Beruflich Dokumente
Kultur Dokumente
945
Abstract—In this work scaling of an optical broadcast- ditional requirements pertaining to reliability, power con-
and-select network based on a passive star coupler is ex-
sumption, package size, weight, and tolerance to extreme
plored for avionics applications. Each client in the network
is equipped with a transmitter unit and a multichannel temperature variation and vibration. For instance, the
receiver capable of receiving signals from all other clients temperature of an avionics computing system can range
connected to the star coupler. We propose a connecting from −40°C to 100°C [3], while many consumer-grade
node concept to scale the number of clients supported by components are only designed to operate from 0°C to
the architecture. These connecting nodes act as bridges be-
tween star couplers, enabling the organization of several 70°C [4].
star couplers into a topology with additional clients. This Fiber-optics interconnection networks, particularly with
design is modeled in the PhoenixSim simulation environ-
wavelength division multiplexing (WDM) technology,
ment, and system-level simulation results are reported.
We then propose the ring topology and dimension-N present a potentially attractive solution to the avionics re-
topology to interconnect and scale star couplers. Finally quirements of high bandwidth for data processing applica-
we compare the ring and dimension-N topologies in terms tions and low latency for control-related signals. Compared
of scalability limit at different crossing traffic loads,
revealing the trade-offs between latency, system complex-
with traditional copper-based wires, interconnects based
ity, and scalability. Our study shows that a robust, low- on fiber optics have additional advantages, such as resis-
latency network of up to hundreds of clients, sufficient tance to electromagnetic interference and low weight.
for current and next-generation avionics applications, Fiber-optic cables have already been deployed in avionics
can be built using off-the-shelf and near-term commercial
technology. platforms, but the network switches are still electrical.
For example, the F-35 Joint Strike Fighter uses a
Index Terms—Avionics; Network architecture; Optical 2 Gbit∕s fiber channel and two 32-port electrical switch
interconnect; System simulation. modules for its high-speed data network [5]. The first com-
mercial application of a fiber-optic local area network
(LAN) was established in the Boeing 777 [6]. Based on
I. INTRODUCTION the Fiber Distributed Data Interface (FDDI) standard,
100 Mbit∕s connections were provided between the net-
work units. New network protocols such as the ARINC
data rate and modulation format transparency, high and optical Clos. In order to avoid the high cost of testbed
reliability, and ease of maintenance. This qualifies the prototyping and integration with existing systems,
simple and completely passive star coupler as an attractive an analysis and comparison of the network architectures
solution for avionic systems. Compared with electrical are performed using the Library for Integrated Optical
switches, which exhibit latencies of the order of microsec- Networks (LION) discrete-event simulation model.
onds [9], the latency of the star coupler can approach the Network performance such as average packet latency is
time-of-flight latency. (Details of latency of the full link are evaluated and compared with different network architec-
discussed in Subsection IV.D.) Another important concern tures, which is similar to our approach.
with the avionics platform is failure probability. The com- Habiby and Vaidyanathan [12] present the networking
plete passiveness of the star coupler results in a lower fail- challenges of introducing optical networks in aircraft appli-
ure rate than with an active switch. Since aircraft may cations and standards progress of a WDM-based optical
need to communicate with different external sensors with backbone network. A case study comparing copper-based
disparate protocols, the format transparency of the optical and fiber-based cable infrastructures is also highlighted.
network is also particularly attractive.
The benefits and cost of optical networking in the aero-
To evaluate the potential of a star-coupler-based space platform and recent technology progress in terms
broadcast-and-select optical network, this architecture is of device components, network design, simulation, and
studied and evaluated within the PhoenixSim simulation modeling are examined in [13]. This study also highlights
environment. The WDM network utilizes silicon photonics the need for simulation tools in the evaluation of different
technology [10] and commercial off-the-shelf components. network architectures and the exploration of hypothetical
Each client in the network has one transmitter unit, emit- operational scenarios. Reference [14] describes a four-node
ting on one wavelength, and one receiver unit, which passive optical backplane network connecting sensors, pro-
includes an arrayed waveguide grating (AWG) and 32 pho- cessors, and shared storage. Each module in the network is
todetectors, receiving on all wavelengths. The clients are given a unique wavelength on which it receives informa-
connected by a star coupler, which broadcasts any message tion. Information destined for a particular end-point mod-
that is incident on an input fiber port to all output fiber ule must be sent on the wavelength associated with that
ports. destination.
