Dylan A. P. Davis

Renewable Energy-Aware Manycast Overlays

Manycasting allows a single source to reach multiple destinations while providing flexibility in destination selection. Our goal in this paper is to improve the cost of the manycast drop at member node (MA-DMN) overlay algorithm in terms of energy consumption and associated greenhouse gas (GHG) emissions. To reduce the environmental impact, ideally, a large percentage of the network nodes along the transmission and the chosen destinations need to be green. We present a novel energy-conservative emission-aware variant of the MA-DMN algorithm. We then propose further modifications to increase the utilization of those destinations that are powered by renewable energy sources: manycast drop at greenest nodes (MA-DGN). The potential for emission reduction by those algorithms is two-fold: The data are transported in the most efficient way and processed at the greenest available data centers. We compare the approaches by simulating realistic quantities of dynamic traffic. We assume heterogeneously distributed and time-dependent availability of renewable energy sources to power nodes throughout the network. We find that the energy-source-aware algorithms lower both energy-consumption and GHG emissions at stable network performance levels, in some cases even lowers blocking rate.

Energy source-aware manycast overlay in WDM networks

Manycasting is an emerging communication paradigm which allows a single source to reach multiple destinations while providing flexibility in the selection of which destinations to connect with. Traditional wavelength division multiplexed (WDM) networks do not support the all-optical splitting of signals to multiple output ports as required by point-to-multipoint communication schemes. Previous work has proposed an overlay approach known as Manycasting with Drop at Member Node (MA-DMN) to provide manycast support as a logical overlay to basic point-to-point lightpath connections. This approach has been studied extensively and compared to alternative overlay models, and has emerged the obvious candidate for supporting manycast overlays. Throughout its evaluation though, MA-DMN has never been scrutinized in terms of its costs for energy consumption and associated greenhouse gas (GHG) emissions. In this work, we subject MA-DMN to these evaluations, while also proposing a new more energy-conservative emission-aware variant known as MA-DMN using Least Impact Trees (MA-DMN-LIT). We compare these two approaches by simulating realistic quantities of dynamic traffic, and uniformly distributing renewable energy sources to power nodes throughout the network. We find that MA-DMN-LIT reduces energy consumption over MA-DMN by 6-10% across the network, while also reducing CO2 emissions by as much as 27%. We further conclude that MA-DMN-LIT also provides lower connection blocking by not over-subscribing shorter paths in the network as its emission-blind counterpart does.

Critical resource multicast protection in data center networks

Resources in a network are imperfect, and equipment failure can have detrimental effects on data and transmission success rates. Attempts to improve the survivability of network communications when these failures occur focus primarily on protection against the common occurrence of link failures, while nodal failure has been largely overlooked. In data-critical infrastructures, such as cloud computing or data center networks, wherein the critical points of interest are at the nodes, a natural disaster or directed attack could have catastrophic consequences for the localized data. To overcome single points of failure, replicated multicast transmissions can be used to distribute copies of critical data to various geographically dispersed locations on the grid. We therefore explore strategies to protect against single critical node failures during multicast transmissions. We propose three novel multicast survivable heuristics and quantitatively analyze and then compare them to traditional multicast provisioning schemes through extensive simulation1.

Parallel circuit provisioning in ESnet's OSCARS

Large-scale science applications generate great volumes of data, which are frequently stored in remote data repositories or shared with cooperating laboratories across the network through the use of advance reservation connections. The groups that utilize these data transfers would benefit from having their applications simultaneously transmit data over multiple channels in parallel. Many of todays networks do not however provide the resource-aware scheduling to support this parallelization. As part of a framework for providing said applications with parallel resource-optimized provisioning of end-to-end requests, we propose and develop a scheduling enhancement to ESnets On-demand Secure Circuits and Advance Reservation System (OSCARS). This enhancement comes in the form of a front-end client, the behavior of which we quantitatively evaluate to compare the performance of parallel resource-provisioning to serial resource usage for both unicast and anycast scenarios.

