You are on page 1of 4

Flexible-sense Optical Transmission

V. N. Rozental1, G. Bruno2, A. Soso2, M. Camera2 and D. A. A. Mello1


1
OCNLab, University of Brasilia
2
Ericsson Telecomunicazioni, via Calda 5, 16153 Genova

e-mail: valery@unb.br
ABSTRACT
We present a platform for the optical interconnection of traffic nodes where the sense of transmission is
exploited as an additional degree of flexibility. Implementation of the optical arrangement is shown. The
requirements about the negotiation of the available resources, such as fibers and wavelengths, are discussed. We
evaluate the implications of flexible-sense transmission in terms of cost, power and reach, under realistic
assumptions of traffic peak-to-average bandwidth ratio and degree of asymmetry.
Keywords: Flexible-sense transmission, IP traffic asymmetry, software defined parameters, peak traffic
accommodation, resource negotiation.

1. INTRODUCTION
Internet bandwidth explosion of the recent years resulted in IP traffic and storage capacity increase of 30-40%
and 40-50% per year [1]. These trends drive costly infrastructural upgrades in all network levels, namely,
creating and expanding datacenters, deploying new backbone fiber links and increasing transmission capacities.
In addition to the infrastructural investments, improvements can be made by attributing new degrees of freedom
to system parameters, to be defined and controlled by software, so that the system is able to adjust to the
changing transmission and network conditions. Solutions, such as bandwidth flexible networks and software-
defined transponders, have been presented [2]-[4].
In this scope, we propose the concept of flexible-sense optical transmission (FST), where the total capacity of a
duplex system can be used in both senses of transmission, according to the temporal traffic requirements. FST
explores the Internet traffic asymmetry and its peak-to-average characteristics to increase bandwidth in a given
direction when required. Several benefits may be achieved: primarily, FST allows accommodating high
unidirectional traffic peaks, reducing the probability of data loss and latency caused by congestion. Secondly, it
allows a more careful system dimensioning, reducing system cost due to the optimized use of bandwidth.
The topic of single-fiber bidirectional transmission has been studied in numerous works ([5]-[7]), showing that
Rayleigh backscattering (RB) produces severe crosstalk penalties when counter-propagating channels overlap in
frequency. RB crosstalk, however, can be reduced to a minimum by appropriate channel separation [5]. FST
avoids overlapped bidirectional transmission and the resultant RB-associated penalties.
In this paper we introduce the FST platform, discuss different implementation scenarios and assess system
implications. The rest of the paper is structured as follows: Section 2.1 explains the motivation behind FST;
Section 2.2 introduces a generic physical layer implementation of FST; Section 2.3 discusses several application
scenarios; Section 2.4 approaches the MAC layer resource negotiation; Section 3 discusses the relevant aspects
of FST, and Section 4 concludes the paper.

2. FLEXIBLE-SENSE ARRANGEMENT
2.1 Motivation
Fig. 1 shows the Internet exchange point traffic for the city of São Paulo at a random instant [8]. With few
exceptions, equally up and downstream dimensioned links exhibit asymmetry (i.e., different rates for each

Fig. 1. Provider Internet exchange (Pix) traffic for São Paulo.


direction), varying from roughly a factor of 2 (Americanet, Telium, etc.), up to factor of 5 and more (Locaweb,
Algar1, etc.). The links have high peak-to-average traffic rates, as well as dynamic asymmetry changes
throughout the day. Datacenter traffic, characterized by high data flow between clusters, is another important
example of asymmetry and high peak-to-average characteristics, where spurts of application demands may lead
to traffic congestion [9].
However, optical interfaces of the aforementioned systems are designed to cope with the peak traffic requests
for both transmission senses, resulting in link over-engineering. We propose to exploit IP asymmetries in high
peak-to-average rate scenarios to reduce system cost by increasing the bandwidth in a given direction when
required, using temporally unemployed resources in the opposite sense.
2.2 Optical implementation
In its simplest form, the FST arrangement connects two nodes, A (left) and B (right), as shown in Fig. 2. The
internal structure is shown only partially for node A, for simplicity. The interface consists of a duplicated
transmitter and receiver at both nodes to allow flexible sense operation. Thus, the signal can flow in either
direction in both fibers. If multiple transmitters are employed, each working at a specified wavelength, the FST
concept is replicated for all of them and a WDM mux/demux are used at both sides. Note that the Tx and Rx data
management modules, characterized by a constant operation format in standard interfaces, become software
defined, supporting different data mapping/demapping modes. Mode switch is driven by the control module, a
cross-layer interface for the resource negotiation algorithms.
Node A Node B

