John Wissinger

Design and Preliminary Implementation of an N

We have demonstrated a diffraction-based nonblocking, scalable N × N optical switch employing a digital micromirror display (DMD) with 12 μs switching speed, performing 100 times faster than the currently available technology. The distributed nature of diffraction makes this switch more robust than one-to-one reflective systems where a single mirror failure incapacitates an entire connection. We thereby address a key bottleneck in data centers and optical aggregation networks by decreasing circuit-switching speed and allowing for facile port count scalability.

\times

Abstract. Digital micromirror devices (DMDs) by their high-switching speed, stability, and repeatability are promising devices for fast, reconfigurable telecommunication switches. However, their binary mirror orientation is an issue for conventional redirection of a large number of incoming ports to a similarly large number of output fibers, like with analog micro-opto electro-mechanical systems. We are presenting here the use of the DMD as a diffraction-based optical switch, where Fourier diffraction patterns are used to steer the incoming beams to any output configuration. Fourier diffraction patterns are computer-generated holograms that structure the incoming light into any shape in the output plane. This way, the light from any fiber can be redirected to any position in the output plane. The incoming light can also be split to any positions in the output plane. This technique has the potential to make an “any-to-any,” true nonblocking, optical switch with high-port count, solving some of the problems of the present technology.

N Diffractive All-Optical Fiber Optic Switch

Optical functionality is being used to realize new data center architectures that minimize electronic switching overheads, pushing the processing to the edge of the network. A challenge in optically interconnected data center networks is to identify the large, bandwidth hungry flows (i.e., elephants) and efficiently establish the optical circuits. Moreover, the amount of optical resources to be provisioned during the network planning phase is a critical design problem. Flow classification accuracy affects the efficiency of optical circuits. Optical channel bandwidth, on the other hand, directly relates to the additive-increase, multiplicative-decrease congestion control mechanism of the transmission control protocol and affects the effective bandwidth allocated to elephant flows. In this paper, we simultaneously investigate the impact of two important mechanisms on data center network performance: traffic flow classification accuracy and optical bandwidth aggregation (i.e., the consolidation of several low-capacity channels into a single high-capacity one by employing advanced modulation formats for short-reach communications). We develop a discrete-event simulator for a hybrid data center network, enabling the tuning of flow classification parameters. Our simulations indicate that data center performance is highly sensitive to the aggregation level.We could observe up to a 74.5% improvement in network throughput only due to consolidating the optical channel bandwidth. We further noticed that the role of flow classification becomes more pronounced with higher bandwidth per wavelength as well as with more hot-spot traffic. Compared to a random classification benchmark, adaptive flow classification could lead to throughput improvements as large as 54.7%.

Digital micromirror device as a diffractive reconfigurable optical switch for telecommunication

OpenFlow, as a unified operator-friendly manageable network control approach, has benefits of supporting the convergence of electronic packet and optical circuit networks as well as quality-of-transmission (QoT) awareness. We experimentally present the QoT-awareness in converged OpenFlow networks with (i) QoT-aware wavelength reassignment (ii) QoT-aware path re-routing if the QoT is below the requirement. The experimental work validates efficient networking approach and provides a key direction for next generation software defined networks.

TCP flow classification and bandwidth aggregation in optically interconnected data center networks

Considering that high performance electronic computation has become extremely efficient, for an optical hardware accelerator to be relevant, it must solve a type or a set of problems where its electronic counterpart is still struggling in term of size, energy, or time. We have identified one such challenge as the minimization of large scale Ising Hamiltonians when the number of particles is on the order of a million. Here we discuss an algorithmic approach based on probabilistic inference using graphical model and message passing.

Quality of Transmission Awareness in Converged Electronic and Optical Networks with OpenFlow

We are investigating the use of optics to solve highly connected graphical models by probabilistic inference, and more specifically the sum-product message passing algorithm. We are examining the fundamental limit of size and power requirement according to the best multiplexing strategy we have found. For a million nodes, and an alphabet of a hundred, we found that the minimum size for the optical implementation is 10mm3, and the lowest bound for the power is 200 watts for operation at the shot noise limit. The various functions required for the algorithm to be operational are presented and potential implementations are discussed. These include a vector matrix multiplication using spectral hole burning, a logarithm carried out with two photon absorption, an exponential performed with saturable absorption, a normalization executed with an thermo-optics interferometer, and a wavelength remapping accomplished with a pump-probe amplifier.

All-optical graphical models for probabilistic inference

We optimize flow placement for a hybrid network implementing an adaptive neural network classifier. We predict elephant flows with high accuracy on anonymized university network traffic. We also demonstrate the capability to perform highly complex actions at 40 Gbps using less than 5% of co-processor capacity. This shows that it is possible to implement intelligent actions such as a neural network in a data center using fully programmable NICs without handicapping the server CPU.

Optical implementation of probabilistic graphical models

We demonstrate for the first time an SDN-based OFDM elastic optical network with pilot-tone assisted distributed control. Improvement in spectral efficiency and a fast reconfiguration time of 30ms have been achieved in our experiment.

Machine learning based adaptive flow classification for optically interconnected data centers

Presented here is a 32 × 32 optical switch for telecommunications applications capable of reconfiguring at speeds of up to 12 microseconds. The free space switching mechanism in this interconnect is a digital micromirror device (DMD) consisting of a 2D array of 10.8μm mirrors optimized for implementation at 1.55μm. Hinged along one axis, each micromirror is capable of accessing one of two positions in binary fashion. In general reflection based applications this corresponds to the ability to manifest only two display states with each mirror, but by employing this binary state system to display a set of binary amplitude holograms, we are able to access hundreds of distinct locations in space. We previously demonstrated a 7 × 7 switch employing this technology, providing a proof of concept device validating our initial design principles but exhibiting high insertion and wavelength dependent losses. The current system employs 1920 × 1080 DMD, allowing us to increase the number of accessible ports to 32 × 32. Adjustments in imaging, coupling component design and wavelength control were also made in order to improve the overall loss of the switch. This optical switch performs in a bit-rate and protocol independent manner, enabling its use across various network fabrics and data rates. Additionally, by employing a diffractive switching mechanism, we are able to implement a variety of ancillary features such as dynamic beam pick-off for monitoring purposes, beam division for multicasting applications and in situ attenuation control.

OFDM pilot-tone assisted distributed control for SDN-based elastic optical networks

We present here the use the DMD as a diffraction-based optical switch, where Fourier diffraction patterns are used to steer the incoming beams to any output configuration. We have implemented a single-mode fiber coupled N X N switch and demonstrated its ability to operate over the entire telecommunication C-band centered at 1550 nm. The all-optical switch was built primarily with off-the-shelf components and a Texas Instruments DLP7000™with an array of 1024 X 768 micromirrors. This DMD is capable of switching 100 times faster than currently available technology (3D MOEMS). The switch is robust to typical failure modes, protocol and bit-rate agnostic, and permits full reconfigurable optical add drop multiplexing (ROADM). The switch demonstrator was inserted into a networking testbed for the majority of the measurements. The testbed assembled under the Center for Integrated Access Networks (ClAN), a National Science Foundation (NSF) Engineering Research Center (ERC), provided an environment in which to simulate and test the data routing functionality of the switch. A Fujitsu Flashwave 9500 PS was used to provide the data signal, which was sent through the switch and received by a second Flashwave node. We successfully transmitted an HD video stream through a switched channel without any measurable data loss.