RESUMEN
The growth of data centers (DC) and high-performance computing (HPC) requires higher bandwidth, but traffic usually occurs between a small number of computing nodes, and the location of the communication bottleneck dynamically changes as the application runs. Therefore, the traditional static network that solves the communication bottleneck by providing excessive bandwidth cannot meet the demand of high performance and low cost at the same time. The reconfiguration of optical interconnects greatly improves the flexibility of the network, which can allocate unutilized bandwidth to node pairs with dense communication and improve resource utilization. However, this flexibility relies on a fast control plane to achieve efficient interaction between devices in the network. We made improvements in traffic collection, topology calculation, and optical switch configuration, and built an experimental platform to evaluate our control plane. The flexibility of optical interconnects shows a good acceleration effect when running applications that solve large-scale problems, and the experimental results show that a proper reconfiguration cycle can reduce the completion time of 3-D Fast Fourier Transform application by up to 53%.
RESUMEN
Optical network-on-chip (ONoC) is an effective communication architecture to realize high performance and energy efï¬ciency. Diverse routing algorithms are proposed to avoid the congestion, tolerate the faults, and reduce the insertion loss or energy consumption. However, existing algorithms did not consider the characteristic optical circuit-switching of ONoC, which aggravates the network congestion and degrades the associated performance severely. In this paper, by exploiting congestion prediction technique, we propose a new routing algorithm for ONoC, named loophole-routing, to improve the success rate of path-setup and decrease the latency. We use the congestion prediction technique to analyze the latency and predict the port condition caused by the network congestion. Theoretical analysis and experimental results of different synthetic traffic patterns show that the loophole-routing improves network latency over XY routing and OE-turn routing by 15.56%, 25.71%, 18.92%, 66.67% and 42.86% under uniform, hotspot1, hotspot2, transpose2 and transpose3 traffic patterns while improving the saturation throughput by 31.43%, 34.33%, 35.29%, 67.86% and 99.5% under uniform, hotspot1, hotspot2, transpose2 and transpose3 traffic patterns on average than XY routing. In addition, our proposed loophole-routing has the benefits of high path diversity and adaptive degree and low computing complexity and overhead and the potential to make fault-tolerant path selection.
RESUMEN
The dc-biased optical orthogonal frequency division multiplexing (DCO-OFDM) system is experimentally demonstrated as an appealing candidate in future visible light communication (VLC) system. However, the intrinsic high PAPR drawback that the DCO-OFDM system suffers from still needs to be addressed and few effective approach has been found so far. To deal with this problem, in this paper, the tone reservation (TR) technique based the time domain kernel matrix (TKM-TR) schemes for reducing the PAPR are studied and applied to DCO-OFDM system. Aiming at the drawback of its severe tailing in previous TKM-TR schemes, first an improved TKM-TR scheme is proposed, in which the peak regrowth caused by severe tailing is eliminated by optimizing the scaling factors. In addition, considering the clipping ratio (CR) value in TKM-TR scheme is greatly related to the PAPR reduction performance, an extensively used heuristic global optimization algorithm, the particle swarm optimization (PSO) method is employed in TKM-TR to obtain a better CR for more PAPR reduction. Simulation results show that the improved TKM-TR scheme can efficiently address the tailing problem in previous TKM-TR schemes and achieve better PAPR reduction. Moreover, due to the powerful searching ability, PSO based TKM-TR scheme achieves more PAPR reduction and lower bit error rate (BER).
RESUMEN
With the number of cores increasing, there is an emerging need for a high-bandwidth low-latency interconnection network, serving core-to-memory communication. In this paper, aiming at the goal of simultaneous access to multi-rank memory, we propose an optical interconnection network for core-to-memory communication. In the proposed network, the wavelength usage is delicately arranged so that cores can communicate with different ranks at the same time and broadcast for flow control can be achieved. A distributed memory controller architecture that works in a pipeline mode is also designed for efficient optical communication and transaction address processes. The scaling method and wavelength assignment for the proposed network are investigated. Compared with traditional electronic bus-based core-to-memory communication, the simulation results based on the PARSEC benchmark show that the bandwidth enhancement and latency reduction are apparent.