Christopher Nitta

LIONS: An AWGR-Based Low-Latency Optical Switch for High-Performance Computing and Data Centers

This paper discusses the architecture of an arrayed waveguide grating router (AWGR)-based low-latency interconnect optical network switch called LIONS, and its different loopback buffering schemes. A proof of concept is demonstrated with a 4 × 4 experimental testbed. A simulator was developed to model the LIONS architecture and was validated by comparing experimentally obtained statistics such as average end-to-end latency with the results produced by the simulator. Considering the complexity and cost in implementing loopback buffers in LIONS, we propose an all-optical negative acknowledgement (AO-NACK) architecture in order to remove the need for loopback buffers. Simulation results for LIONS with AO-NACK architecture and distributed loopback buffer architecture are compared with the performance of the flattened butterfly electrical switching network.

A Scalable, Low-Latency, High-Throughput, Optical Interconnect Architecture Based on Arrayed Waveguide Grating Routers

This paper proposes, simulates, and experimentally demonstrates an optical interconnect architecture for large-scale computing systems. The proposed architecture, Hierarchical Lightwave Optical Interconnect Network (H-LION), leverages wavelength routing in arrayed waveguide grating routers (AWGRs), and computing nodes (or servers) with embedded routers and wavelength-specific optical I/Os. Within the racks and clusters, the interconnect topology is hierarchical all-to-all exploiting passive AWGRs. For the intercluster communication, the proposed architecture exploits a flat and distributed Thin-CLOS topology based on AWGR-based optical switches. H-LION can scale beyond 100 000 nodes while guaranteeing up to 1.83×saving in number of inter-rack cables, and up to 1.5×saving in number of inter-rack switches, when compared with a legacy three-tier Fat Tree network. Network simulation results show a system-wide network throughput reaching as high as 90% of the total possible capacity in case of synthetic traffic with uniform random distribution. Experiments show 97% intracluster throughput for uniform random traffic, and error-free intercluster communication at 10 Gb/s.

Y-Threads: supporting concurrency in wireless sensor networks

Resource constrained systems often are programmed using an event-based model. Many applications do not lend themselves well to an event-based approach, but preemptive multithreading pre-allocates resources that cannot be used even while not in use by the owning thread. In this paper, we propose a hybrid approach called Y-Threads. Y-Threads provide separate small stacks for blocking portions of applications, while allowing for shared stacks for non-blocking computations. We have implemented Y-Threads on Mica and Telos wireless sensor network platforms. The results show that Y-Threads provide a preemptive multithreaded programming model with resource utilization closer to an event-based approach. In addition, relatively large memory buffers can be allocated for temporary use with less overhead than conventional dynamic memory allocation methods.

All-Optical Physical Layer NACK in AWGR-Based Optical Interconnects

This letter, proposes and experimentally demonstrates an all-optical physical layer negative acknowledgment (AO-NACK) technique to handle contention in array waveguide grating router (AWGR)-based optical interconnects. By using back-propagation in AWGR, the packets experiencing contention are reflected back to the senders in the optical domain to serve as a physical layer negative acknowledgement to trigger the retransmission. A host-switch distance of ≈ 20 m and a packet length of 204.8 ns are used in this proof-of-principle demonstration. Notification of AO-NACKs messages and successful packet retransmission and switching is demonstrated with error-free operation at 10 and 40 Gb/s.

Scalable Optical Interconnect Architecture Using AWGR-Based TONAK LION Switch With Limited Number of Wavelengths

This paper analyzes the scalability in arrayed waveguide grating router (AWGR)-based interconnect architectures and demonstrates active AWGR-based switching using a distributed control plane. First, the paper analyses an all-to-all single AWGR passive interconnection with N nodes and proposes a new architecture that overcomes the scalability limitation given by wavelength registration and crosstalk, by introducing multiples of smaller AWGRs (W × W) operating on a fewer number of wavelengths . Second, this paper demonstrates active AWGR switching with a distributed control plane, to be used when the size of the interconnection network makes the all-to-all approach using passive AWGRs impractical. In particular, an active AWGR-based TONAK switch is introduced. TONAK combines an all-optical NACK technique, which removes the need for electrical buffers at the switch input/output ports, and a TOKEN technique, which enables a distributed all-optical arbiter to handle packet contention. The experimental validation and performance study of the AWGR-based TONAK switch is presented, demonstrating the feasibility of the TONAK solution and the high throughput and low average packet latency for an up to 75% offered load.

