An Improved Polling Mechanism for Dynamic Bandwidth Allocation of XGPON Networks

Abstract

One of the most well-known technologies that offers a large range of services at user’s fingertips without requiring the least amount of infrastructure is passive optical networks. This can be due to a particular component known as the passive optical spiller. In order to facilitate time division access for individual ONUs on a single optical fiber, the appropriate dynamic bandwidth allocation mechanism must be present at the OLT with the advent of user support. For bandwidth assignment by DBA, it is essential to determine the precious upstream bandwidth demand, and can be aided by the more frequent polling mechanism. The more frequent polling mechanism lead to erroneous bandwidth report problem, become more serious when the mechanism is employed in larger coverage area network (specifically larger channel delay). This IBR problem lead increase in upstream delay at higher load condition, it’s because of improper assignment of bandwidth and waste of bandwidth. The IBR issue is resolved by the improved polling and scheduling method presented in this paper. Instead of transmitting the bandwidth allotment in each downstream frame, the OLT uses this scheduling technique to send it all at once to a TCONT (i) of ONU (i) for the duration of the SI period. To counteract the channel delay impact, a unique polling technique is also presented that use the hybrid methodology in which the OLTs sends polling mechanism once per the Service interval but According to the simulation study’s findings, the suggested system reduces mean upstream delay for type 2, 3, and 4 traffic classes by 62% when compared to the EBU scheme and by up to 20% when compared to the XGIANT design.

Introduction

I. INTRODUCTION

Excessive internet traffic demand is increasing in recent years has prompted more research on advanced networks that can handle the high bandwidth demands of enterprises, public sector organizations, single individuals, and the mobile industry. A technology called passive spillter/combiner [1] of passive optical networks has proven a promise and demonstrated its significance in offering high throughput, scalability, and energy efficiency in the direction of a highly reliable bandwidth infrastructure. Upstream speeds of 2.5 Gbps and downstream speeds of 10 Gbps are supported by XGPON [2]. However, because of the symmetrical broadcast access mechanism, the line-rate can approach Gbits for individual users, but because of the time division based bandwidth-sharing mechanism employed in upstream, it cannot reach Gbits for individual users. The point-to-multipoint architecture of XGPON network is depicted in the figure 1, along with the various components that represent a network system. For example, An Optical Line Terminal(OLT) serves as the central controller, the Optical Line Unit(OLU) acts as the router at user ends, the passive spillter distributes signals to each ONU in a passive manner, and the upstream unit functions as a combiner, combining all of the signals from ONUs into a single optical signal that can be received and processed by a single OLT.

A four types of transmission container types (TCONTs) with provisioned allocation policies and quality of services have been recommended by the XGPON standard [2]. Each ONU with different traffic class is assigned a unique Allocation Id, which might be useful in classifying each of them separately. Each of the TCONTs has a corresponding type of bandwidth associated with it, such as fixed bandwidth for TCONT- T1, ensured bandwidth for TCONT-T2, non-assured and best effort for TCONT-T3 and TCONT-T4. To support with various traffic container, the various with their supported application with their operational service parameter is given in Table I. In XGPON, the synchronous frame has been standardized, with value of 125 µsec for both the downstream and the upstream. The OLT obtained the queue status report from the ONUs via Dynamic Bandwidth Report Upstream (DBRu) field in upstream direction. Following to save a scheduling operation for Granting of bandwidth is carried out and delivered back to ONUs via field BWmap to avoid the collision during the data transmission in upstream direction. This process is carried out at a instance called the service interval (e.g. SImin, SImax). The duration of this service interval (SI) is integer multiple of frame duration.

