A Sort-based DBA Algorithm for EPON System

(1)

A Sort-based DBA Algorithm for EPON System

Wen-Ping Chen, Shih-Wei Kau, Bin-Yih Liao

+

, Wen-Shyang Hwang, and Wen-Tsung Hu

Department of Electrical Engineering, Department of Electronic Engineering

+

National Kaohsiung University of Applied Science

[email protected], [email protected], [email protected], [email protected], [email protected]

Abstract

EPON technology is being considered as a promising solution for next-generation broadband access network due to the convergence of low-cost equipments. To prevent data collision and ensure efficient transmission, a MAC protocol to share resource of upstream transmission medium is necessary. This article proposes a sort-based DBA algorithm that sorts all REPORT messages by length to fully utilize the idle time as long as at least one ONU requests a long enough transmission window; or some requests be lain out together in idle period to compensate the idle time when no request length is long enough. Besides, a queue management scheme also be developed, which improves efficiency further by decreasing the overhead of REPORT messages when the ONU queue is long. Event-driven simulations show that sort-based DBA can improve network performance in terms of packet delay, average queue length and throughput, as compared with the well-known IPACT and DBA2 algorithms.

Keywords: EPON, Polling, DBA, and Sort

1. Introduction

Passive optical networks (PONs) have received some attention from both industry and academia as a possible cost-effective solution. A great deal of effort has gone into developing and standardizing various PON technologies. Asynchronous transfer mode (ATM) PON [1] and gigabit PON [2], for example, have been standardized by the International Telecommunication Union (ITU-T). ATM, however, has a significant shortcoming: a dropped or corrupted cell will invalidate the entire IP datagram. When this happens, any remaining cells carrying portions of the invalid datagram will continue to propagate, consuming network resources unnecessarily. Furthermore, ATM imposes a cell tax on variable-length IP packets.

EPONs represent the convergence of low-cost Ethernet equipment (e.g., switches and network interface cards) and optical fiber architecture. Considering that more than ninety percent of today’s data traffic originates from and terminates

in Ethernet LANs, EPONs appear to be a natural candidate for future last-mile solutions.

Another possible solution is to use wavelength division multiplexing (WDM) PON [7]. This approach allows each optical network unit (ONU) to operate at a different wavelength, thus avoiding any mutual interference. This technology is simple to implement and provides enormous bandwidth, but requires either a tunable receiver or a receiver array at the optical line terminal (OLT) to receive data over multiple channels. Unfortunately, these devices are currently rather costly.

An EPON is a point-to-multipoint (P2MP) network

consisting of one OLT and multiple ONUS. In the downstream

direction the OLT broadcasts to all ONUS simultaneously,

while in the upstream direction a single optical fiber channel is

shared by all ONUS. To avoid data collision, a scheduling

algorithm is needed to prevent simultaneous transmissions. The well-known media access control (MAC) protocol CSMA/CD is standard for Ethernet LANs [3], but doesn’t represent a good choice for EPONs. Because the OLT will

receive all data packets transmitted by the ONUS and discard

those involved in collisions, each ONU would require an additional receiver operating at the upstream wavelength and a carrier sensing circuit. Time division multiplexing (TDM) technology is a popular alternative for EPONs. Each ONU is assigned a timeslot for data transmission in each cycle, and can only transmit data in the allocated window. It can be either static or dynamic, depending on the arbitration mechanism implemented by the OLT. Kramer and Mukherjee [4] have studied the performance of EPON under TDM using a fixed bandwidth assignment algorithm. Although this scheme is easy to implement and performs well at heavy load conditions, it cannot handle statistical multiplexing between ONUS. Static

schemes such as this are also very inefficient due to the bursty nature of access network traffic.

To cope with this problem, Kramer [5] proposed the polling-based scheme called interleaved polling with adaptive cycle time (IPACT). This algorithm achieves good performance by combining limited service with a maximum transmission window defined over 2 ms polling cycles. However, the idle time issue isn’t effectively resolved, and the fact that IPACT allocates bandwidth based on a single ONU REPORT which is not globally optimized. Due to the bursty

Eighth International Conference on Intelligent Systems Design and Applications

(2)

nature of Ethernet traffic and encapsulation of Ethernet packets (That is, packet fragmentation is prohibited), however,

some ONUS may have less traffic to transmit and other ONUS

may have more traffic to transmit and need larger bandwidth in each transmission cycle.

To address the issue, Luo et al. [9] proposed a DBA scheme called limited sharing with traffic prediction (LSTP) by predicting the traffic arrived during the waiting time and maintaining a portion of bandwidth for delivery. However, the prediction scheme has the behavior of bursty traffic, but some bandwidth may be wasted because it could not exactly

estimate the real demand of traffic load for all ONUS at the

next transmission cycle. Assi et al. [6] proposed a DBA algorithm, which make use of the excessive bandwidth of

lightly loaded ONUS to allocate the bandwidth demand of

heavily loaded ONUS in each transmission cycle and thus

improve the performance of the limited allocation scheme. Besides, for addressing the idle time issue, the authors also proposed an early allocation mechanism, called DBA2, which schedules a lightly loaded ONU instantaneously without any

delay, whereas schedules those heavily loaded ONUS after the

OLT receives all REPORT messages and perform computation for bandwidth allocation. However, the idle period is only improved under low or medium traffic load by

DBA2 algorithm. Moreover, all ONUS may have a bandwidth

demand larger than the minimum guaranteed bandwidth at high traffic load, the GATE message cannot be transmitted early to ONU for compensating the idle time.