The number of clients connected by a star coupler is lim-
ited if no signal amplification is employed. We address this
scalability challenge by proposing a method that scales B. Optical Passive Star Network
the star coupler network by using connecting nodes. The
extended scalability provided by interconnecting star cou- Several contributions [15,16] discuss the optical passive
plers in two topologies, ring and dimension-N, is analyzed star topology and study multihop multistar interconnection
with mathematical formulations and also presented networks and the associated protocols. A key assumption in
numerically. these papers is the utilization of a single-wavelength optical
The remainder of this paper is organized as follows. passive star (i.e., only one optical signal can be guided
Section II presents a generalized overview of interconnec- through the device at one time). The authors in [15] study
tion networks in an avionics platform, related prior litera- architectures based on Kautz graphs and stack-graphs, an-
ture on passive star networks, and the PhoenixSim alyze the issue of access and control protocols, and show
simulation environment. Section III introduces photonic evaluation through simulation. The work in [16] introduces
device models and the network performance of the optical the partitioned optical passive stars (POPS) topology, an
star coupler architecture. Section IV discusses the connect- all-optical nonhierarchical star-based topology to achieve
ing node structure, performance evaluation of star-coupler- a single-hop network. High performance and utilization is
based networks using connecting nodes, latency issues, shown with random communication patterns. An efficient
ring and dimension-N topologies, and the scalability protocol design coordinating data transmission in a star-
limit of these two topologies. Concluding remarks are made coupler-based network is presented in [17]. In this design,
in Section V. The detailed mathematical formulation of the stations are equipped with fixed transmitters and tunable
two topologies studied is included in Appendixes A and B. receivers. The authors in [18] propose time-division-
multiplexing-based channel sharing for a passive-star-
based multihop network.
II. RELATED WORK
Different from the above network architectures, the cli-
ent nodes in our network have multichannel receivers, and
A. Avionics Communication Network each client is able to receive messages from all other clients
without collisions. This eliminates the need for a compli-
A large body of research has been dedicated to the design cated scheduling protocol design. The LAMBDANET net-
and evaluation of communication networks in an avionics work [19], with a similar network architecture, has been
context. Many of these have focused on designing a WDM- proposed for voice traffic transport and video distribution
based optical backbone network. The work in [11] gives an applications. An experimental demonstration was per-
overview of a WDM LAN for military avionics network formed to establish the feasibility of a 16-port star-coupler
architectures including interconnected rings, optical tree, network using commercially available components.
Li et al. VOL. 5, NO. 9/SEPTEMBER 2013/J. OPT. COMMUN. NETW. 947
However, this scheme becomes impractical when the feasibility of the device parameters can be studied. The
number of clients increases [20]. use of PhoenixSim is not limited to network-on-chip appli-
cations; the event-driven kernel and modular-based device
libraries make it easily extensible to other network simu-
C. Network Simulations in PhoenixSim lation applications such as the avionics environment.
Developed at Columbia University, PhoenixSim (Fig. 1) III. DEVICE MODELS AND NETWORK PERFORMANCE
is a software toolkit that enables design space exploration
of photonic interconnection networks through physical-
level-aware event-driven simulations. PhoenixSim in-
A. Photonic Device Models
cludes a library of electronic and photonic device models,
including modulator [21], photodetector [22], and fiber con- Our baseline star coupler network model is illustrated
nection, which can be tailored to measured parameters. in Fig. 2. A 32-port star coupler connects 32 compute
New device models can also be easily added by extending nodes (network clients). Each client is linked to a trans-
existing device models [23]. The primary purpose of Phoe- mitter and receiver unit. The design utilizes 32 unique
nixSim is to reveal the system-level effects of integrating wavelengths on the 100 GHz ITU grid for node-to-node
photonic devices into interconnection networks. Recent communications.
work has focused on using PhoenixSim to show the benefits Each client is attached to a transmitter unit, which is
and trade-offs of optically enhanced chip-scale network configured to emit at a wavelength unique to that transmit-
architectures [23,24]. ting node, thereby avoiding collisions. This transmitter
PhoenixSim is based on the OMNeT++ discrete-event unit, assumed to have a 10 Gbit∕s data rate in this case,
simulation environment [25] and includes a library of elec- has an integrated modulator and can be configured to work
trical and photonic device models that are highly parame- at one of the 32 wavelengths. All transmitters in the
terized. PhoenixSim models photonic devices using a network have a common hardware configuration, which
relatively high level of abstraction by establishing device reduces hardware complexity and cost. Based on typical
parameters that are essential for the system-level under- device characterizations, we assume an input laser power
standing of a photonic interconnection network. Device of 10 dBm and transmitter insertion loss of 7 dB, for a chip
characteristics such as insertion loss, propagation latency, output optical power of 3 dBm.