The XGPON compliant DBA has two types of polling mechanisms: one that schedules grant for each alloc − ID as well as, polls them once during a service interval (SI); the other that schedules grant for each alloc − ID during every downstream cycles as well as polls them in every DS cycle. The first method having a problem improper reporting of buffer occupancy and lead to improper assignment of bandwidth to individuals. This cause the higher uplink delay and larger frame loss specifically the lower priority traffic class. To combat larger delay and frame loss of first method, the other method is suggested. the performance of lower priority class is improved at low load condition but at higher load situation the bandwidth assignment is become more severe and inefficient, lead to higher delay and larger frame loss for the all the traffic class. The second method having a problem of lack of bandwidth, it’s because of more frequent assignment of polling field called DBRu. Nevertheless, no specific polling method is suggested for DBA, that may be used for small or large Chanel delay e.g. 100µSec, 600µSec for 20km, 60km coverage respectively. More attention has been paid to appropriate and effective polling methods for the XGPON system.

TABLE I

Various Traffic class with their Operational Parameters

In this work, a Modified EBU for XGPON is present that improve the scheduling and polling mechanism of existing state of art DBA scheme called Efficient bandwidth utilization (EBU). EBU grant phase suffers from Borrowed-Refund problem and polling phase suffer from the static report inconsistency, and increased delay for all traffic class at high loads and bandwidth waste at lower priority traffic loads. The BR problem result in reduced bandwidth assignment to T3 and T4 class and T4 is severely affected its delay and frame loss increase. Modified EBU resolve the both issues and improve the performance of all the traffic class. However, the packet loss is slightly increasing because of solving the problem of BR. The flow of rest of this paper is as: section II is covered related works and existing works is reviewed. The section ?? covers the polling methodology used in proposed method and modified scheduling methodology have been explained in section III-B. The results are presented in section IV and the conclusion is presented in last section.

II. RELATED WORK

In 2006 [3], the first standard compliant DBA for GPON network was introduced. The MAC operation is divided in two phases that this DBA is working on are the Guaranteed Part of Allocation (GPA) and the Surplus Bandwidth Part of Allocation. The SPAs’ allocation for TCONT-2 is done as guaranteed bandwidth allocation; TCONT-3 allocates a portion of non-assured bandwidth; and TCONT-4 allocates one word in order for DBRu to be function for buffer occupancy of ONUs. In SPA, remaining portion non-assured bandwidth is carried for TCONT-3, and in in the end, for TCONT-4’s remaining non-guaranteed bandwidth.

The aforementioned type of scheduling order is implemented using the AB provisioned counter, as Table I illustrates for each traffic class. With an operational value of 5 to 10, the other parameter known as the SI is introduced in this DBA named Timers/Down Counter and maintained for each alloc−ID of each traffic class maintained. The polling and scheduling operation is carried out once the SI value reaches to 1, where the SI is recharged to its initial value irrespective to the request size. The granting mechanism follows the gated policy as; min (Request, AB, FB), where FB is called the Total Frame Bytes of upstream burst. The major problem of this DBA is ideal channel time, it because of polling is scheduling is carried out once the SI expiry; lead to higher upstream delay.

The Immediate Allocation with Colorless Grant (IACG) algorithm [4] provides a solution to the ideal time delay problem. The available byte counter (V B) was added for every traffic class’s alloc − Id. It stores the bytes allotted for each DS frame cycle and decremented them according to the bytes granted for each frame action as; min (Request, V B, FB). This V B stores the AB once the SI expire. And same is carried out once in every SI. Finally, by adding extra TCONT-5, the unused bandwidth in each grant process is allotted to each ONU without any polling operation. However, some bandwidth that is not being used during the grant operation is still available which may not be used by other types of alloc − ID.

