Iyswarya Narayanan

One meter to find them all: water network leak localization using a single flow meter

Leak localization is a major issue faced by water utilities worldwide. Leaks are ideally detected and localized by a network-wide metering infrastructure. However, in many utilities, in-network metering is minimally present at just the inlets of subnetworks called District Metering Area (DMA). We consider the problem of leak localization using data from a single flow meter placed at the inlet of a DMA. We use standard time-series based modeling to detect if a current meter reading is a leak or not, and if so, to estimate the excess flow. Conventional approaches use an a-priori fully calibrated hydraulic model to map the excess flow back to a set of candidate leak locations. However, obtaining an accurate hydraulic model is expensive and hence, beyond the reach of many water utilities. We present an alternate approach that exploits the network structure and static properties in a novel way. Specifically, we extend the use of centrality metrics to infrastructure domains and use these metrics to map from the excess leak flow to the candidate leak location(s). We evaluate our approach on benchmark water utility network topologies as well as on real data obtained from an European water utility. On benchmark topologies, the localization obtained by our method is comparable to that obtained from a complete hydraulic model. On a real-world network, we were able to localize two out of the three leaks whose data we had access to. Of these two cases, we find that the actual leak location was in the candidate set identified by our approach; further, the approach pruned as much as 78% of the DMA locations, indicating a high degree of localization.

Multi-Objective Optimization to Minimize Battery Degradation and Electricity Cost for Demand Response in Datacenters

This paper presents a Lithium-ion battery control framework to achieve minimum health degradation and electricity cost when batteries are used for datacenter demand response (DR). Demand response in datacenters refers to the adjustment of demand for grid electricity to minimize electricity cost. Utilizing batteries for demand response will reduce the electricity cost but might accelerate health degradation. This tradeoff makes battery control for demand response a multi-objective optimization problem. Current research focuses only on minimizing the cost of demand response and does not capture battery transient and degradation dynamics. We address this multi-objective optimization problem using a second-order equivalent circuit model and an empirical capacity fade model of Lithium-ion batteries. To the best of our knowledge, this is the first study to use a nonlinear Lithium-ion battery and health degradation model for health-aware optimal control in the context of datacenters. The optimization problem is solved using a differential evolution (DE) algorithm and repeated for different battery pack sizes. Simulation results furnish a Pareto front that makes it possible to examine tradeoffs between the two optimization objectives and size the battery pack accordingly.Copyright

Battery health-conscious online power management for stochastic datacenter demand response

This paper presents a stochastic control framework for optimizing datacenter power management. The paper focuses on datacenters employing lithium-ion batteries for demand response. The use of batteries for demand response can reduce electricity costs, at the expense of battery degradation. We minimize this degradation using a control policy that takes into account uncertainties in power demand. We perform this optimization using a second-order model of battery charge dynamics, coupled with a physics-based model of battery aging via solid-electrolyte interphase (SEI) growth. To the best of our knowledge, this is the first study that uses battery models capturing diffusion dynamics and nonlinear aging effects, together with a model of demand uncertainty, for datacenter energy management. We formulate this as a stochastic dynamic programming (SDP) problem, where uncertain power demand is modeled as a Markov chain. The resulting control policy keeps grid power within a predefined range while minimizing battery degradation.

A Stochastic Optimal Control Approach for Exploring Tradeoffs between Cost Savings and Battery Aging in Datacenter Demand Response

This brief paper optimizes power management for datacenters employing lithium-ion battery storage, with the specific goal of addressing the tradeoff between: 1) the cost saving achievable through the peak demand shaving and 2) the corresponding battery aging. To the best of the authors’ knowledge, this tradeoff has never been addressed using physics-based models of battery performance and degradation combined with stochastic models of datacenter demand. We build: 1) a Markov chain model of datacenter power demand; 2) a second-order model of battery diffusion/reaction dynamics; and 3) a physics-based model of battery aging via solid electrolyte interphase growth. Together, these models enable the solution of the battery health-conscious demand response problem via stochastic dynamic programming (SDP). A penalty function is used for enforcing a datacenter “power cap” within this SDP problem. By varying this power cap, we traverse the Pareto tradeoff between the cost savings due to demand response and battery health degradation.

Little Knowledge Isn’t Always Dangerous—Understanding Water Distribution Networks Using Centrality Metrics

Addressing nonrevenue water, a major issue for water utilities, requires identification of strategic metering locations using calibrated hydraulic models of the water network. However, calibrated hydraulic models use both static and dynamic network data and are often prohibitively expensive. We present an approach to understand water network operations that uses only the static information of the network. Specifically, we analyze water networks using augmented centrality measures. We use readily available static information about network elements (e.g., diameters of pipes) rather than calibrated dynamic information (e.g., roughness coefficients of pipes, demands at nodes), and model each network element appropriately for analysis using customized centrality measures. Our approach identifies: 1) pipes carrying higher flows; 2) nodes with higher delivery heads; and 3) pipes with higher failure impact. Each of the above helps in determining strategic instrumentation locations. We validate our analysis by comparison with fully calibrated hydraulic models for three benchmark topologies. Our experimental evaluation shows that centrality analysis yields results which have a match of more than 85% with those obtained using calibrated hydraulic models on benchmark networks without significant over-provisioning. We also present results from a real-life case study where our approach matched 78% with locations picked by experts.

