Content uploaded by Glen Kramer
Author content
All content in this area was uploaded by Glen Kramer
Content may be subject to copyright.
What is Next for Ethernet PON?
Glen Kramer
(Teknovus, Inc., 1351 Redwood Way, Petaluma, CA 94954, USA.
Tel: +707-665-0400, Fax: +707-665-0491,
Email: glen.kramer@teknovus.com)
Abstract
Standardized in 2004, EPON has emerged as a highly
successful technology – 3 million lines has been
deployed in less than 2 years. The next generation of
EPON will bring 10 Gb/s bandwidth to access
networks.
1 Brief history of EPON
In 2003, the Ethernet protocol celebrated its 30th
birthday. All these years it has been adapting and
evolving to become the very inexpensive and
ubiquitous networking protocol that we know it today.
In January 2001, IEEE formed a study group called
Ethernet in the First Mile (EFM). This group was
chartered with extending existing Ethernet technology
into the subscriber access area, focusing on both
residential and business access networks. Keeping
with Ethernet tradition, the group set a goal of
providing a significant increase in performance while
minimizing equipment, operational, and maintenance
costs. Ethernet Passive Optical Networks (EPONs)
became one of focus areas of EFM.
Ethernet PON is a PON-based network that carries
data traffic encapsulated in Ethernet frames as defined
by the IEEE 802.3 standard. Where possible, EPON
utilizes the existing 802.3 specification, including
usage of the existing 802.3 full-duplex Media Access
Control (MAC).
2 Various PON architectures
There exist several standards for PONs:
x BPON – Broadband Passive Optical Network
specification was developed by FSAN and
standardized by the ITU-T (recommendation
G.983) over the period 1998-2003. BPON
uses ATM as a bearer protocol.
x GPON – Gigabit-Capable Passive Optical
Network was standardized by ITU-T G.984 in
2003-2004. It is based on unique derivative of
the Generic Framing Procedure (G.7041)
x EPON – Ethernet Passive Optical Network was
developed by the IEEE and standardized in
June 2004. EPON uses Ethernet and
Multi-Point Control Protocol.
Both BPON and GPON architectures were conceived
by the FSAN group, which is driven by major
incumbent telecommunications operators. Most of
the operators are heavily invested in providing legacy
TDM services. Accordingly, both BPON and GPON
are optimized for TDM traffic and rely on framing
structures with very strict timing and synchronization
requirements.
In BPON, an upstream frame consists of 53 timeslots,
were each timeslot comprised of one ATM cell and 3
bytes of overhead. When two consecutive timeslots
are given to different ONUs, these 3 bytes, or
approximately 154 ns, of the overhead should be
sufficient to shut down the laser in the first ONU, turn
on the laser in the second ONU, and perform gain
adjustment and clock synchronization at the OLT.
- 49 -
MoB2-1
The 5th International Conference on Optical Internet (COIN 2006)
Hyatt Regency Jeju, Korea / July 9 - 13, 2006
89-955301-4-6 98560 2006 OSIA
Similarly, very tight timing is specified for GPON. For
example, in GPON with 1.244 Gbps line rate, only 16
bit times (less than 13 ns) are allocated for the laser-on
and laser-off times. Such short intervals require
more expensive, higher-speed laser drivers at the
ONU. Similarly, a very tight bound of 44 bit times
(less than 36 ns) is allotted for the gain control and
clock recovery. In many cases, the dynamic range of
the signal arriving from different ONUs will require a
longer AGC time than the allotted overhead (guard
interval). To reduce the range of necessary gain
adjustment, BPON and GPON perform a
power-leveling operation in which the OLT instructs
individual ONUs to adjust their transmitting power so
that the level of signals received by the OLT from
different ONUs is approximately equal.
In the IEEE 802.3ah task force, a subject of selecting
EPON PHY timing parameters, such as laser turn-on
and turn-off times, and gain control time required a
technical discussion lasting almost a year. The task
force has considered several alternatives for
burst-mode timing specification, including a proposal
to use very short laser-on, laser-off, AGC, and CDR
time intervals, similar to GPON spec.
After extensive analysis the task force has decided to
take an approach different from GPON specification
and has settled on relaxed timing parameters, arguing
that this would lead to higher component yields, and
therefore would lower the costs. The IEEE 802.3ah
standard specifies the following parameters: laser-on
time = 512 ns, laser-off time = 512 ns, and a gain
adjustment time d 400 ns (negotiable). The
reasoning was that the ONUs, being mass-deployment
devices, must be as simple and inexpensive as
possible. For this, the PMD components should have
high yield and should not mandate implementation of
digital interfaces, which otherwise would be
mandatory if ONUs were required to negotiate laser
on/off times. The OLT device can be more
expensive as only a single device is used per EPON
network. Therefore, the OLT is allowed to negotiate
and adjust its receiver parameters such as the
automatic gain control (AGC) time.