Later in 2013, the EBU [5] had been proposed with Borrow Refund policy to allocate such unused bandwidth of individual traffic class. For borrowing, the possibilities of negative value for V B is suggested, means the requested bandwidth is higher than available byte counter value. In an update operation, the common counter counterfeits the return of such a borrowed byte. A novel polling methodology is introduced in it, where the DBRu slot is assigned to each alloc − ID irrespective their queue size and additional DBRu can be allocated whose grant size is higher than zero. Still the problem is persisting in EBU that if there is no bandwidth allocation in common counter than it leads to reduce the bandwidth allocation in least priority traffic class. One accessible byte counter for each traffic class was used to address the deficiency of EBU in a Simple and Feasible DBA (SFDBA) [6]. One of the byte counters may exhaust the common counter, preventing other service classes from using the services of other classes. For the purpose of resolving such a difficulty, the DBA known as the High Utilization algorithm (DBAHU) [7] was proposed as an updated version of SFDBA using a concept of BR of EBU. The grant policy is carried out based on the min (report, A, FB), where A is called common maximum allocation byte of same traffic class.

In subsequent year 2015, X-GIANT algorithm was introduced by author [8]. A problem of zero unused bandwidth, which may have affected the next cycle allocation is solved by allocating the bandwidth irrespective of expiry of Down counter. Author proposed smaller SI value to reduce the ideal channel timing of XGPON, e.g. SImin = 2, and SImax = 1. The priority order among the TCONT class is maintained via proposing a GIR:PIR ratio to 0.4, which reflect in their upstream delay result. The polling is occurred in every frame duration for TCONT-2, TCONT-3 and TCONT-4, while twice frame duration for TCONT- 3 and TCONT-4. The performance of TCONT-3 is much better up to certain load due to excess grant allocation to TCONT-3.

The main challenge is as the polling carried out each frame duration for TCONT- 2, it leads to overhead allocation for T2 traffic class. The latter the enhance polling and scheduling mechanism was introduced in DBA called Enhanced Bandwidth utilization (IBU) [9].

The priority to TCONT-2 and TCONT-3 in assignment the delay of least container TCONT-4 is increased. The Comprehensive Bandwidth Utilization (CBU) [10] is useful in improving the ICAG and EBU scheduling policy and polling policy. This improve the delay for TCONT-4 but fails to TCONT-3 and TCONT-2 traffic as compared to that of EBU. The CBU introduced the polling frequency in every ODD frame duration e.g. 8, 5, 2. However, the clear justification was not explained for the same.

The memory mechanism was introduced in EBU to counterfeit the static report inconsistency problem due to larger coverage areas. It was suggested that it required to store the at least three granted bytes that need to subtract from the reported bandwidth for 20km operation. This may increase to eight cycle for the larger coverage operation. The latter in 2022, the more efficient bandwidth allocation algorithm was introducing namely EEBU [11]. This polling was carried out once per SI based on Polling flag value. This polling flag in reset once when the Down counter value to zero value unlike to that of EBU where the PF flag may reset to zero once the grant to the particular alloy − ID is greater than zero.

The goal of this algorithm is improvising the utilization of remaining bandwidth in the available Byte counter. The problem of Static report inconsistency still present in this algorithm. All together, the most fundamental algorithm known as XGIANT [8], GIANT [3], ICAG [4], and EBU [5] serves as the foundation for the improvement of other DBA mechanisms. This work developed a novel technique, named MEBU, to solve the issue of static report inconsistency with better polling approach, utilizing the unused bandwidth totally and without any more complexity.

III. MODIFIED EBU

A bandwidth is allocation in MEBU is illustrated in figure 2, according to service interval SI. The SI duration is divided into total 10 frame duration, C0, C1,…C9 cycles, in order to easily understand the polling mechanism. Every one of them is 125uSec in duration. Round-trip time is another crucial aspect in polling, and it is calculated using the Ranging technique [2]. The term” time to allocated bandwidth reached to specific ONUs” or” obtaining the DBRu value from the ONUs” refers to RTT/2. By equalization timing, ONUs receive the given DBRu field once it has been assigned. Equalization time plus processing time equals the total polling time. The ONU processed the Queue Occupancy report after receiving the DBRu and transmitted it to OLT. Following reception of the buffer report, OLT processes it and determines the bandwidth demand by factoring in the time required for OLT processing as well as bandwidth allocation. The grant is distributed to each ONU by the DBA using BWmap; the total amount of time required is known as the dynamic bandwidth allocation time.

