http://cloud.pubs.dbs.uni-leipzig.de/sites/cloud.pubs.dbs.uni-leipzig.de/files/Brandic2008CloudcomputingandemergingITplatformsVisionhypeand.pdf

Cloud computing and emerging IT platforms: Vision, hype, and reality for delivering computing as the 5th utility

Matrix Factorization Techniques for Recommender Systems

Large-scale Internet services require a computing infrastructure that can beappropriately described as a warehouse-sized computing system. The cost ofbuilding datacenter facilities capable of delivering a given power capacity tosuch a computer can rival the recurring energy consumption costs themselves.Therefore, there are strong economic incentives to operate facilities as closeas possible to maximum capacity, so that the non-recurring facility costs canbe best amortized. That is difficult to achieve in practice because ofuncertainties in equipment power ratings and because power consumption tends tovary significantly with the actual computing activity. Effective powerprovisioning strategies are needed to determine how much computing equipmentcan be safely and efficiently hosted within a given power budget.In this paper we present the aggregate power usage characteristics of largecollections of servers (up to 15 thousand) for different classes ofapplications over a period of approximately six months. Those observationsallow us to evaluate opportunities for maximizing the use of the deployed powercapacity of datacenters, and assess the risks of over-subscribing it. We findthat even in well-tuned applications there is a noticeable gap (7 - 16%)between achieved and theoretical aggregate peak power usage at the clusterlevel (thousands of servers). The gap grows to almost 40% in wholedatacenters. This headroom can be used to deploy additional compute equipmentwithin the same power budget with minimal risk of exceeding it. We use ourmodeling framework to estimate the potential of power management schemes toreduce peak power and energy usage. We find that the opportunities for powerand energy savings are significant, but greater at the cluster-level (thousandsof servers) than at the rack-level (tens). Finally we argue that systems needto be power efficient across the activity range, and not only at peakperformance levels.

/pdf/power-provisioning-for-a-warehouse-sized-computer-3iyo1da2iy.pdf

Power provisioning for a warehouse-sized computer

Usually, a proof of a theorem contains more knowledge than the mere fact that the theorem is true. For instance, to prove that a graph is Hamiltonian it suffices to exhibit a Hamiltonian tour in it; however, this seems to contain more knowledge than the single bit Hamiltonian/non-Hamiltonian.In this paper a computational complexity theory of the “knowledge” contained in a proof is developed. Zero-knowledge proofs are defined as those proofs that convey no additional knowledge other than the correctness of the proposition in question. Examples of zero-knowledge proof systems are given for the languages of quadratic residuosity and 'quadratic nonresiduosity. These are the first examples of zero-knowledge proofs for languages not known to be efficiently recognizable.

/pdf/the-knowledge-complexity-of-interactive-proof-systems-31apre0ecf.pdf

The knowledge complexity of interactive proof-systems

Cloud computing systems fundamentally provide access to large pools of data and computational resources through a variety of interfaces similar in spirit to existing grid and HPC resource management and programming systems.  These types of systems offer a new programming target for scalable application developers and have gained popularity over the past few years.  However, most cloud computing systems in operation today are proprietary, rely upon infrastructure that is invisible to the research community, or are not explicitly designed to be instrumented and modified by systems researchers. In this work, we present Eucalyptus -- an open-source software framework for cloud computing that implements what is commonly referred to as Infrastructure as a Service (IaaS); systems that give users the ability to run and control entire virtual machine instances deployed across a variety physical resources. We outline the basic principles of the Eucalyptus design, detail important operational aspects of the system, and discuss architectural trade-offs that we have made in order to allow Eucalyptus to be portable, modular and simple to use on infrastructure commonly found within academic settings.  Finally, we provide evidence that Eucalyptus enables users familiar with existing Grid and HPC systems to explore new cloud computing functionality while maintaining access to existing, familiar application development software and Grid middle-ware.

/pdf/the-eucalyptus-open-source-cloud-computing-system-2ahud2ixme.pdf

The Eucalyptus Open-Source Cloud-Computing System

Battery-based energy storage has emerged as an enabling technology for a variety of grid energy optimizations, such as peak shaving and cost arbitrage. A key component of battery-driven peak shaving optimizations is peak forecasting, which predicts the hours of the day that see the greatest demand. While there has been significant prior work on load forecasting, we argue that the problem of predicting periods where the demand peaks for individual consumers or micro-grids is more challenging than forecasting load at a grid scale. We propose a new model for peak forecasting, based on deep learning, that predicts the k hours of each day with the highest and lowest demand. We evaluate our approach using a two year trace from a real micro-grid of 156 buildings and show that it outperforms the state of the art load forecasting techniques adapted for peak predictions by 11--32%. When used for battery-based peak shaving, our model yields annual savings of $496,320 for a 4 MWhr battery for this micro-grid.