Therefore, this paper proposes a new DBA algorithm called sort-based DBA algorithm which the transmission order of all grant data are allowed by the REPORT length at next cycle. It can completely eliminate the idle time between cycles as long as at least one ONU has a sufficiently long data transmission.

Even when most of the ONUS are operating at medium load,

only if some guard-band distance are wasted, the new algorithm still can achieve good outcome to eliminate idle time. Besides, this paper also proposed a queue management which reduces the unnecessary overhead of REPORT messages when ONU is operating heavy traffic load. It can obtain more bandwidth utilization at heavy traffic load for EPON system. As this paper will demonstrate, sort-based DBA has best bandwidth utilization, low delay, and low queue length than IPACT, and DBA2 on an EPON system.

The remainder of the article is organized as follows. Section 2 defines basic EPON architecture. The new sort-based DBA algorithm supporting EPON is detailed in Section 3. Section 4 describes the results of a computer simulation comparing sort-based DBA to IPACT, and DBA2. Section 5 concludes with a few remarks.

2. System Architecture

Ethernet PON is a candidate architecture for “last mile” access networks. In this section, we describe the network and its nodes to give a global view of the proposed system. There are several multipoint topologies suitable for an EPON: bus, ring, tree, and tree-and-branch. The most popular choice is based on a tree topology; it consists of one optical line

terminal (OLT), a 1:N passive star splitter (and combiner), and

multiple optical network units (ONUS). The number of ONUS

(N) is typically between 4 and 64, but networks with N=128 have also been fabricated. The OLT resides in a central office (CO) that connects the access network to a metropolitan area network (MAN) or wide area network (WAN). The OLT is connected to the passive star splitter by a single optical fiber. The passive splitter is generally located far from the CO, but close to the subscriber premises. An ONU may be located at the curb or building, wireless, or even on the subscriber premises, and is connected to the passive splitter by a short, dedicated optical fiber. The distances between the OLT and each ONU typically range between 10 and 20 km. All data are encapsulated in Ethernet packets for transmission, which are presumed to be compatible with the IEEE 802.3 standard. The fragmentation of Ethernet packet into to the transmission window is allowed. All transmissions occur between the OLT and the ONUS.

In the downstream direction, the OLT connects all of

ONUS such as a point-to-multipoint (P2MP) architecture. It

broadcasts Ethernet frames to all of ONUS simultaneously

through the 1:N splitter on a single wavelength (for example, 1550 nm). This behavior is similar to that of a shared media network. In the downstream direction, Ethernet standards fit the EPON architecture perfectly: packets broadcast by the OLT are given a media access control (MAC) address, so that they will be extracted only at the intended destination (that is, an ONU). In the upstream direction, an EPON is a

multipoint-to-point (MP2P) network. All ONUS transmit their data to the

OLT on a common wavelength (for example, 1310 nm)

through the 1:N passive combiner. Since ONUS share the

upstream transmission medium, an EPON must efficiently allocate uplink access and avoid data collisions. A MAC-based mechanism is generally chosen for this purpose.

3. Sort-based DBA Algorithms

In a global polling based DBA algorithm, each ONU normally sends a REPORT message to the OLT after transmitting its grant data. The purpose of these messages is to request bandwidth for the next cycle time. Once OLT has received all REPORT messages, it calculates the appropriate bandwidth allocation and broadcasts GATE messages to all

ONUS while the upstream bandwidth is idle between the time

of the last REPORT message is received and the time of the

GATE message is sent to all of ONUS. Hence the idle time is

given by

ONU DBA

idle T RTT T

T (1)

Where RTT is the round-trip time from ONU to OLT, TDBA

is the calculation time of DBA algorithm, and TONU is the

processing time of the ONU (on receiving a GATA message). In terms of bandwidth utilization, the idle time defined above is undoubtedly wasted. To address this problem, we propose an adaptive DBA algorithm by sorted all REPORT messages with request length to given a transmission order. This algorithm can be divided into two cases to eliminate the idle

(3)

also able to serve needy ONUS earlier than IPACT or DBA2

algorithms.