cross talk, bandwidth, and energy dissipation are ab- Each client is also equipped with a multichannel receiver
stracted to describe the photonic devices, which can be de- with 32 photodetectors, permitting it to receive messages
termined through experimental measurements or through from all other clients connected to the star coupler. The mul-
performance projections of future devices [8,26]. A logical tiwavelength receiver design is fabricated on silicon-on-
routing table is used to determine the path a message takes insulator using a 100 GHz spacing AWG for demultiplexing
through the device. Each property is represented as an the WDM optical signal into separate channels and 32
N × N matrix where the row corresponds to the input port monolithically integrated epitaxially grown germanium
and the column represents the output port. Each entry in a waveguide photodetectors. The multichannel high-speed
matrix corresponds to the value used for the particular receiver operates at a 10 Gbit∕s data rate, as experimentally
input/output combination. As a simple example, the optical demonstrated in [22]. In order to stabilize the AWG temper-
fiber is modeled as a two-port device with its latency rep- ature in the avionics environment, a thermoelectric cooler
resented as a 2 × 2 matrix, and the (1,2), (2,1) item is L × t, module is used; athermal AWGs with large numbers of chan-
where L is the fiber length and t is the propagation delay nels have been demonstrated to have a small wavelength
per unit length. By capturing the physical-layer aspects of shift with large temperature variations [27–29]. The electri-
the devices, the implications in terms of system-level cal circuits associated with the multichannel receiver selects
the channel to be sent to the client. The additional optical-to-
Photonic Elements electronic and electronic-to-optical conversion latency is of
Interconnection the order of a few nanoseconds [30].
Network
Design
System-Level
Simulation
Throughput
Latency
requirements diminish the scalability and achievability of lengths λ1 –λ20 , this clearly constitutes a bottleneck. The
the design. Finally, the star coupler itself would only scale bottleneck can be alleviated by adding additional connect-
at the price of a higher manufacturing cost. ing nodes. In this way, in Fig. 4(b), 10 connecting nodes are
The proposed approach for increasing port count uses placed between the two stars by using 10 wavelengths
connecting nodes to connect different star couplers and repeating the messages received in each star on λ21 –λ30 .
keeps the 32-port star coupler component unmodified. To The capacity provided by these multiple connecting
avoid significant modification of the 32-port star coupler nodes can be allocated equally. For instance, connecting
network and the endpoint node structure, the connecting node 1 will retransmit in the other star any messages sent
node embeds the transmitter unit and multichannel by client nodes 1 and 2 only (thus received on wavelengths
receiver unit previously introduced, thus saving the addi- λ1 and λ2 ), connecting node 2 will be in charge of clients
tional development time and associated cost. A connecting 3 and 4, and so on. Alternatively, this capacity can be allo-
node is responsible for receiving messages on one or more cated according to the traffic requirements. In the above
wavelengths from star coupler A and broadcasting them example, with 20 clients and 10 connecting nodes, connect-
(or a selection of them, as will be further explained) to star ing nodes 1–4 can be devoted entirely to clients 1–4, respec-
coupler B (Fig. 4) on a given wavelength (and vice versa; tively, while the 6 other connecting nodes are shared among
therefore two transmitters and receiver units are required clients 5–20. This allocation is optimal if clients 1–4 are
per connecting node). Since two or more messages can over- sending four times more messages than clients 5–20.
lap on the connection node reception side, a connecting It is worth noting that this allocation, achieved through
node is able to deserialize concurrently incoming messages wavelength assignment, can be implemented either stati-
on several wavelengths. Messages that are not immedi- cally or dynamically. In the static case, the connecting
ately retransmitted are buffered. nodes can be equipped with the corresponding photodetec-
Two examples are provided in Fig. 4. The same architec- tors and interfaces only, resulting in a lower cost. In the
tures are designed to connect 40 nodes by connecting two dynamic scheme, each node is equipped with all photode-
passive star couplers either through one connecting node tectors, some being turned off. Also note that to prevent
[Fig. 4(a)] or ten connecting nodes [Fig. 4(b)]. In Fig. 4(a), messages from endless circulation from one star to another,
the single connecting node listens on the client wave- connecting nodes will not react to self-originated messages.
lengths 1–20 (on each side) and repeats these messages For instance, in Fig. 4(b), connecting nodes will only repeat
on the other side on wavelength λ21 . Since only one wave- messages received on wavelengths λ1 to λ20 .