https://dl.acm.org/doi/pdf/10.1145/3396851.3397751

Peak Forecasting for Battery-based Energy Optimizations in Campus Microgrids

Designing server clusters for intermittent power introduces new possibilities to make them cheaper, greener, and more reliable, including leveraging variable electricity prices to buy more power when it is cheap, increasing the use of clean renewable energy, and capping power at low levels to extend UPS lifetime during blackouts. However, regulating power usage to take advantage of these possibilities is challenging, since applications often access persistent distributed state, where power fluctuations impact I/O performance and data availability. To address the problem, we design and implement BlinkFS, which combines a blinking abstraction with a power-balanced data layout and popularity-based replication/reclamation to optimize I/O throughput and latency as power varies. Our experiments show that BlinkFS outperforms existing approaches, particularly at low steady power levels and high levels of intermittency. As one example of our results, we show that BlinkFS reduces completion time for MapReduce-style jobs by 42% at 50% full power compared to an existing energy-proportional DFS.

/pdf/a-distributed-file-system-for-intermittent-power-1j16b2nbhc.pdf

A distributed file system for intermittent power

Cloud platforms' rapid growth is raising significant concerns about their carbon emissions. To reduce carbon emissions, future cloud platforms will need to increase their reliance on renewable energy sources, such as solar and wind, which have zero emissions but are highly unreliable. Unfortunately, today's energy systems effectively mask this unreliability in hardware, which prevents applications from optimizing their carbon-efficiency, or work done per kilogram of carbon emitted. To address the problem, we design an "ecovisor", which virtualizes the energy system and exposes software-defined control of it to applications. An ecovisor enables each application to handle clean energy's unreliability in software based on its own specific requirements. We implement a small-scale ecovisor prototype that virtualizes a physical energy system to enable software-based application-level i) visibility into variable grid carbon-intensity and local renewable generation and ii) control of server power usage and battery charging and discharging. We evaluate the ecovisor approach by showing how multiple applications can concurrently exercise their virtual energy system in different ways to better optimize carbon-efficiency based on their specific requirements compared to general system-wide policies.

/pdf/ecovisor-a-virtual-energy-system-for-carbon-efficient-gngq61zx.pdf

Ecovisor: A Virtual Energy System for Carbon-Efficient Applications

In recent years, researchers have proposed numerous advanced load scheduling algorithms for smart homes with the goal of reducing the grid's peak power usage. In parallel, utilities have introduced variable rate pricing plans to incentivize residential consumers to shift more of their power usage to low-price, off-peak periods, also with the goal of reducing the grid's peak power usage. In this paper, we argue that variable rate pricing plans do not incentivize consumers to adopt advanced load scheduling algorithms. While beneficial to the grid, these algorithms do not significantly lower a consumer's electric bill. To address the problem, we propose flat-power pricing, which directly incentivizes consumers to flatten their own demand profile, rather than shift as much of their power usage as possible to low-cost, off-peak periods. Since most loads have only limited scheduling freedom, load scheduling algorithms often cannot shift much, if any, power usage to low-cost, off-peak periods, which are often many hours in the future. In contrast, flat-power pricing encourages consumers to shift power usage even over short time intervals to flatten demand. We evaluate the benefits of advanced load scheduling algorithms using flat-power pricing, showing that consumers save up to 40% on their electric bill, compared with 11% using an existing time-of-use rate plan.

/pdf/incentivizing-advanced-load-scheduling-in-smart-homes-353w611cea.pdf

Incentivizing Advanced Load Scheduling in Smart Homes

Edge computing has emerged as a popular paradigm for supporting mobile and IoT applications with low latency or high bandwidth needs. The attractiveness of edge computing has been further enhanced due to the recent availability of special-purpose hardware to accelerate specific compute tasks, such as deep learning inference, on edge nodes. In this paper, we experimentally compare the benefits and limitations of using specialized edge systems, built using edge accelerators, to more traditional forms of edge and cloud computing. Our experimental study using edge-based AI workloads shows that today's edge accelerators can provide comparable, and in many cases better, performance, when normalized for power or cost, than traditional edge and cloud servers. They also provide latency and bandwidth benefits for split processing, across and within tiers, when using model compression or model splitting, but require dynamic methods to determine the optimal split across tiers. We find that edge accelerators can support varying degrees of concurrency for multi-tenant inference applications, but lack isolation mechanisms necessary for edge cloud multi-tenant hosting.

David Irwin

Papers

Peak Forecasting for Battery-based Energy Optimizations in Campus Microgrids

A distributed file system for intermittent power

Ecovisor: A Virtual Energy System for Carbon-Efficient Applications

Incentivizing Advanced Load Scheduling in Smart Homes

AI on the Edge: Rethinking AI-based IoT Applications Using Specialized Edge Architectures