Tx Rx
D
M E RX
U M DATA Rx Port
X U MGMT
X
Rx Tx
A to B

CONTROL

Tx Rx
D
E M TX
M U DATA Tx Port
U X MGMT
X
Rx Tx
B to A

Fig. 2. Generic schematics for flexible-sense interface connecting two nodes, A and B. The arrows A→B and
B→A refer to the fixed-sense case for reference.
2.3 Application scenarios
FST finds its applications in datacenter interconnections, characterized by bursty and asymmetric traffic and in
local and metropolitan area networks (LAN/MAN). FST is not primarily intended for long-haul amplified
networks, because these systems would require bidirectional optical amplification, significantly increasing their
cost.
In particular, FST may be successfully employed in a framework of 100 Gigabit Ethernet [10] and beyond, to
enhance the flexibility of its optical interfaces. Applications which may benefit from FST are primarily those
over SMF:
 long range (10 km), single mode fiber (SMF) based 40GBASE-LR4 and 100GBASE-LR4;
 extended range (40 km) SMF-based 100GBASE-ER4;
 2/10/40 km 100GbE/OTU4, 10x10 Multi Source Agreement (MSA) [11].
The FST concept may be also applied to MMF applications like 40GBASE-SR4 and 100GBASE-SR10, where
the links between the nodes A and B in Fig. 2 represent MMF ribbons of 4 and 10 fibers for each direction.
However, these applications are highly cost-optimized, so their upgrade with the FST capacities is less likely.
It is important to note that flexible-sense interfaces can coexist safely with standard fixed-sense interfaces
because they employ different wavelengths. For example, passive point-to-point and ring DWDM system can be
safely upgraded to FST-enabled interfaces provided that all the installed components are bidirectional, which is
usually true for optical mux/demuxes and add/drop filters.
2.4 Resource negotiation
In FST interfaces, transmission direction changes with time for each subcarrier, to cope with variable traffic
needs between the end nodes. For this purpose, a dedicated protocol needs to negotiate the use of the available
bandwidth, avoiding “collisions,” i.e., use of the same subcarrier to transmit in both directions simultaneously.
A new in-band protocol is preferable over out-of-band protocols that involve dedicated channels, having the
advantage of not requiring additional resources, and of a direct and fast communication between the two cards at
the two end points. This protocol does not need differential delay accommodation, like required by ITU-T LCAS
[12], because all subcarriers flow across the same path; and does not need intermediate node awareness, as
required by ITU-T ODUflex hitless adjustment [13], because all the traffic remains at optical level.
Protocol activation for request or release of subcarriers can be driven by bandwidth request from the packet
layer, like the instantiation of a new multiprotocol label switching (MPLS) label-switched paths (LSPs), or it can
be forced by the application layers. It could also be automatically initiated by any end node, based on the
transmission buffer queues and the rate of receiving data from the other node, but feasibility of this last option
needs to be further investigated in terms of class of service management, interworking with TCP-IP mechanisms,
reaction times to traffic changes, etc.

3. DISCUSSION
Table 1 summarizes different aspects of FST application in 40/100 Gigabit SMF-based Ethernet and MSA
standards. In WDM links Rayleigh backscattering may become a performance limiter only when subcarrier
spacing is comparable with the signal baud rate [5]. Because 100G Ethernet interfaces employ 800 GHz spaced
WDM, there is no Rayleigh backscattering penalty due to counter-propagation [5]. The optical circulators
introduce maximum 2 x 0.5 dB loss for all implementations: the standard power budgets can be still satisfied by
proper subsystem design, e.g. at the expense of slightly more powerful transmit lasers.
Table 1. Parameters for 40/100 Gigabit SMF-based FST-capable standards.

Transmission rate FST Min. FST Max.


Num. of Subcarrier unidirectional
Standard per wavelength unidirectional
subcarriers spacing rate [Gb/s]
[Gb/s] rate [Gb/s]
40GBASE-
4 20 nm 10 10 70
LR4
100GBASE-
4 800 GHz 25 25 175
LR4/ER4
100G 10x10
10 8 nm 10 10 190
MSA

Other considerations involve size and cost of the FST, where the degree of photonic integration is a key factor:
discrete component implementation results in doubling the size and cost of the line card, making FST platform
less attractive. Fully integrated components, in turn, would yield a marginal increase in both size and cost.
When traffic from both system nodes is sufficiently low, FST flexible data mapping capabilities may be
additionally exploited for power consumption and latency reduction by controlling the degree of parallelization
of the whole subsystem, depending on the desired application: for latency reduction, the traffic load can be
distributed over all available dual-sense bandwidth, so that all transmit and receive side data processing is
implemented in parallel. On the other hand, to maximize power efficiency, the incoming traffic can be mapped
to as few carriers/fibers as possible. In this way, a subset of lasers and modulators may be turned off. Possible
power reduction, of course, depends on the particular implementation and degree of photonic integration. To
exemplify the above concepts, Fig. 3 illustrates possible transmission regimes for a 4x25G 100GBASE-
LR4/ER4, FST capable Ethernet interface. Fig. 3(a) depicts a fixed-sense operation, where the bandwidth is
equally distributed for A→B and B→A links. If the actual payload is much lower than the nominal capacity of
100 Gb/s (for example, 20 Gb/s), this figure also represents minimum latency regime, because the payload is
distributed among the subcarriers, allowing to parallelize its processing at transmit and receive nodes. Fig. 3(b)
depicts an extreme case in terms of client traffic asymmetry, where A→B traffic is much higher than 100 Gb/s
(for example, 170 Gb/s), and B→A traffic is extremely low or inexistent. Here, FST diverts almost all bandwidth
resources to A→B sense. A single subcarrier is always left for B→A sense for continuous resource negotiation.
Finally, Fig. 3(c) shows a minimum power operation regime. Here, the payload in both senses is much lower
than 100 Gb/s, allowing FST to allocate for transmission only few of the available subcarriers. In that way,
inactive subcarriers may be put in low energy mode (by switching off modulator drivers, transmit and receive-
side framers, etc.).
Turned off