length is used to repeat all messages received on wave-
B. Dumb Versus Smart Connecting Nodes
nodes, while there will be large delays with only one In fact, with full crossing traffic load, the maximum number
connecting node. of supported clients of the star topology is only 45—three
It is worth noting that having the star couplers spread star couplers with 15 clients and 10 connecting nodes
around the aircraft permits greater flexibility in the final connected to a central star coupler. Therefore a star topology
implementation of the network. Compute node pairs re- in this case has minimal scalability.
quiring particularly low latency can be connected to the The drawback of the ring topology is that, in the worst
same star coupler to avoid the latency related to the multi- case, the crossing traffic has to traverse half the ring to
ple hops. Additionally, star couplers can be located very reach its destination. This not only adds to the overall
close to some clients, further improving the latency by system latency, it also places a greater burden on the
reducing traveling distances. The topology design space connecting nodes, decreasing the number of clients that
exploration of connecting star couplers can be found in can be supported. In order to minimize the system latency
Subsections IV.E and IV.F. and maximize the number of clients, we propose the
dimension-N topology (N number of star couplers − 1).
An example of connecting five star couplers with
E. Ring and Dimension-N Topology dimension-4 topology is shown in Fig. 8. There are 15 cli-
ents per star coupler; since the generated traffic has 0.8
Star couplers can be further interconnected to add more probability of being crossing traffic, 15 × 0.8 12 connect-
clients in different topologies. One particularly interesting ing nodes are needed. Because there are four links for each
topology is a ring topology (Fig. 7). In the ring topology, the star coupler, three connecting nodes are needed per link. A
connecting nodes not only have to transmit messages link consists of three connecting nodes and the associated
generated within the same star coupler, they will also be fibers that are used to connect between the star coupler and
responsible for retransmitting messages that require multi- the connecting node. Note that the ring topology can be
ple hops to reach their destination. This puts additional viewed as a dimension-2 topology, and the dimension-N
pressure on the connecting nodes. In such scenarios, smart topology is effectively a fully connected network.
connecting nodes must be used to prevent an avalanche ef- The dimension-N topology and the ring topology re-
fect in messages. The wavelengths that the connecting no- present two extreme cases in the design space aiming for
des receive have to be specified, which, depending on the low latency, low system complexity, and low cost. The dimen-
application, may require careful distribution of the avail- sion-N topology always achieves lower latency than the ring
able wavelengths among the connecting nodes. This can topology, since the worst-case latency is only one hop. How-
be done dynamically. One way of implementing it is to have ever, the disadvantage of the dimension-N topology is the
a central controller that interfaces with the connecting no- complex fiber connections and the number of connecting no-
des and assigns the wavelengths according to the traffic re- des required, since the number of connections scales with
quirements. This approach will work for smaller scale ON 2 , while ring topology only scales with ON.
networks, which is the case for the avionics applications.
The star couplers can also be interconnected in a star top-
ology in which the peripheral star couplers are connected to F. Scaling Limit
a central star coupler. However, in this case all the crossing
traffic has to go through the central star coupler; therefore, First we analyze the maximum number of clients sup-
the number of available connecting nodes for each star ported by a ring topology. We consider the case where each
coupler is limited by the central star coupler connectivity.
This in turn limits the total number of supported clients.
client has equal probability to communicate with any other load of 4 Gbits∕s), 120 clients can be supported using
client. As the number of coupled stars increases, an only six star couplers in the ring topology, which is an
increasing number of connecting nodes must be devoted approximately 1.7× increase compared with full traffic
to crossing traffic, thus decreasing the number of wave- load. The oscillations in the maximum number of clients
lengths available for compute nodes. For the ring topology, are due to the fact that increasing the number of star cou-
the increase in the number of hops further increases the plers requires more connecting nodes, thus, given the
load on the connecting nodes. In situations where some wavelength constraint, leaving less clients per star coupler.
source–destination pairs having intense communications For instance, when traffic load is equal to 1, 90 clients can
are placed in the same star coupler, the crossing traffic load be supported by using 10 star couplers, i.e., 9 clients and 23
can be significantly reduced, requiring fewer connecting connecting nodes per star coupler. However, when 11 star
nodes and increasing the scaling limit. Therefore the couplers are connected in a ring, only 88 clients can be sup-
scaling limits are also examined at different crossing ported, since 24 connecting nodes are required to fulfill the
traffic loads. crossing traffic load, leaving only 8 clients per star coupler.