Resource survivability for multicast in elastic optical networks

While the multicast paradigm offers tremendous benefits in efficiency for transmitting data across optical networks, the loss of a destination, or resource, node can result in both the loss of data at the node and the disconnection of an entire established request. We propose an integer linear programming based approach to optimally solve this problem in elastic optical networks, and show that the generalized problem of protecting against this failure is NP-hard. We find that protection against this type of failure can be provided with the trade-off of increased frequency slot consumption, compared to less-protected solutions.

Static protection against single multicast resource failure in optical WDM networks

The multicast paradigm offers tremendous benefits in efficiency for transmitting data across optical networks, allowing a single client to send information to an entire set of endpoints. A multicast request is most efficiently provisioned through the creation of a tree, with the endpoints, or resources, occasionally serving as branching points. This practice can lead to the source of the request becoming disconnected from the associated resources should one of those branching resources fail. In cases where a large amount of data are currently in transmission, the ramifications of this failure can be severe. We propose an optimal solution through integer linear programming for the static protected multicast routing and wavelength assignment problem, where an entire set of requests is provisioned with built-in redundancy against single resource node failure. We compare the optimal performance against several heuristics and find that protection against this type of failure can be provided with the trade-off of increased wavelength consumption, compared to less-protected solutions.

Static protection against single multicast resource failure

The multicast paradigm offers tremendous benefits in efficiency for transmitting data across optical networks, allowing a single client to send information to an entire set of endpoints. A multicast request is most efficiently provisioned through the creation of a tree, with the end points, or resources, occasionally serving as branching points. This practice can lead to the source of the request becoming disconnected from the associated resources should one of those branching resources fail. We propose an optimal solution through Integer Linear Programming (ILP) for the static protected multicast Routing and Wavelength Assignment (RWA) problem, where an entire set of requests are provisioned with built-in redundancy against single resource node failure. We compare the optimal performance against several heuristics, and find that protection against this type of failure can be provided without excessive wavelength consumption.

Parallel and survivable multipath circuit provisioning in ESnet's OSCARS

Data generation is approaching petascale and exascale rates by cutting-edge science and research applications varying from material informatics to physics. With data generation and management comes the necessity to transmit such vast collections of information across the world’s networks for processing, analysis, storage, or peer-sharing. This practice is becoming the norm to the large-scale scientific community, but complications can arise during networking. There are countless situations such as component failure due to a harmless construction accident or a devastating natural disaster that may lead to catastrophic interruption of service. Furthermore, given the size of datasets, there is a strong need to support intelligent and fast parallelism throughout the network to allow end users to efficiently consume available bandwidth. We therefore propose a multipath extension for ESnet’s On-demand Secure Circuits and Advance Reservation System (OSCARS), the network research community’s most popular long-lived circuit-provisioning software package. Presently, OSCARS supports purely point-to-point circuits; however, our proposed client software provides an overlay onto the default OSCARS path computation engine that enables end users to route their data along multiple link-disjoint paths to provide session survivability and increase the degree of parallelism. We have also adapted the proposed multipath extension to an existing anycast OSCARS deployment, which allows for the selection of one preferred destination node from among a set of potential candidates. Through thorough simulation analysis and exposure to realistic failure event distributions, we quantitatively evaluate the multipath client performance and showcase the relative benefits when compared to the standard single-path OSCARS deployment.

Electricity cost and emissions reduction in optical networks

The goal of energy cost-aware Routing and Wavelength Assignment (RWA) is to minimize the total electricity expenditure in an optical network. While effective, just aiming to reduce the electricity consumed does not necessarily mitigate the environmental impact. A new approach is required to reduce the emissions produced as a by-product of RWA. We present a method for doing so through the use of Mixed Integer Linear Programming, which can find the optimal solution that minimizes the electricity cost of RWA for a static set of requests. This objective is quantitatively compared to alternative goals, including directly minimizing the emissions produced, reducing the length of established paths, and balancing reductions in both emissions and electricity cost simultaneously.