Time has shown that the relaxed physical specification
of EPON to be one of the most important and
insightful decisions made by the EFM task force.
There are many suppliers for EPON optics, the
performance and yield are increasing while the cost
decreasing. At the same time, suppliers of GPON
transceivers are struggling with the more demanding
optical requirements on the ITU-T specification [1, 2].
3 Myths and Facts about EPON
In the literature, there have been many claims about
EPON’s shortcomings compared to alternatives,
especially GPON. Most of these claims are simply a
result of misunderstanding of IEEE 802.3ah standard
specification. In reality, IEEE 802.3ah provides a
quite flexible EPON specification that allows and
expects future extensibility and improvements.
3.1 Maximum Distance and Split Ratio
There have been some incorrect claims that the EPON
specification has a maximum distance and a split ratio
that are inadequate deployments in access networks.
The first step in understanding the EPON specification
is to look at the objectives that the EFM task force
established for self- guidance. Related to EPON, these
objectives were:
Provide a family of physical layer specifications:
x PHY for PON, >= 10km, 1000Mbps, single SM
fiber, >= 1:16
x PHY for PON, >= 20km, 1000Mbps, single SM
fiber, >= 1:16
In simpler terms, this means that the task force’s
objective was to produce a specification for the
physical layer of a PON, supporting distances of at
least 10 km, using single strand of a single-mode fiber
for bidirectional communications, and supporting at
least a 1:16 split ratio. The second item specifies a
separate physical-layer specification that supports at
- 50 -
least a 20 km distance, all other parameters being the
same. In the IEEE Std. 802.3, these physical layers are
referred to as 1000BASE-PX10 and
1000BASE-PX20.
It is important to understand that the IEEE 802.3ah
specification only defines a minimum boundary that a
PON device should achieve to be considered
standards-compliant. A performance exceeding the
minimal level is acceptable. In fact, today most
EPONs are deployed with a 1:32 split ratio, with some
trials done for 1:64 or even 1:128 split ratios.
The physical layer specification in EPON does not
limit the maximum distance or the maximum split ratio.
The physical layer performance is dependent on the
state of the art for optical transceivers and is
improving as this technology matures. No physical
phenomenon limits the split or distance in EPON more
than it limits them in GPON.
In addition to physical layer specifications, PON
systems require arbitration protocols to provide the
necessary connectivity and resource allocation in a
point-to-multipoint environment. These arbitration
protocols may and sometimes do limit the split ratio
and maximum distance of the PON. Table 1 shows a
few such logical parameters: maximum logical reach
(distance to the farthest ONU), maximum logical
range (distance differential between the farthest and
the closest ONUs), and maximum logical split ratio.
It can be seen that the GPON protocol specification
imposes stricter constraints on the distance and the
split, compared to that of EPON.
Table 1: Protocol constraints in GPON and EPON
GPON EPON
Max Logical Reach (km) 60 unlimited
Max Logical Range (km) 20 unlimited
Max Logical Split 128 32767
3.2 Security
Although, PONs are vulnerable to eavesdropping and
theft-of-service attacks, proposals to include security
mechanisms in objectives of 802.3ah task force did
not find the necessary support. Instead, security
mechanism has just been standardized by IEEE
802.1ae task group.
The Ethernet security specification mechanism has
just been standardized by IEEE 802.1ae task group.
Because, the 802.1ae specification was not completed
by the time the EPON standard was approved, most
EPON deployments in the world today use proprietary
security solutions. In several cases, large
telecommunications operators issued their own
security specifications. In several cases, large
telecommunications operators have issued their own
security specifications, which not only tailored to their
specific technical requirements, but also with regard to
local regulatory environment. For example, the AES
cipher suite does not have a regulatory approval for
use in China and EPON vendors are implementing a
China-specific churning mechanism developed by
China Telecom to enhance the downstream data
security.
By comparison, ITU-T recommendation G.984.3
specifies an AES-based encryption mechanism for
GPON, thus making GPON’s use in China
problematic.
4 EPON is a Successful Technology
The Ethernet in the First Mile task force completed its
charter in June 2004, culminating in ratification of
IEEE Std. 802.3ah-2004 (now merged into IEEE Std.