Resilient microring resonator based photonic networks

Microring resonator-based photonic interconnects are being considered for both on-chip and off-chip communication in order to satisfy the power and bandwidth requirements of future large scale chip multiprocessors. However, microring resonators are prone to malfunction due to fabrication errors, and they are also extremely sensitive to fluctuations in temperature. In this paper we derive a fault model for microring based optical links that can be used by computer architects to make informed design choices. We evaluate different schemes for improving resilience, such as retransmission versus error-correction, using an optical fault simulator based on our fault model. We show how meeting a target mean time between failures (MTBF) affects the choice of resilience scheme - our investigation indicates that until fault rates are in the range of 10−21 to 10−24 per cycle, error detection/correction schemes will be needed in order to meet a 1M hour MTBF. We also evaluate how the resilience scheme impacts the performance of the link, which will help an architect choose the appropriate scheme based on the throughput requirements of a particular design.

An All-Optical Token Technique Enabling a Fully-Distributed Control Plane in AWGR-Based Optical Interconnects

This paper proposes and experimentally demonstrates a fully-distributed All-Optical TOKEN (AO-TOKEN) contention resolution technique for AWGR-based optical interconnects. The AO-TOKEN technique is implemented by exploiting the saturation effect in SOAs placed at the AWGR outputs. A polarization-diversity scheme allows thoe data and control planes to share the same physical link. The AO-TOKEN is more scalable than alternative electrical/optical solutions since it eliminates the need for a centralized electrical control plane. Our experimental results show that the technique can work over a wavelength-range of ≈ 23 nm using off-the-shelf components. We also successfully demonstrate all-optical contention resolution, packet transmission, and switching with error-free operation at 10 Gb/s.

Scalable and Distributed Contention Resolution in AWGR-Based Data Center Switches Using RSOA-Based Optical Mutual Exclusion

We describe a mutual exclusion element using a reflective semiconductor optical amplifier (RSOA) and a simple scheme for contention resolution in arrayed waveguide grating router (AWGR)-based optical switches in data centers. We describe a hardware demonstration and detailed performance analysis of an AWGR-based optical switch based on the proposed concept. We show that the proposed RSOA-based contention resolution significantly reduces latency compared to existing methods and that it does not require any global or centralized coordination, which makes it inherently scalable and suitable for emerging data center networks.

AWGR-based all-to-all optical interconnects using limited number of wavelengths

This paper discusses all-to-all interconnections for N nodes using W wavelengths and arrayed waveguide grating routers (AWGRs). This scheme allows for reducing the optical wiring by a factor of W(N-1)/2N and isolating the crosstalk.

DCAF - A Directly Connected Arbitration-Free Photonic Crossbar for Energy-Efficient High Performance Computing

DCAF is a directly connected arbitration free photonic crossbar that is realized by taking advantage of multiple photonic layers connected with photonic vias. In order to evaluate DCAF we developed a detailed implementation model for the network and analyzed the power and performance on a variety of benchmarks, including SPLASH-2 and synthetic traces. Our results demonstrate that the overhead required by arbitration is non-trivial, especially at high loads. Eliminating the need for arbitration, sizing the buffers carefully and retransmitting lost packets when there is contention results in a 44% reduction in average packet latency without additional power overhead. We also use an analytical model for ScaLAPACK QR decomposition and find that a 64 processor DCAF could outperform a 1024 node cluster connected with 40Gbps links on matrices up to 500MB in size.