Let ρ be the offered load per client, S be the number of As a comparison, we investigate the scalability of the
star couplers, and W be the maximum number of ports per ring and the dimension-N topology at different crossing
star coupler; then the maximum number of clients T max can traffic loads. The scalability of dimension-N is computed
be computed as (proof is included in Appendix A) by sweeping the number of clients until we find the maxi-
mum number of clients that can be supported by a star cou-
W W pler at a given traffic load (Appendix B includes the
T max ≤ S · ρ . (1)
ρ
1 ·
S
S2 −1
4 4 S1 − 4Sρ 2 detailed mathematical formulation). The result is com-
pared with the ring topology at crossing traffic loads of
T max serves as the upper bound for the maximum number 1 and 0.1 (Fig. 10). At full crossing traffic load, the
of supported clients given a number of star couplers and a dimension-N topology can support more clients compared
traffic load. From Eq. (1) we can see that the asymptotic with the ring topology, and this advantage becomes larger
scalability limit of the ring topology is as the number of star couplers increases beyond five. At low
crossing traffic load (0.1 in this case), the ring topology is
4W able to achieve a higher scalability, especially when the
T asym : (2) number of star couplers is larger than ten. This is expected:
ρ
When there is full crossing traffic demand, the best way
For W 32 and ρ 1, T asym 128. of providing enough bandwidth is to construct a fully
connected network. However, when the crossing traffic
Figure 9 plots the maximum number of clients that can demand is low, maintaining a fully connected network is
be supported with the ring topology at different crossing costly, since too many ports are used by the links to other
traffic loads. Compared with a single star coupler, using star couplers, reducing resources for clients.
a ring topology can achieve up to a 4× increase in the num-
ber of supported clients at full crossing traffic load. The We illustrate the advantage of a star-coupler-based net-
scalability can also be greatly enhanced if some com- work with the following example. Suppose there are 140
promises can be made on the crossing traffic load. For ex-
ample, if the crossing traffic load is 0.4 (i.e., the 10 Gbit∕s
line is active 40% of the time, resulting in a time averaged
~2 ρ C
ρ~ s C ~2 ρ C
~ ρ; (A10) T ~ W
~ ~ S C ; (A18)
T CS ~S ~2
1 ρ · S −1
S~ 4
and we have (for the odd case only; the even case follows ~ Eq. (A18) can be solved to find S:
~
the same idea) with C ≤ C.
r
2
~lN
N ~ ρ · S − 1:
~ r ρs · T l C (A11)
T T 2 − ρT W − ρT 4
S 8 S~ : (A19)
ρT
2 W− 4
Using Eq. (A9), we can write
~ is the upper bound of the number of clients, S~ is the
Since C
8
> W
< 1 ρ ·S2 −1 if S is odd lower bound of the number of required star couplers.
~
C S 8
: (A12)
>
: W
2 if S is even
1Sρ ·S8
APPENDIX B: SCALABILITY OF THE DIMENSION-N TOPOLOGY
This constitutes an approximation of C. We can easily
prove that this approximation is the upper bound of In the dimension-N topology, the offered load between
~ ≥ Cmax . From Eq. (A9) we know N
Cmax , i.e., C ~lN ~r two clients ρsd is
~ Since N l is the ceiling of N
W − C. ~ l (and so for N r ), we have ρ ρ
N~lN ~ r ≤ N l N r . Combining these two statements, we ρsd : (B1)
T CS
get that W − C ~ ≤ N l N r . Constraint (A8) must hold such
that N l N r ≤ W − Cmax . So we have that Therefore the load exchanged between two star couplers
ρs is
W−C ~
~ ≤ W − Cmax ⇔ Cmax ≤ C: (A13) Cρ
ρs C2 · ρsd ; (B2)
S
To remove the odd or even cases for the upper bound, we
can use the fact that for the same S ≥ 1, the upper bound of and the number of connecting nodes and clients per star
the even case is always smaller than the odd case, so that coupler is constrained by
Cmax ≤
W
: (A14)
C S − 1 ⌈ CρS ⌉ ≤ W: (B3)
ρ S2 −1
1 · S 4
The maximum number of clients can be obtained by sweep-
~ approximation (as well ing C from 1 to W until Eq. (B3) cannot be satisfied. Here
From this point we can use the C
we provide a mathematical upper bound of C. Again, if we
as the bounds) to approximate (and bound) T max . Since
temporarily allow fractional connecting nodes and clients,
T max Cmax · S, therefore
the number of clients per star coupler becomes
W W
T max ≤ S · ρ : (A15) ~ W
1 · ρ S2 −1 S1 − 4Sρ 2 C : (B4)
S 4 4 ρ 1 − Sρ
From Eq. (A15) we can see that the asymptotic scalabil- We want to prove that C ~ is the upper bound of C, i.e.,
ity limit of the scheme is ~
Cmax ≤ C. This can be proved by
T asym
4W
ρ
: (A16) Cmax S − 1
Cmax ρ
S
C ρ
≤ Cmax S − 1 max ≤ W
S
⌈ ⌉
~
For W 32 and ρ 1, T asym 128. C~ S − 1 Cρ : (B5)
S
Another interesting problem is finding the minimum
number of star couplers to connect T clients, i.e., given ~
From Eq. (B5) we can see easily that Cmax ≤ C.