802.3-2005). EPON became the first optical
technology cost-effective enough to justify its
mass-deployment in an access network. Today, only
2 years after the standard ratification, more than 3
million EPON lines are deployed and the CO-installed
capacity exceeds 10 million lines.
True to its Ethernet heritage, EPON products keep
evolving; they gain functionality and performance, all
- 51 -
while lowering the cost. Since the standard
introduction, EPON system costs have decreased by
50% or more, while the cost of optical transceivers has
decreased by about 70%. Recent industry
announcements have introduced a slew of new EPON
offerings such as a quad-OLT ASIC, EPONs
supporting T1/E1 circuit emulation with jitter and
wonder well within ITU-T specs, or even an entire
ONU integrated in a GBIC module (Figure 1).
Figure 1: ONU integrated into a GBIC module
The GBIC ONU is a representative example of EPON
evolution: from a stand-alone device with a large
number of sub-components (EPON ASIC, external
SerDes, external memory, external PHY) to a
cost-reduced stand-alone device using a highly-
integrated EPON SoC, and then to a mass-produced
sub-component used in a larger system. The GBIC
module integrates a burst-mode transceiver and an
ONU ASIC. The ONU ASIC itself is a
highly-integrated system-on-a-chip device, which
includes an 802.3ah EPON MAC/MPCP, a SERDES,
a line-rate L2/L3/L4 classification engine, encryption,
forward error correction, a switch, an integrated
packet buffer, and an embedded processor. All these
functions are packed in a small LQFP package
consuming a paltry 0.6 W. This evolutionary step
allows turning any switch or router with a GBIC
interface into an ONU switch or router (Figure 2).
Figure 2: A 24-port switch attached to EPON using
GBIC ONU module.
5 The Road Ahead
What is in store for EPON? Recently, this question
was answered by the IEEE 802.3 working group
during its March 2006 Plenary session when it
approved formation of 10 Gb/s EPON study group.
The 10 Gb/s EPON call-for-interest materials identify
several major drivers for an increased capacity EPON:
emergence and growing acceptance of high-definition
television, continued development of markets with a
significant share of the population living in
multi-dwelling units, and the need to support next
generation wireless back-haul [3].
Support for HDTV and other advanced video services
is arguably the main driving force for higher-speed
EPON.
Wide adoption of 1 Gb/s EPON provided a significant
jump in access network capacity and allowed carriers
to deploy advanced digital video services. For
example, in Japan, KDDI is offering DVD-grade
multi-channel broadcasting and video on demand
(VOD), as well as high-grade IP telephony and
high-speed Internet connections (“Hikari Plus” FTTH
service).
In the Hikari Plus service, a subscriber can view 30
channels of broadcasting TV programs supported by
- 52 -
IP multicast technology and more than 2,000 VOD
programs. Karaoke-on-demand service is also
available.
Subscribers have accepted the new services, enabled
by gigabit-capable optical access networks, with great
enthusiasm, driving up the demand for yet more
bandwidth-intensive applications and services.
Many R&D labs around the world are working to
supplement the IP broadcast and VoD services
available today with such services as time-shifted
broadcast/narrowcast, all-channel personal video
recorder, picture-in-picture/split screen, digital cinema
distribution, personal multimedia publishing,
residential and business digital video surveillance, and
so on. Another example - distribution of DVD
content to in-home DVD recorders through FTTH
systems, is being field-tested by Poweredcom, Toshiba
and Tokyo Electric Power Company.
While the rich offering of new services is expected to
provide a boost to subscriber take-rates, bandwidth
consumption per subscriber is expected to grow as
well. Newer TV sets and set-top boxes, in addition
to standard definition television (SDTV) channels
(~2Mb/s) now support high-definition television
(HDTV) channels (~10Mb/s). According to a market
research report published by Technology Futures, Inc.,
45% of US households will be using HDTV by 2010,
and that number will continue to grow to
approximately 90% by 2020 [4]. The recently
approved ITU-T standard J.601 for Large Screen
Digital Imagery (LSDI) requires 40 to 160 Mb/s per
channel.
The increasing demand for advanced video services,
together with the need to support multi-dwelling units
and next-generation wireless backhaul, prompted the
IEEE 802.3 working group to initiate a study of 10
Gb/s EPON architectures.
5.1 Next-Generation EPON
The 10 Gb/s EPON effort in IEEE 802.3 will focus on
defining a new point-to-multipoint physical layer,
keeping the MAC, MAC Control and all the layers
above unchanged to the greatest extent possible.