As explained above, Upstream frame delay (DUS) depends upon four factor as given by (1).

T2 class mean queuing delay well with range the decreasing delay performance at higher load is due to higher priority assignment in GPA cycle as well as additional cycle. T3 have constant delay performance due to GPA, SPA and additional cycle. T4 performance is also decreasing at higher load condition, this is unusual case. To refine this additional refinement has been suggested in M-XGIANT.

V. ACKNOWLEDGMENT

I would like to express my deepest gratitude to my supervisor for his valuable suggestion and guidance. Without his immense support, this work would not be possible to happen on time.

Conclusion

The performance XGPON network is improvement in terms of supporting higher traffic load. The Mean queuing delay improvised as compared to that of RR, EBU and XGIANT. For T2 the performance improvement is 3% to 5% for T2 Class, 5% to 7% for T3 class and 2% to 4% for T4 class. However, with fixed assignment scheme the performance of MEBU is well bounded event at higher traffic load. It is best suits for the bounded delay application like 5G communication, the bounded delay specification for Best effort traffic is within 10mS. So this novel algorithm is best suits for the application like FTTh, FTTb and backhaul communication networks.

References

[1] Gigabit-Capable Passive Optical Networks (G-PON), Rec. G.984.x, ITU Std., October 2008. [2] 10-Gigabit-Capable Passive Optical Networks (GPON) Series of Recommendations, G.987.x, ITU Std., March 2010. [Online]. [3] Available: http://www.itu.int/rec/T-REC-G/e [4] H. C. Leligoun, C. Linardakis, K. Kanonakis, J. D. Angelopoulos, and T. Orphanoudakis, “Efficient medium arbitration of FSAN- compliant GPONs,” International Journal of Communication Systems, vol. 19, pp. 603—617, 2006. [5] M.-S. Han, H. Yoo, B.-Y. Yoon, B. Kim, and J.-S. Koh, “Efficient dynamic bandwidth allocation for fsan-compliant gpon,” J. Opt. [6] Netw., vol. 7, no. 8, pp. 783–795, Aug 2008. [Online]. Available: http://jon.osa.org/abstract.cfm?URI=jon-7-8-783 [7] M. Soo Han, H. Yoo, and D. Soo Lee, “Development of efficient dynamic bandwidth allocation algorithm for xgpon,” ETRI Journal, vol. 35, 02 2013. [8] M. Han, “Simple and feasible dynamic bandwidth and polling allocation for xgpon,” in 16th International Conference on Advanced Communication Technology, Feb 2014, pp. 298–304. [9] M. Han, “Dynamic bandwidth allocation with high utilization for XG-PON,” in 16th International Conference on Advanced Communication Technology, Feb 2014, pp. 994–997. [10] J. A. Arokkim, K. Brown, and C. J. Sreenam, “Refining the giant dynamic bandwidth allocation mechanism for xg-pon,” in IEEE ICC 2015 SAC- Access Networks and Systems, Nov 2015, pp. 1006–1011. [11] R. Butt, S. Idrus, K. Qureshi, N. Zul, and S. H. Mohammad, “Improved dynamic bandwidth allocation algorithm for xgpon,” Journal of Optical Communications and Networking, vol. 9, 01 2017. [12] R. Butt, S. Idrus, S. Rehman, P. M. Shah, and N. Zul, “Comprehensive polling and scheduling mechanism for long reach gigabit passive optical network,” Journal of Optical Communications, 01 2017. [13] T. Thangappan and B. Therese, “Implementation of extra efficient bandwidth utilization dynamic bandwidth allocation algorithm to support differentiated service classes in xgpon,” Wireless Personal Communication, 05 2021.

Copyright

Copyright © 2023 Patel M. C., Mandloi A. S.. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.