ρ, W, and T, minimize S subject to
REFERENCES
C · S ≥ T; C N l N r ≤ W: (A17)
[1] S. F. Habiby and M. J. Hackert, “RONIA results: WDM-based
We again relax integrity constraints and allow a frac- optical networks in aircraft applications,” in IEEE Avionics,
tional number of clients, connecting nodes, and star cou- Fiber-Optics and Photonics Technology Conf., 2008, pp. 71–72.
plers. To minimize S,~ we have to maximize C. ~ Maximal
[2] M.-Y. Nam, E. Seo, L. Sha, K.-J. Park, and K. Kang, “Limiting
C~ can be approximated by using the odd case of Eq. (A12), worst-case end-to-end latency when traffic increases in a
since it returns the largest value of C; we have switched avionics network,” in IEEE 17th Int. Conf. on
Li et al. VOL. 5, NO. 9/SEPTEMBER 2013/J. OPT. COMMUN. NETW. 955
Embedded and Real-Time Computing Systems and Applica- [20] K. R. Desai and K. Ghose, “An evaluation of communication
tions (RTCSA), 2011, vol. 1, pp. 285–294. protocols for star-coupled multidimensional WDM networks
[3] M. Masanovic, L. A. Johansson, and J. Barton, “Widely for multiprocessors,” in Proc. 2nd Int. Conf. on Massively
tunable optical transceiver for avionic WDM networks,” in Parallel Processing Using Optical Interconnections, 1995,
IEEE Avionics, Fiber-Optics and Photonics Technology Conf. pp. 42–49.
(AVFOP), 2011, pp. 23–24. [21] Q. Li, N. Ophir, L. Xu, K. Padmaraju, L. Chen, M. Lipson, and
[4] G. J. Whaley and R. J. Karnopp, “Air Force highly integrated K. Bergman, “Experimental characterization of the optical-
photonics program: Development and demonstration of an op- power upper bound in a silicon microring modulator,” in IEEE
tically transparent fiber optic network for avionics applica- Optical Interconnects Conf., May 2012, pp. 38–39.
tions,” Proc. SPIE, vol. 7700, 77000A, Apr. 2010. [22] Y. Kuo, M. Kwakernaak, X. Sun, J. Pescatore, M. Gilmer, J.
[5] J. Levis, B. Sutterfield, and R. Stevens, “Fiber optic communi- Oakley, Z. Ji, and A. Nikolov, “Integrated multi-wavelength
cation within the F-35 mission systems,” in IEEE Conf. Avionics silicon germanium high speed receivers,” in Integrated Pho-
Fiber-Optics and Photonics, 12–14 Sept. 2006, pp. 12–13. tonics Research, Silicon and Nanophotonics (IPRSN), 2010,
paper IWF3.
[6] D. E. Anderson and M. W. Beranek, “777 optical LAN technol-
ogy review,” in 48th IEEE Electronic Components & Technol- [23] J. Chan, G. Hendry, K. Bergman, and L. P. Carloni, “Physical-
ogy Conf., 1998, pp. 386–390. layer modeling and system-level design of chip-scale photonic
[7] A. A. R. Lee and S. D. Rayner, “Avionic architectures incorpo- interconnection networks,” IEEE Trans. Comput.-Aided
rating optical fibre technology,” in IEEE Conf. Avionics Fiber- Des. Integr. Circuits Syst., vol. 30, no. 10, pp. 1507–1520,
Optics and Photonics, 12–14 Sept. 2006, pp. 10–11. Oct. 2011.
[8] J. Chan, G. Hendry, A. Biberman, K. Bergman, and L. P. [24] G. Hendry, E. Robinson, V. Gleyzer, J. Chan, L. P. Carloni, N.
Carloni, “PhoenixSim: A simulator for physical-layer analysis Bliss, and K. Bergman, “Time-division-multiplexed arbitra-
of chip-scale photonic interconnection networks,” in Proc. tion in silicon nanophotonic networks-on-chip for high-
Conf. on Design, Automation and Test in Europe, Mar. 2010, performance chip multiprocessors,” J. Parallel Distrib.
pp. 691–696. Comput., vol. 71, no. 5, pp. 641–650, May 2011.