This means that carriers can expect architectural
continuity and backward compatibility of network
management system (NMS), PON-layer operations,
administrations, and maintenance (OAM) system,
DBA and scheduling, and so on.
The 10 Gb/s EPON study group has set objectives of
specifying both symmetric line rate operation as well
as asymmetric line rate operation. The symmetric
option will operate at 10 Gb/s in both the downstream
and upstream directions. The asymmetric option will
use 10 Gb/s in the downstream and 1 Gb/s upstream,
most likely reusing the existing IEEE 802.3ah
specification for the upstream.
The asymmetric option reflects the fact that the
advanced video services create capacity pressure
mostly in the downstream direction.
The asymmetric EPON product will most likely
appear first, as this specification relies on fairly
mature technologies. The upstream transmission will
remain identical to that of the existing 1 Gb/s EPON,
and will rely on field-proven and mass deployed
burst-mode optical transceivers. The downstream
transmission, which uses continuous-mode optics, will
rely on the maturity of 10 Gb/s point-to-point devices.
The main emphasis of the symmetric option will be on
defining burst-mode operation at 10 Gb/s. The
64b/66b line coding used by the 10 Gb/s Ethernet PHY,
would likely necessitate a new FEC scheme.
6 10 Gb/s EPON Efficiency
As described in [5], EPON efficiency depends on the
values of various overhead components associated
with frame encapsulation and scheduling. These
components include line-coding overhead,
- 53 -
encapsulation overhead, control message overhead,
guard-band overhead, discovery overhead, and frame
delineation overhead. There also may be present an
FEC overhead, if FEC is implemented. The effects
of the above components were analyzed in [5] for 1
Gb/s EPON. Among the above components, only the
line-coding overhead, control message overhead, and
frame-delineation overhead would change for 10 Gb/s
EPON. We will refrain from any guesswork regarding
the expected FEC overhead, since the 10Gb/s FEC
method has not been determined.
6.1 Line-Coding Overhead
The 64b/66b line coding reduces the bit-to-baud
overhead from 20% in the current 1 Gb/s EPON to
only 3.03%.
6.2 Control Message Overhead
The control channel overhead represents bandwidth
lost due to use of in-band control messages such as
GATEs and REPORTs. The amount of overhead
depends on the number of ONUs and cycle time, i.e.,
an interval of time in which each ONU should receive
a GATE message and send a REPORT message. For
32 ONUs and a 1 ms cycle time, we can estimate a
control channel overhead of 0.215%.
6.3 Frame Delineation Overhead
According to the IEEE 802.3 standard, the
variable-sized Ethernet frames cannot be fragmented.
The non-fragmentability of Ethernet frames was the
main reason for introducing multiple queue sets in the
REPORT messages – the scheduler always selects one
of the reported queue lengths, so that the granted
timeslot is filled completely. In this case, as it is the
case in most commercially deployed EPONs, the
delineation overhead will be zero. However, even if
the scheduler completely ignores the reported frame
boundaries, the total bandwidth lost due to unused slot
remainders will only be ~1.52%.
7 Conclusion
10 Gb/s EPON is expected to be a highly-efficient
specification. As was indicated in the IEEE
call-for-interest materials[3], the 10 Gb/s EPON
provides more bandwidth capacity than CATV
network using DOCSIS 3.0. This makes 10 Gb/s
EPON a good candidate replacement architecture for
next generation CATV networks; it will allow
significant increase in data bandwidth available to
subscribers without forcing any drastic changes to the
existing video distribution model employed by cable
operators.
7 References
1. J. Redd and C. Lyon, “Challenges Mark GPON FEC
Receiver Designs,” Lightwave, PennWell, vol. 23, no. 5,
pp. 11 – 14, May 2006.
2. M. Fuller, “GPON burst-Mode Receiver Electronics
Prove Challenging,” Lightwave, PennWell, vol. 23, no.
5, pp. 11 – 17, May 2006.
3. 10Gbps PHY for EPON - Call for Interest, Presentation
at IEEE 802 Plenary meeting in Denver, CO, March 6,
2006. Available at
http://www.ieee802.org/3/cfi/0306_1/cfi_0306_1.pdf
.
4. L. K. Vanston, R. L. Hodges, and J. Savage, "Forecasts
for Higher Bandwidth Broadband Services," ISBN
1-884154-22-0, December 2004.
5. G. Kramer, Ethernet Passive Optical Networks,
McGraw-Hill Professional, ISBN: 0071445625, March
2005.
- 54 -