1295 1300 1305 1310 λ [nm] 1295 1300 1305 1310 λ [nm] 1295 1300 1305 1310 λ [nm]
A to B (fixed sense) A to B (fixed sense)
A to B (fixed sense)
Turned off

1295 1300 1305 1310 λ [nm] 1295 1300 1305 1310 λ [nm] 1295 1300 1305 1310 λ [nm]

B to A (fixed sense) A to B B to A (fixed sense) A to B B to A (fixed sense) A to B


B to A B to A B to A

(a) Fixed-sense operation (b) A→B bandwidth maximization (c) Minimum power operation
Fig. 3. FST operation scenarios

4. CONCLUSIONS
We proposed a platform for flexible-sense optical transmission, FST, for duplex system bandwidth allocation in
both senses of transmission, according to the temporal traffic requirements in asymmetric optical links with high
peak-to-average traffic characteristics. FST interface architecture relies on duplicating transmitter and receiver at
both end nodes to allow flexible sense operation, and is applicable without changing the installed equipment
since the optical components are bidirectional (e.g. passive WDM systems). We estimate that photonic
integration-based implementation would result in a relatively low additional manufacture cost with little increase
in packaging size, comparing to standard modules. In addition, FST may be explored to reduce system latency
and power consumption.

ACKNOWLEDGEMENTS
This work was supported by the Innovation Center, Ericsson Telecomunicações S.A., Brazil.

REFERENCES
[1] “IEEE 802.3™ Industry Connections Ethernet Bandwidth Assessment”, IEEE 802.3 Ethernet Working
Group, July, 2012. Available at: http://www.ieee802.org/3/ad_hoc/bwa/BWA_Report.pdf
[2] “Toward Deeply Virtualized Elastic Optical Networks” M. Jinno, A. Hirano OFC 2013 Tech. Digest.
[3] “Software Defined Optical Networks Technology and Infrastructure: Enabling Software-Defined Optical
Network Operations”, D. Simeonidou, R. Nejabati, M. Channegowda, OFC 2013, Tech. Digest.
[4] “BER-adaptive Flexible-format Transmitter for Elastic Optical Networks”, H. Choi, T. Tsuritani, I.
Morita., Optics Express, No. 20, 2012.
[5] “Limitations in Dense Bidirectional Transmission in Absence of Optical Amplification”, S. Radic,
Photonics Techn. Lett., IEEE , vol.14, no.1, Jan. 2002.
[6] “Limitations Imposed by Rayleigh Backscattering in Closely Interleaved, Bidirectional WDM
Transmission Systems,” M. Sumida, T. Kubo, T. Imai, Photonics Techn. Lett., IEEE , vol.15, no.1, Jan. 2003.
[7] “Bidirectional WDM Transmission Technique Utilizing Two Identical Sets of Wavelengths for Both
Directions Over a Single Fiber”, H. Obara, J. Lightwave Techn., vol.25, no.1, Jan. 2007.
[8] Center of Study and Research in Network Technology and Operations – CEPTRO, Brazil. Available at:
http://www.ptt.br/trafego/pix/sp
[9] “The Nature of Datacenter Traffic: Measurements & Analysis” S. Kandula et al. In Proc. ACM
SIGCOMM, ACM, 2009.
[10] “Amendment 4: Media Access Control Parameters, Physical Layers, and Management Parameters for 40
Gb/s and 100 Gb/s Operation”, IEEE Standard 802.3ba™ -2010. June 2010.
[11] “10X10 MSA Technical Specifications, Rev 2.5,” Aug. 2012. Available at:
http://www.10x10msa.org/documents/MSA%20Technical%20Rev2-5.pdf
[12] “ITU-T G.7042/Y.1305, Link Capacity Adjustment Scheme (LCAS) for Virtual Concatenated Signals”,
March 2006.
[13] “ITU-T G.7044/Y.1347, Hitless adjustment of ODUflex(GFP)”, Oct. 2011.

You might also like