[9] Demartek, “High performance fibre channel switch vs. unified [25] A. Varga, “OMNeT++ discrete event simulation system”
port switch technology” [Online]. Available: http://www [Online]. Available: http://www.omnetpp.org.
.demartek.com/Reports_Free/Demartek_Brocade_6510_FC_ [26] Q. Li, R. Hendry, J. Chan, K. Bergman, M. Glick, and R. Dutt,
Switch_Latency_Evaluation_2012‑09.pdf. “Network simulation of passive optical broadcast-and-select
[10] R. Soref, “The past, present, and future of silicon photonics,” network for avionics applications,” presented at the Govern-
IEEE J. Sel. Top. Quantum Electron., vol. 12, no. 6, pp. 1678– ment Microcircuit Applications and Critical Technology Conf.
1687, 2006. (GOMACTech-13), Las Vegas, Nov. 2013.
[11] C. Reardon, J. Profumo, and A. George, “Comparative simu- [27] T. Saito, K. Nara, K. Tanaka, Y. Nekado, J. I. Hasegawa, and
lative analysis of WDM LANs for avionics platforms,” in IEEE K. Kashihara, “Temperature-insensitive (athermal) AWG
Military Communications Conf., 2006. modules,” Furukawa Rev., vol. 24, pp. 29–33, 2003.
[12] S. F. Habiby and R. Vaidyanathan, “WDM optical backbone [28] S. Kamei, Y. Inoue, A. Kaneko, T. Shibata, and H. Takahashi,
networks in aircraft applications: Networking challenges “Recent progress on athermal AWG wavelength multiplexer,”
and standards progress,” in IEEE Military Communications Proc. SPIE, vol. 6014, 60140H, Oct. 2005.
Conf., 18–21 Oct. 2009, pp. 1–6. [29] X. Wang, S. Xiao, W. Zheng, F. Wang, Y. Hao, X. Jiang,
[13] J. Jackel, “Advances in optical networking for aerospace M. Wang, and J. Yang, “Slot-based athermal silicon
platform applications,” in IEEE Avionics, Fiber-Optics and arrayed-waveguide grating (AWG),” Proc. SPIE, vol. 7134,
Photonics Technology Conf. (AVFOP), 2011, pp. 5–6. 71340X, 2008.
[14] D. Drury, “Satellite optical backplane,” in IEEE Avionics, [30] S. Nishimura, H. Inoue, S. Hanatani, H. Matsuoka, and T.
Fiber-Optics and Photonics Technology Conf. (AVFOP), 2011, Yokota, “Optical interconnections for the massively
pp. 17–18. parallel computer,” IEEE Photon. Technol. Lett., vol. 9,
no. 7, pp. 1029–1031, July 1997.
[15] D. Coudert, A. Ferreira, and X. Munoz, “A multihop multi-
OPS optical interconnection network,” J. Lightwave Technol., [31] The F-35 Lightning II Program [Online]. Available: http://
vol. 18, no. 12, pp. 2076–2085, Dec. 2000. www.jsf.mil/f35/f35_variants.htm.
[16] G. Gravenstreter, R. G. Melhem, D. M. Chiarulli, S. P. [32] Boeing, “747–8 Technical Characterizations” [Online].
Levitan, and J. P. Teza, “The partitioned optical passive stars Available: http://www.boeing.com/boeing/commercial/747family/
(POPS) topology,” in Proc. 9th Int. Parallel Processing 747‑8_fact_sheet.page.
Symp., 25–28 Apr. 1995, pp. 4–10. [33] J. Zhang, Y. An, M. S. Berger, and A. T. Clausen, “Wavelength
[17] R. Chipalkatti, Z. Zhang, and A. S. Acampora, “Protocols and fiber assignment problems on avionic networks,” in IEEE
for optical star-coupler network using WDM: Performance Avionics, Fiber-Optics and Photonics Technology Conf.
and complexity study,” IEEE J. Sel. Areas Commun., vol. 11, (AVFOP), 4–6 Oct. 2011, pp. 15–16.
no. 4, pp. 579–589, May 1993. [34] S. Assefa, S. B. G. Lee, C. L. Schow, W. M. J. Green, A. V.
[18] S. B. Tridandapani and B. Mukherjee, “Channel sharing in Rylyakov, R. A. John, and Y. A. Vlasov, “20 Gbps receiver
multi-hop WDM lightwave networks: Realization and perfor- based on germanium photodetector hybrid-integrated with
mance of multicast traffic,” IEEE J. Sel. Areas Commun., 90 nm CMOS amplifier,” in Conf. on Lasers and Electro-Optics
vol. 15, no. 3, pp. 488–500, 1997. (CLEO), 1–6 May 2011, pp. 1–2.