Networking lessons: From computers to water

As an instance of using IT to green non-IT domains, we consider the question whether lessons from computer networking can be applied in water distribution networks to improve their energy footprint and/or efficiency. Our contributions in this work are: (i) we identify several areas where principles from computer networking can be used to better water distribution; (ii) we focus on a specific infrastructure enhancement problem caused by increasing demands on a water utility network and present solutions (similar to those used in computer networks) that optimize both operational expenditure and total cost of ownership. We validate our solutions through simulations and compare their efficacy against techniques that are traditionally used in enhancing water networks. Our results show that lessons from computer networks can help in enhancing water networks.

Efficient booster pump placement in water networks using graph theoretic principles

Municipal water delivery networks face increasing demands due to population growth. We focus on enhancing a water utilitys infrastructure to meet its growing demands in a cost effective manner. Specifically, we consider the problem of placing pressure boosting pumps in a water network such that the minimum required delivery pressure is maintained at all the consumption points in the network for a given demand and the pump placement costs are minimized. The cost could be either energy cost or total cost dollars. Iterative optimization strategies and evolutionary computation techniques are typically used for solving such enhancement problems. We take a different perspective exploiting the structure of the network using graph theoretic principles. For water networks with tree topologies, we determine the optimal pump placement in terms of energy costs. We find that this energy optimal solution need not always minimize the total cost of ownership (TCO) involved in the pump placement. Therefore, we propose heuristic methodologies that reduce the TCO involved in placing pressure boosting pumps in tree networks. For water networks with complex topologies involving loops, we use the best of our tree solutions to find the initial seeds for iterative search strategies such as genetic algorithms (GA) and successive linear programming (SLP). We use EPANET for hydraulic modeling and study the efficacy of the proposed solution in terms of the TCO. In real-world topologies we considered, our heuristic seeding improves the performance of GA and SLP by about 68 % and 26 % respectively.

Right-Sizing Geo-distributed Data Centers for Availability and Latency

We show cloud developers how to right size data center (DC) capacity for geo-distributed applications deployed on several multi-megawatt DCs, possibly also using many smaller edge DCs. Note that capacity considerations for a geo-distributed infrastructure do not decompose into individual DC capacity planning. When edge DCs are used, heterogeneous availability and costs affect the capacity split between the edge and core DCs. Non-uniform spatial distribution of clients and interdependence between latency and availability constraints make it non-trivial to provision the right capacity at each DC. We develop a geo-distributed capacity planning framework to capture the key factors that influence capacity, ranging from application demand patterns, latency and availability requirements, DC cost-availability trade-offs, and data replication overheads. We apply our framework to a realistic application and DC infrastructure setting to gather insights into how capacity should be provisioned and allocated across DCs for a representative set of requirements and costs.

Rain or Shine? — Making Sense of Cloudy Reliability Data

Cloud datacenters must ensure high availability for the hosted applications and failures can be the bane of datacenter operators. Understanding the what, when and why of failures can help tremendously to mitigate their occurrence and impact. Failures can, however, depend on numerous spatial and temporal factors spanning hardware, workloads, support facilities, and even the environment. One has to rely on failure data from the field to quantify the influence of these factors on failures. Towards this goal, we collect failures data along with many parameters that might influence failures from two large production datacenters with very diverse characteristics. We show that multiple factors simultaneously affect failures, and these factors may interact in non-trivial ways. This makes conventional approaches that study aggregate characteristics or single parameter influences, rather inaccurate. Instead, we build a multi-factor analysis framework to systematically identify influencing factors, quantify their relative impact, and help in more accurate decision making for failure mitigation. We demonstrate this approach for three important decisions: spare capacity provisioning, comparing the reliability of hardware for vendor selection, and quantifying flexibility in datacenter climate control for cost-reliability trade-offs.

Evaluating energy storage for a multitude of uses in the datacenter

Datacenters often are a power utilitys largest consumers, and are expected to participate in several power management scenarios with diverse characteristics in which Energy Storage Devices (ESDs) are expected to play important roles. Different ESD technologies exist, including little explored technologies such as flow batteries, that offer different performance characteristics in cost, size, and environmental impact. While prior works in datacenter ESD literature have considered one of usage aspect, technology, performance metric (typically cost), the whole three-dimensional space is little explored. Towards understanding this design space, this paper presents first such study towards joint characterization of ESD usages based on their provisioning and operating demands, under ideal and realistic ESD technologies, and quantify their impact on datacenter performance. We expect our work can help datacenter operators to characterize this three-dimensional space in a systematic manner, and make design decisions targeted towards cost-effective and environmental impact aware datacenter energy management.