Showing papers by "Priya Mahadevan published in 2010"

ElasticTree: saving energy in data center networks

Abstract: Networks are a shared resource connecting critical IT infrastructure, and the general practice is to always leave them on. Yet, meaningful energy savings can result from improving a network's ability to scale up and down, as traffic demands ebb and flow. We present ElasticTree, a network-wide power1 manager, which dynamically adjusts the set of active network elements -- links and switches--to satisfy changing data center traffic loads.We first compare multiple strategies for finding minimum-power network subsets across a range of traffic patterns. We implement and analyze ElasticTree on a prototype testbed built with production OpenFlow switches from three network vendors. Further, we examine the trade-offs between energy efficiency, performance and robustness, with real traces from a production e-commerce website. Our results demonstrate that for data center workloads, ElasticTree can save up to 50% of network energy, while maintaining the ability to handle traffic surges. Our fast heuristic for computing network subsets enables ElasticTree to scale to data centers containing thousands of nodes. We finish by showing how a network admin might configure ElasticTree to satisfy their needs for performance and fault tolerance, while minimizing their network power bill.

Energy proportionality of an enterprise network

Abstract: Energy efficiency is becoming increasingly important in the operation of networking infrastructure, especially in enterprise and data center networks. While strategies for lowering the energy consumption of network devices have been proposed, what is lacking is a comprehensive measurement study conducted across a large network (such as an enterprise), that monitors power usage as a function of traffic flowing through the network. We present a large power profile study that we conducted in an enterprise network, comprising of 90 live switches from various vendors. We first describe Urja, the system that we built, that collects required configurations from a wide variety of deployed switches and uses them to accurately predict the power consumed by individual devices and the network as a whole. Urja is vendor neutral, and relies on standard SNMP MIBs to gather the required configuration and traffic information. Further, based on available knobs in current devices, the analysis engine in Urja lists various configuration and rewiring changes that can be made to the devices in order to make the network more energy proportional. Urja has been deployed in an enterprise sub-network for about 4 months; through comprehensive analysis of the data collected over this period, we present various changes (in increasing order of cost and complexity) that network administrators can perform; in this segment of an enterprise network, we can save over 30% of the network energy through simple configuration and rewiring changes, and without any performance impact.

Deep sleep mode management for a network switch

Abstract: A switch (100) connected to a network is configured to have one or more components that can be placed in a deep sleep mode. The switch (100) includes a management circuit (150) that is configured to wake up the components in deep sleep mode. The management circuit (150) includes a port (151) that receives packets and a wake-up circuit (152) that determines whether a packet received via the port (151) is a magic packet including a unique ID for the port (151) or the switch (100). If the packet is the magic packet including the unique ID for the port (151) or the switch (100), the wake-up circuit (151) is configured to send a wake-up signal to components in the switch (100) to wake up from the deep sleep mode.

Reducing lifecycle energy use of network switches

Abstract: The networking infrastructure is becoming a growing portion of the energy footprint in data centers as well as information technology in general. In this paper, we evaluate energy management strategies for network switches. We begin by performing a lifecycle assessment of existing switches in a data center, and find that the use phase of the lifecycle dominates. We then parametrically examine various energy management techniques to reduce this operational footprint, and find that advances in operational energy efficiency may soon increase the share of environmental burden from switch manufacturing. We conclude by discussing how these findings may influence network design in the future.

On the Complexity of Power Minimization Schemes in Data Center Networks

Abstract: In this paper, we consider migration of virtual machines in a data center to minimize network power consumption. Network power is consumed when switches are turned on, and conserved when they are turned off; the optimization problem then is to site virtual machines within the data center to achieve connectivity and desired bandwidth while turning on as few switches as possible. Depending upon specifics of the permissible optimization, and topology considerations within the data center, this optimization problem can be easy, or hard. We fully taxonomize the suite of optimization problems in this general space, showing that the most complex problem (placement and routing of virtual machines in a topologically-rich data center network) is NP- hard. We offer a placement technique based on a classic VLSI placement algorithm, and demonstrate efficacy on a trace set derived from a production data center.

Systems and methods for network and server power management

Abstract: The present disclosure includes a system and method for managing network and server power. In an example of managing network and server power according to the present disclosure, routing network traffic is routed onto a number of core networks based on core network statistics, capacity requirements are determined based on core network statistics for the number of core networks and for a number of servers operating a number of virtual machines on the number of core networks, wherein the number of core networks include a number of core switches and a number of edge switches, and the capacity is set for the number of core switches based on the capacity requirements for the number of core networks and for the number of servers based on the capacity requirements for the number of servers.

System for controlling power consumption of a network

Abstract: A system for controlling power consumption of a network includes at least one terminal to receive a plurality of requests to route data from a plurality of data sources to a plurality of data sinks, where the data sources and the data sinks are connected to each other through a plurality of network nodes forming the network, and a network configuration unit. The network configuration unit includes a selection module configured to select a configuration of the network nodes that allows the network to have a lowest overall power consumption of the network among a plurality of configurations of the network, and an output module configured to output a plurality of instruction signals to the network nodes to perform the network configuration. A network path for transmitting a network flow is selected that does not allow the network flow to be split and flow through another network path.

Cloud Sustainability Dashboard

Abstract: Cloud computing is gaining in popularity [1]. At the same time, the global community is becoming increasingly conscious about sustainability [2]. While intuitively Cloud computing, due to resource consolidation and virtualization, economies of large scale, delivery on demand, etc., has the potential to be more sustainable than well tuned data centers, this is not guaranteed. To evaluate and understand Cloud sustainability, we propose the Cloud Sustainability Dashboard (CSD), which models and assesses the overall sustainability impact of services hosted in the Cloud. We employ our approach to empirically evaluate the sustainability of Open Cirrus, an open source Cloud computing environment.

Managing replacement of an existing component

Abstract: In a method of managing replacement of at least one existing component with at least one replacement component, in which the at least one replacement component comprises a different number of components than the at least one existing component, at least one cost associated with implementing the at least one existing component is obtained. The at least one cost includes an environmental cost and a utility cost associated with implementing the at least one existing component. An existing net benefit of continuing implementation of the at least one existing component is calculated based upon the obtained at least one cost. A new net benefit of at least one of procuring and implementing the at least one replacement component is calculated.