[19] M. S. Goodman, H. Kobrinski, M. P. Vecchi, R. M. Bulley, and [35] B. Grot, J. Hestness, S. W. Keckler, and O. Mutlu, “Express
J. L. Gimlett, “The LAMBDANET multiwavelength network: cube topologies for on-chip interconnects,” in IEEE 15th
Architecture, applications, and demonstrations,” IEEE J. Sel. Int. Symp. on High Performance Computer Architecture
Areas Commun., vol. 8, no. 6, pp. 995–1004, 1990. (HPCA), 14–18 Feb. 2009, pp. 163–174.
956 J. OPT. COMMUN. NETW./VOL. 5, NO. 9/SEPTEMBER 2013 Li et al.
[36] B. Attia, W. Chouchene, A. Zitouni, A. Nourdin, and R. Tourki, Johnnie Chan (S’08-M’12) received his B.S.
“Design and implementation of low latency network interface degree (with high distinction) in computer
for network on chip,” in 5th Int. Design and Test Workshop and electrical engineering and his M.S.
(IDT), 14–15 Dec. 2010, pp. 37–42. degree in electrical engineering from the
University of Virginia, Charlottesville, in
[37] M. Simmons, “Ethernet theory of operation,” Microchip Tech- 2005 and 2007, respectively, and a Ph.D.
nology Inc., Application Note AN1120, 2008 [Online]. Available: degree in the Department of Electrical Engi-
http://ww1.microchip.com/downloads/en/AppNotes/01120a.pdf. neering, Columbia University, New York in
[38] L.-S. Peh and W. J. Dally, “A delay model and speculative 2012.
architecture for pipelined routers,” in 7th Int. Symp. on His current research interests include sil-
icon photonic devices for chip-scale intercon-
High-Performance Computer Architecture (HPCA), 2001,
nection networks and optical networks for high-performance
pp. 255–266. computing.
[39] J. Kim, W. J. Dally, B. Towles, and A. K. Gupta, “Microarch-
itecture of a high-radix router,” Comput. Archit. News, vol. 33,
no. 2, pp. 420–431, 2005.
[40] A. L. Chiu and E. H. Modiano, “Traffic grooming algorithms for Howard Wang (S’06) received his B.S.
reducing electronic multiplexing costs in WDM ring networks,” degree in electrical engineering (minoring
J. Lightwave Technol., vol. 18, no. 1, pp. 2–12, Jan. 2000. in computer science and economics) from
Columbia University in 2006 and an M.S.
in electrical engineering in 2008. He is cur-
rently working toward a Ph.D. degree in
the Department of Electrical Engineering,
Qi Li (S’10) received his B.Eng. in electrical Columbia University, New York.
and computer engineering with a minor in His research interests involve interfacing
mathematics from the Hong Kong Univer- optical interconnection networks for high-
sity of Science and Technology in 2010. He performance computing.
spent the Spring of 2010 as an exchange stu-
dent at Cornell University. He received his
M.S. in 2012 and is currently working to-
ward a Ph.D. degree in the Department of
Electrical Engineering, Columbia Univer- Keren Bergman (S’87–M’93–SM’07–F’09)
sity, New York. received her B.S. degree from Bucknell Uni-
His research interests include silicon versity, Lewisburg, PA, in 1988 and her M.S.
photonics and optical interconnection networks. and Ph.D. degrees from Massachusetts
Institute of Technology, Cambridge, MA, in
1991 and 1994, respectively, all in electrical
engineering. She is currently a Professor at
Sebastien Rumley (M’12) received his the Department of Electrical Engineering,
M.S. degree in telecommunications engi- Columbia University, New York, where
neering from the Ecole Polytechnique she also directs the Lightwave Research
Fédérale de Lausanne (EPFL) in 2005, after Laboratory.
graduate studies in Lausanne, Zurich, and Her research programs involve optical interconnection net-
Santiago de Chile (PUC). He received the works for advanced computing systems, photonic packet switching,
Ph.D. degree in 2011 from the EPFL, where and nanophotonic networks on-chip.
he worked as a research assistant at the She is a Fellow of the Optical Society of America and a Fellow of
Telecommunications Laboratory between the Institute of Electrical and Electronic Engineers (IEEE). She is
2006 and 2011. the Co-Editor-in-Chief of the Journal of Optical Communications
His research focuses on optical network and Networking.
design, modeling and optimization, and on software improving
the research process.