Fig. 7 and Fig. 8 show the average queue length and grant length as a function of offer load respectively. Again, the sort-based DBA algorithm has the best property. Its queue length is the shortest for traffic loads between 55 Mb/s to 62 Mb/s, where the offered traffic pushes the limits of the network. At light and medium loads, we also find that sort-based DBA gives the shortest average grant bursts. At heavy loads, our

algorithm increases Lmin so that more packets can be served

into the upstream fiber. This increases the average grant lengths considerably relative to IPACT and DBA2. Furthermore, since IPACT defines a maximum transmission window of 15 Kbytes per ONU in a cycle time of 2 ms, its average transmission length of grant data at heavy load is only 14.5 Kbytes in our traffic model. Since fragmenting Ethernet frames is not allowed, limited IPACT produces the same result. The DBA2 algorithm always allocates remaining idle slots to

other ONUS when the transmission demand is excessive, so its

average transmission length can reach 15 Kbytes.

The above discussion makes it clear that the sort-based DBA algorithm has the best performance, with significantly more throughput per ONU than IPACT, and DBA2 under the same conditions. The proposed DBA algorithm also exhibits excellent bandwidth utilization and low delay, making it suitable for EPON access networks.

0 2,000 4,000 6,000 8,000 10,000 55 56 57 58 59 60 61 62 63 64 65 Offer Load (Mb/s) A v er age Q u eu e L en g th ( K B ) IPACT(Fixed) IPACT(Limit) DBA2 Sort-DBA

Figure 7. Average queue length per ONU for IPACT (fixed and limited), DBA2, and sort-based DBA algorithms.

0 5 10 15 20 25 45 50 55 60 65 Offe r Load (Mb/s) A v e r a g e G r a n t L e n g th (K B ) IPACT(Fixed) IPACT(Limit) DBA2 Sort-DBA

Figure 8. Average grant length per ONU for IPACT (fixed and limited), DBA2, and sort-DBA algorithms.

5. Conclusions

A sort-based DBA algorithm is proposed that sorts all REPORT messages with the request length to eliminate the idle period for obtaining high bandwidth utilization on Ethernet passive optical networks (EPONs). It provides collision-free upstream transmission, and achieves higher utilization under most load conditions than other DBA algorithms. In order to fully make use of the idle time between transmission cycles, it calculates the maximum transmission

window for each ONU demand and adjusts the order of ONUS

so that the longest transmission comes last. At medium and light loads, when no ONU transmission is long enough to use up the idle time on its own, a compensation scheme is developed that one or many grant data are lain out together after last grant data to stuff the idle space. By the simulation results shown that the bandwidth utilization can be obviously advanced to eliminate the idle period. Furthermore, the algorithm proposed can achieve more maximum performance by using a queue management scheme to reduce the overhead of REPORT messages when the ONU queues are long. IPACT algorithms (with fixed service and limited service) and DBA2 were compared to our DBA algorithm in an event-driven EPON network simulation. The sort-based DBA algorithm achieved significantly better performance than the other DBA algorithms.

Acknowledgment

The authors would like to thank National Science Council (NSC) in Taiwan R.O.C. for supporting this research; the project number is NSC 96-2221-E-115-017.

References

[1] J.D Angelopoulos, G.C. Boukis, I.S. Venieris, “Delay priorities enhance utilisation of ATM PON access systems”, Computer Communications, vol. 20, no.11, Oct. 1997 pp. 937-949.

[2]J.D. Angelopoulos, H.C. Leligou, T. Argyriou, S. Zontos, E. Ringoot and T. Van Caenegem, “Efficient Transport of Packets with QoS in an FSAN-Aligned GPON”, IEEE Communications Magazine, vol. 42, no. 2, Feb. 2004, pp.92–98.

[3] C. J. Chae, Wong, E. and Tucker, R. S., “Optical CSMA/CD Media Access Scheme for Ethernet over Passive Optical Network”, IEEE Photonic Technolgy Letters, vol. 14, no. 5, 2002, pp. 711–132. [4] G. Kramer and G. Pesavento, “Ethernet Passive Optical Network (EPON): Building a Next-Generation Optical Access Network”, IEEE Commun.Magazi., vol. 40, no. 2, Feb.2002, pp. 66–73. [5] G. Kramer, B. Mukherjee and G. Pesavento, “IPACT: A dynamic protocol for an Ethernet PON (EPON)”, IEEE Communications Maganize, vol. 4, no. 1, Feb. 2002, pp. 74–80.

[6] C. Assi, Y. Ye, S. Dixit and M. A. Ali, “Dynamic bandwidth allocation for quality of service over Ethernet PONs”, IEEE Journal of Selected Areas in Communications, vol. 21, no. 9, Nov. 2003, pp.1467–1477.

[7] G. Maier, M. Martinell, A. Pattavina, and E. Salvadori, “Design and Cost Performance of the Multistage WDM-PON Access Networks”, IEEE Lightwave Technology, vol. 18, no. 2, Feb. 2000, pp. 125–143.

[8] http://www.nlanr.net/NA/Learn/packetsizes.html.

[9] Y. Luo and N. Ansari, “Limited sharing with traffic prediction for dynamic bandwidth allocation and QoS provisioning over EPONs”, Journal of Optical Networking, vol. 4, no. 9, Sep. 2005, pp.561–572.