Pier Giorgio Raponi

Energy efficiency in passive optical networks: where, when, and how?

This article provides an overview of current efforts in reducing energy consumption in passive optical access networks. Both ITU-T and IEEE standardized PONs are considered. The current solutions proposed by standardization authorities, industry, and academia are classified based on the layer they address in the standardized architectures: physical layer, data link layer, and hybrid. Then, the article provides answers to major questions, such as where, when, and how to reduce PON energy consumption in TDM PONs by means of a quantitative evaluation. Results show that to reduce energy consumption, ONUs must be provided with physical devices that are not only power-efficient but also provide improved services (e.g., fast synchronization) to upper layers. For this latter purpose, novel physical ONU architectures are proposed to speed up the synchronization process and enable effective data link layer solutions. Finally, the feasibility of switching ONUs to low power mode in idle slots is assessed through a testbed implementation.

Energy-Efficient Design of a Scalable Optical Multiplane Interconnection Architecture

As the power dissipation of data centers challenges their scalability, architectures for interconnecting computers, or servers must simultaneously achieve high throughput at peak utilization and power consumption proportional to utilization levels. To achieve this goal, this paper proposes the use of an optical multiplane interconnection network, named space-wavelength (SW) switched architecture, able to route and switch packets between servers (on cards) and between processors within a card (or card ports). SW architecture exploits the space domain to address the destination card and the wavelength domain to address the destination port on a per-packet basis. Scalability and energy efficiency of the considered architecture are quantified and compared to typical single-plane architectures. Not only can the SW multiplane architecture achieve higher throughput by exploiting two switching domains, but its performance is shown to be highly scalable with network utilization. More importantly, higher performance is reached with an energy efficiency superior to single-plane architectures. The excellent energy efficiency is achieved using optical devices with low idle power.

A Scalable Space–Time Multi-plane Optical Interconnection Network Using Energy-Efficient Enabling Technologies [Invited]

This paper presents an energy-efficient multi-plane optical interconnection network to interconnect servers in a data center. The novel architecture uses the time domain to individually address each port within a card and the space domain to address each card. Optical enabling technologies passively time-compress serial packets by exploiting the wavelength domain and perform a broadcast-and-select to a destination card with minimum power dissipation. Scalability of both the physical layer and the overall power dissipation of the architecture is shown to be enhanced with respect to the existing interconnection network architectures based on space and wavelength domains. The space-time network architecture is scalable up to 216 ports with space-switches exhibiting energy efficiency of the order of picojoules per bit, thanks to the self-enabled semiconductor-optical-amplifier-based space-switches.

Designing Energy-Efficient Data Center Networks Using Space-Time Optical Interconnection Architectures

This paper considers a space-time interconnection architecture (STIA) based on optical devices and proposes its introduction in data center networks. The power consumption of the STIA is modeled, accounting for the energy proportionality of the optical devices in the STIA. Using such a model, a STIA-based network is designed using three different topologies, tree, folded Clos, and flattened butterfly, and optimized for power efficiency. Results show that, for a fixed topology, small-size STIAs are an energy-efficient solution for data center networks and allow a power reduction of more than an order of magnitude with respect to the Ethernet-based network. The comparison for the same bisection bandwidth shows that folded Clos and flattened butterfly outperform tree, whose power consumption is strongly dependent on the oversubscription ratio selected.

Silicon-based all-optical multi microring network-on-chip

An optical multi microring network-on-chip (MMR NoC) is proposed and evaluated through numerical simulations. The network architecture consists of a central resonating microring with local microrings connected to the input/output ports. A mathematical model based on the transfer matrix method is used to assess the MMR NoC performance and to analyze the fabrication tolerances. Results show that the proposed architecture exhibits a limited coherent crosstalk with a bandwidth suitable for 10  Gb/s signals, and it is robust to coupling ratio variations and ring radii fabrication inaccuracies.

