Академический Документы
Профессиональный Документы
Культура Документы
a r t i c l e in f o
abstract
Article history:
Received 26 March 2014
Received in revised form
5 September 2014
Accepted 6 September 2014
Available online 30 October 2014
Passive optical network (PON) provides enough bandwidth for combining multiple service
providers' in addition to the fiber to the home (FTTH) network. Multiple optical line
terminals (multi-OLT) PON-based hybrid network is an efficient access network that
provides an effective solution for reducing the computational complexity of data packet
processing of multiple service providers having different packet lengths and data rates.
However, in the multi-OLT PON-based hybrid networks, data transmission in the single
upstream channel requires a complicated dynamic bandwidth allocation (DBA) algorithm
for improving the bandwidth sharing efficiency among the multiple service providers. In
this paper, we propose a new network architecture for a PON-based open access network
(OAN) where every optical network unit (ONU) of the PON will be shared by all the service
providers and in the central office multiple OLTs will be connected to handle the data
packets of multiple service providers. In the upstream direction, we propose to use
multiple wavelengths for modulating the data packets of multiple service providers and to
avoid the requirement of a complicated DBA algorithm. The simulation results show that
the multi-OLT and multi-wavelength PON-based OAN (MM-OAN) provides higher bandwidth utilization, higher upstream efficiency and higher throughput with a lower
overhead to data ratio and jitter than the single-OLT and single wavelength PON-based
OAN (SS-OAN). The overall throughput of the proposed scheme has been increased more
than by a factor equal to the number of wavelengths employed than that of the SS-OAN.
& 2014 Elsevier B.V. All rights reserved.
Keywords:
Open access network
PON
FTTH
Multi-OLT PON
Hybrid networks
1. Introduction
An open access network (OAN) is a hybrid network
where multiple service providers can use the network
simultaneously. The OAN is open to any service provider,
local or otherwise, and can use it to offer different services
to the subscribers. The service providers are able to offer
any number of services and the subscribers are also able to
choose any service provider for each of the services those
http://dx.doi.org/10.1016/j.osn.2014.09.001
1573-4277/& 2014 Elsevier B.V. All rights reserved.
they choose to subscribe to [1,2]. Multiple service providers are expected to be deployed in the future modern
cities in addition to the fiber to the home (FTTH) network,
e.g., wireless sensor networks (WSNs), high-definition
television (HDTV) or video on demand (VoD), and Femto
networks (FNs). Usually, in any hybrid network, the central
office (CO) provides independent connection with the
different networks or service providers [3]. The conventional OAN also provides independent connections
between the CO and different service providers through
a common access network and any service provider can
connect to that access network using a single access
terminal. The passive optical network (PON) can be used
149
150
polling table. As all the OLTs use same polling table for RTT
information, in the proposed scheme, it is called shared
polling table. The shared polling table is also used to provide
synchronization among the multiple OLTs [6]. The proposed
MM-OAN architecture is considered to be deployed in the
city area, i.e. short range (SR) PON, to connect several service
providers. However, a longer feeder optical fiber between the
splitter and array waveguide grating (AWG) can be installed
to implement the architecture for a long range (LR) PON.
In the upstream direction of the proposed MM-OAN,
the data packets of a particular service provider will be
accepted by the designated OLT through a passive splitter
and an AWG. The upstream channel in a time cycle will be
divided by the number of active ONUs N, i.e., Tslot Tcycle/N.
Here, Tslot is the time-slot for each ONU in a time cycle and
Tcycle is the length of a time cycle. A Tslot of an ONU will be
occupied by overlapping the data packets from the different service providers by using the different uplink wavelengths. As shown in the figure, the upstream data packets
of the FTTH terminals connected to the ONUs 1N are
modulated by the wavelength u1 and transmitted in the
time slots 1, 2, 3, N. In contrast, the upstream data
packets of the CHs of a WSN connected to the ONUs 1N
are modulated by the wavelength u2 and transmitted in
the time slots 1, 2, 3, N, respectively. Here, the length of
the feeder cable between the AWG and the splitter will be
limited to few tens of kilometers as the network will be
deployed in the cities.
In the downstream direction of the proposed MM-OAN,
each OLT will broadcast data to the network through a
passive splitter in a time division multiple access (TDMA)
manner using a single downstream wavelength d. Here,
the broadcasted data of each OLT contains the downstream
data packets and control messages to a particular service
provider of the entire ONUs. As shown in the Fig. 1, the
OLT1 broadcasts the downstream data packets and control
messages to the FTTH terminal of all the ONUs while the
OLT2 broadcasts the downstream data packets and control
messages to the CHs of a WSN of all the ONUs.
