its queue, so it sends a REPORT packet to OLT requesting for 0 timeslots for the next transmission cycle at t0. At t1, a packet arrives in the queue of ONU_1 waiting to be transmitted. At t2, the REPORT packet arrives in ONU assigning no timeslots for the current transmission cycle. So, the data packet in ONU's queue will have to wait another transmission cycle in order to be transmitted. This causes larger mean end-to-end delay for dynamic algorithm. Other conclusions that can be made are the following:

For different load of network, the dynamic algorithm achieves the best division of the network
bandwidth as it adapts the allocations to each ONU's needs. For ONUs with low queue load in a highly loaded network, the dynamic algorithm shows the worst delay performance. As explained earlier above, this is because the static algorithm allocates more than the low load ONUs need, and they may use the extra allocations to send new packets faster without the need to report them. On the other hand, the rest of the ONUs in the network suffer more because there is wasted bandwidth while they still have outstanding requests. The drawbacks of the static algorithms are the following: it wastes a lot of empty grants that may be needed by other ONUs, and it prevents high utilization of the network. It limits the amount of ONUs/allocated bandwidth per ONU for the network.

46 / 53

7.5 Current condition and future work and research


It is important to notice that although ONUs are scheduled to send packets every transmission cycle duration (0.002 sec), the burstiness of GATE packets produced by the OLT allows ONUs to transmit for specific timeslot quanta (as long as the transmission window) but in non-canonical intervals. Further study has to be made in order to enhance the current algorithm so that the transmission of each ONU takes place strictly every transmission cycle. Finally, another future direction can be the implementation of more dynamic algorithms.

47 / 53

8. Conclusion
This report concentrated on the issue of upstream bandwidth allocation in PON networks (EPON, GPON, 10G-PON). We presented a brief history of PON, discussed the importance of DBA in contemporary networks and refer to Medium Access Control and its importance in bandwidth allocation. We evaluated and reviewed a number of articles and publications which propose new bandwidth allocation algorithms and categorized them based on the aspects covered. Accordingly, we used OPNET, a powerful simulation tool, to design model of a 10GPON network. The model consists of one module for the network's central-office, the Optical Line Terminal (OLT), and another module for the customer premises device, the Optical Network Unit (ONU). The network model implements 10GPONs Time Division Multiple Access (TDMA) multiplexing scheme and two bandwidth algorithms are developed and analyzed. Finally, results were made relating to average packet end to end delay and improvements for potential study were proposed.

48 / 53

Appendix: Source Code


In this side note, we take a closer look at the source code of the simulation. We demonstrate the process models and techniques used to simulate the 10GPON. ONU process model

49 / 53

50 / 53

OLT process model

51 / 53

52 / 53

References

[1] [2] [3] [4] [5] [6]

[7] [8] [9] [10]

Techno-economic study of highsplitting ratio PONs and comparison with conventional FTTH-PONs/FTTH-P2P/ FTTB and FTTC deployments.- Sotiria Chatzi, Ioannis Tomkos A 10-Gb/s 1024-Way-Split 100-km Long-Reach Optical-Access Network- Darren P. Shea, John E. Mitchell A bandwidth assignment polling algorithm to enhance the efficiency in QoS long-reach EPONs- N. Merayo, T. Jimenez, P. Fernandez, R.J. Duran, R.M. Lorenzo, I. de Miguel and E.J. Abril A Highly Flexible and Efficient Passive Optical Network Employing Dynamic Wavelength Allocation- Yu-Li Hsueh, Matthew S. Rogge, Shu Yamamoto and Leonid G. Kazovsky A survey of dynamic bandwidth allocation algorithms for Ethernet Passive Optical Networks- Jun Zheng, Hussein T. Mouftah Energy Management Mechanism for Ethernet Passive Optical Networks (EPONs)- Ying Yan, Shing-Wa Wong, Luca Valcarenghi, She-Hwa Yen, Divanilson R. Campelo, Shinji Yamashita, Leonid Kazovsky, and Lars Dittmann Energy-Efficient Long-Reach Passive Optical Network: A Network Planning Approach Based on User Behaviors- Lei Shi, Sang-Soo Lee, Huan Song, and Biswanath Mukherjee ETHERNET PASSIVE OPTICAL NETWORK ARCHITECTURES AND DYNAMIC BANDWIDTH ALLOCATION ALGORITHMS- MICHAEL P. MCGARRY, MARTIN REISSLEIN, MARTIN MAIER Bandwidth Distribution Solutions for Performance Enhancement in Long-Reach Passive Optical Networks- Burak Kantarci, T. Mouftah QoS Based Bandwidth Allocation Schemes for Passive Optical Networks- Nadav Aharony, Eyal Nave, Nir Naaman and Shay Auster

53 / 53

You might also like