Characterization of the Communication Patterns of Scientific Applications on Blue Gene/P

This paper examines the communication characteristics of a collection of scientific applications selected from the LLNLs Sequoia suite of benchmarks and the ANLs workload. By using an instrumentation library built on top of MPI we collect and characterize the applicationss messaging behavior: the type of communication patterns and primitives used, the amount of time spent for communication, the message sizes, the total amount of data exchanged, and the impact of collective primitives, through communication matrices we visualize the actual communication patterns to highlight symmetries and other relevant peculiarities. Our analysis exposes several similarities between the applications -- namely the utilization of common low-dimensional stencils, and the use of a small set of collective primitives, in particular all-reduces with small vectors. Overall, our study provides a better understanding of the communication characteristics of several important scientific applications and benchmarks.

Heterogeneous Optical Space Switches for Scalable and Energy-Efficient Data Centers

Optical space switches are key elements for the next generation of switching fabrics in backbone routers, high performance computing systems, and large data processing and storage systems. A number of architectures and alternative options for gating elements have been proposed, assessed, and implemented for a limited port count. The challenge is to further enhance the scalability and energy efficiency of space switches to support future traffic loads. This paper proposes a heterogeneous implementation of the space switches based on two different types of gating elements, namely semiconductor optical amplifiers (SOA) and Mach-Zehnder Interferometers (MZI). With respect to the existing homogeneous implementations, a higher energy efficiency can be achieved by minimizing the number of SOAs, but crosstalk is introduced by MZI. To reduce the power consumption while still guaranteeing adequate physical layer performance, the design of both Spanke and multi-stage architectures is optimized by strategically placing the different gating and amplification elements, and a physical layer analysis is carried out to validate the performance. The proposed heterogeneous implementation is able to achieve power savings up to 10% and 50% in the Spanke and multi-stage Beneš architectures, respectively, with respect to SOA-based space-switch implementations. Moreover, an improvement of the physical layer performance is achievable in the Spanke architecture thanks to the different placement of the SOAs.

Energy-efficient switching in optical interconnection networks

Interconnection networks currently used in backbone routers, high performance computing systems, and large data storage systems are based on electronic switching fabrics. Due to the ever increasing amount of data to store and process, these solutions are approaching their fundamental limitations in terms of wiring density, throughput, and power consumption. Optical interconnection networks are a viable solution to cope with such compelling issues. However the architecture must be designed and optimized in order to achieve high throughput at low power consumption. The paper will overview different SOA-based implementations of the space switches required in both single plane and multi plane optical interconnection networks. The objective is to identify the most suitable implementation in terms of number of SOAs needed, scalability and power consumption, when considering todays available technologies.

Energy efficiency and scalability of multi-plane optical interconnection networks for computing platforms and data centers

Multi-plane optical interconnection networks are proposed as scalable architectures with optimized energy-efficiency for large computing platforms. The space-time architecture is found to have scalable energy consumption greater than both the space-wavelength architecture and single-plane architecture.

Heterogeneous space switches for power-efficient optical interconnection networks

Space switches are the fundamental elements in single plane and multi-plane optical interconnection networks. This paper proposes a heterogeneous implementation of optical space switches based on two gating elements, amplifiers such as semiconductor optical amplifier (SOA) and modulators such as Mach-Zehnder modulators. The heterogenous implementation has the advantage of being more power-efficient than a homogeneous implementation based only on SOAs and more scalable than a homogeneous implementation based only on modulator-based gates as the optical power loss can be compensated by the SOAs. The paper derives the requirements in number of gating elements for different space switch architectures (i.e., crossbar, Beneš, Spanke, Clos, Spanke-Benes, and hybrid Clos). Based on the such derivation, the scalability and the power-efficiency of the different non-blocking architectures with heterogeneous design are assessed, with the aim at identifying the most suitable architecture based on todays available optical technologies. Results show that power saving around 80% are achievable when using a heterogeneous implementation instead of a homogenous SOA-based implementation.