The multi point control protocol (MPCP) provides timing reference to synchronize all the ONUs in the network.
In the MPCP, timing synchronization among ONUs is
151
152
153
i;j
max _2 for the lightly loaded ONUs
max
is the maximum granted window for the
where grant_
i;j
ONU i at the time cycle j.
However, if the requested window sizes are less than
the maximum transmission windows then the granted
window size grant
for the ONU i at the time cycle j is
i;j
calculated as below
8
R_2
< R_1
if R_1
i;j
i;j 4i;j
grant
i;j
2
R_2
: R_2
if R_1
i;j
i;j oi:j
T cycle;j grant
x;j
Fig. 5. Internal buffer architecture of an ONU for the upstream transmission
in an MM-OAN.
x1
154
BWU
UE
grant
Nm
k 1 k
Nm
grant NT C
k1 k
grant
Nm
k 1 k
;
max
m
Nk 1 P k B
NT C
NT C
Nm
grant
k1 k
Table 1
Simulation parameters.
Symbol Quantity
NOLT
N
m
d
u
Tproc
Tcycle
TGI
R
BE
BR
Bmax _1
Bmax _2
P
Number of OLTs
Number of ONUs
Number of service providers
Number of downstream wavelengths
Number of upstream wavelengths
Data processing time per service provider
Cycle time
Guard time
Transmission speed
Length of Ethernet overhead
Length of Report message
Maximum packet length for service providers
in the group one
Maximum packet length for service providers
in the group two
Number of generated packets and each packet
Value
4
16
4
1
m
10ms
0.53.0 ms
5 ms
1 Gbps
576 bits
304 bits
1500 bytes
1024 bytes
010
length is Bmax _1 for service providers in group packets
one and Bmax _2 for service providers in group
two
can be calculated by
s
n
1
1 2
T jcycle
;
Jitter
T j
n j 1 cycle
155
multiple upstream packets by using the multiple wavelengths for the multiple service providers connected to an
ONU. Fig. 8(a) and (b) compare the throughput vs offered
load in the upstream direction between the SS-OAN and
MM-OAN for 2-ms and 4-ms cycle times, respectively.
Here, the MM-OAN uses only 1 Gbps bandwidth while
both the 4 Gbps and 1 Gbps bandwidths are used for the
SS-OAN. From this figure we can see that the difference of
throughput between the MM-OAN and SS-OAN increases
gradually from the lowest offered load to the highest
offered load. At the highest offered load of 1.0 the MMOAN provides more than 400% higher throughput than the
SS-OAN for 1 Gbps bandwidth in both the schemes. In
contrast, the MM-OAN provides more than 25% higher
throughput than the SS-OAN when the MM-OAN uses
1 Gbps and SS-OAN uses 4 Gbps bandwidth. This comparison proves than the MM-OAN provides higher throughput
even though 4 times larger bandwidth is used in the
SS-OAN for 4 different service providers.
From the analysis of the contour plots in Fig. 7(a) and (b)
and comparison of throughput for a 2 ms and 4 ms cycle
times in Fig. 8(a) and (b) we found that the proposed MMOAN provides almost more than 400% and 25% higher
throughput than the SS-OAN for 1 Gbps and 4 Gbps bandwidth, respectively, at the highest offered load of 1.0. The
main reason of this improvement in the proposed MM-OAN
than the SS-OAN is the use of the multiple upstream
wavelengths and simultaneous transmissions of the
Fig.7. Throughput in the upstream direction for the SS-OAN and MM-OAN.
Fig. 8. Comparison of throughput in the upstream direction. (a) 2 ms cycle time and (b) 4 ms cycle time.
156
Fig. 9. Comparison of performances between the SS-OAN and MM-OAN, (a) Bandwidth utilization, (b) upstream efficiency, (c) overhead to data ratio, and
(d) Jitter in ms.
157
the Ethernet overhead but the MM-OAN provides significantly less ODR and jitter in both the directions of
maximal cycle times and offered loads than the SS-OAN
as shown in the contour plots of Fig. 9(c) and (d). The
reduction of the ODR and jitter are achieved due to the
new bandwidth allocation principle that uses two different maximum transmission windows for improving the
bandwidth sharing efficiency and reducing the bandwidth
wastage.
This paper considers a PON-based OAN that will be
deployed in the city area. Usually, a SR PON will be
enough to cover the whole city area. However, in this
paper, the jitter performances of the proposed MM-OAN
have been analyzed for both the SR PON and LR PON to
prove that the proposed scheme is capable to provide
better performance than the SS-OAN in both the SR and
LR PONs. Typically, the jitter is increased with the
distances between the OLTs and ONUs. Fig. 10(a) and (b)
shows the comparison of jitter between the MM-OAN and
SS-OAN for both the SR and LR PONs and for 2 ms and
4 ms cycle times, respectively. Here, the results show that
if the cycle time is increased then the jitters are decreased
for both the MM-OAN and SS-OAN because in the larger
cycle time the accumulated traffics suffer from lower
congestion. From the results it is also clear that the
proposed scheme provides far lower jitter, i.e., about
400% lower jitter at an offered load of 1.0, than the SSOAN for both the SR and LR PONs. Finally, we can
conclude that the proposed scheme utilizes the light load
effect and reduces the bandwidth wastage and provides
lower jitter with higher throughput in both the SR and LR
PONs. Therefore, the proposed scheme can be used as
either the city area network or the LR PON.
From the analysis of all the performances it is clear that
the proposed MM-OAN provides superior performances
than the SS-OAN. However, to prove the effectiveness for
implementation of the proposed scheme a tradeoff analysis between the performance enhancement and cost
comparison is important. Table 2 shows a relative cost
comparison between the proposed MM-OAN and SS-OAN.
Here, 4 service providers, and 4 OLTs are considered for the
MM-OAN while a single OLT and 4 four service providers
are considered for the SS-OAN.
From the above relative cost comparison table we can
say that no additional cost is required to implement the
proposed scheme instead the proposed MM-OAN provides
far better performances than the SS-OAN.
Fig. 10. Comparison of Jitter in ms, here, SR short range PON and LRlong range PON. (a) 2 ms cycle time and (b) 4 ms cycle time.
158
Table 2
Relative cost comparison between the MM-OAN and SS-OAN.
Quantity
SS-OAN
MM-OAN
One
4 16
4 16
No addition power is required
Four
16
16
Small amout of additional power is required
for 3 more OLT cards
No additional power is required
7. Conclusion
In this paper, the performances of the proposed network architecture of the PON-based Multi-OLT and multiwavelength OAN, using the modified version of the LS
scheme have been numerically analyzed in terms of
throughput, BWU, UE, ODR, and jitter. All the analyses
were done for a range of non-uniform traffic loads and
variable maximal cycle times. The proposed MM-OAN
commendably utilizes and shares the huge bandwidth of
the optical network for the multiple service providers in a
network and reduces the computational complexity of
data packet processing. Finally, the performances of the
proposed MM-OAN are compared to those of the SS-OAN.
The MM-OAN provides tremendous improvements in
several performance parameters than the SS-OAN because
it effectively utilizes both the multi-OLT and multiwavelength effect on the upstream channel. Specially,
throughput in the MM-OAN is almost 400% higher than
that of the SS-OAN for the four service providers connected to an ONU at an offered load of 1.0 ms and a 2 ms
cycle time. This difference will be increased proportionally
with the number of service providers connected to an
ONU. It will not cause the lower power budget due to
increasing the number of service providers in an ONU
because all the service providers will use different wavelengths in the upstream transmission. However, the maximum number of service providers connected to an ONU
will be limited by the cost of that ONU. A tradeoff analysis
between the cost and performance enhancement with the
number of service providers connected to an ONU is
required to decide the optimum number of service providers connected to an ONU.
References
[1] M. Forzati, C. Larsen, C. Mattsson, Open access networks, the
Swedish experience, in: Proceedings of the 12th International
Conference on Transparent Optical Networks (ICTON), June 27July
1, 2010, pp. 14.
[2] M. Matson, R. Michell, Study on local open access networks for
communities and municipalities, ICT Regulation Toolkit, March
2006. Available online: http://www.ictregulationtoolkit.org/en/
toolkit/docs/Document/3368.
[3] Available online: http://www.ehow.com/about_6495481_hybridnetworks_.html\.
[4] L. Kazovsky, W. Shaw, D. Gutierrez, N. Cheng, S. Wong, Nextgeneration optical access networks, J. Lightwave Technol. 25 (11)
(2007) 34283442.
[5] M. McGarry, M. Reisslein, M. Maier, Ethernet passive optical network architectures and dynamic bandwidth allocation algorithms,,
Proc. IEEE Commun. Surv. Tutor. 10 (3) (2008) 4660. (3rd Quarter).
159