The rapid advancement of metaverse applications in wireless environments necessitates efficient resource management to enhance Quality of Experience (QoE). This paper presents a novel framework for optimizing wireless resource allocation within the metaverse to optimize QoE using convex optimization and matching theory. We formulate a QoE optimization problem considering packet error rate (PER) and immersive experience. Our problem also enables us to trade off between immersive experience and PER while computing QoE. The formulated problem is a mixed-integer non-linear programming (MINLP) problem, which is addressed through decomposition, convex optimization, matching theory, and block successive upper-bound minimization (BSUM). Specifically, for a solution, our proposed model integrates matching theory, BSUM, and convex optimization to optimize the association, transmit power allocation, and resource allocation. Finally, numerical results are provided.

Federated learning (FL) offers many benefits, such as better privacy preservation and less communication overhead for scenarios with frequent data generation. In FL, local models are trained on end-devices and then migrated to the network edge or cloud for global aggregation. This aggregated model is shared back with end-devices to further improve their local models. This iterative process continues until convergence is achieved. Although FL has many merits, it has many challenges. The prominent one is computing resource constraints. End-devices typically have fewer computing resources and are unable to learn well the local models. Therefore, split FL (SFL) was introduced to address this problem. However, enabling SFL is also challenging due to wireless resource constraints and uncertainties. We formulate a joint end-devices computing resources optimization, task-offloading, and resource allocation problem for SFL at the network edge. Our problem formulation has a mixed-integer non-linear programming problem nature and hard to solve due to the presence of both binary and continuous variables. We propose a double deep Q-network (DDDQN) and optimization-based solution. Finally, we validate the proposed method using extensive simulation results.

This paper introduces LiSWARM, a low-cost LiDAR system to detect and track individual drones in a large swarm. LiSWARM provides robust and precise localization and recognition of drones in 3D space, which is not possible with state-of-the-art drone tracking systems that rely on radio-frequency (RF), acoustic, or RGB image signatures. It includes (1) an efficient data processing pipeline to process the point clouds, (2) robust priority-aware clustering algorithms to isolate swarm data from the background, (3) a reliable neural network-based algorithm to recognize the drones, and (4) a technique to track the trajectory of every drone in the swarm. We develop the LiSWARM prototype and validate it through both in-lab and field experiments. Notably, we measure its performance during two drone light shows involving 150 and 500 drones and confirm that the system achieves up to 98% accuracy in recognizing drones and reliably tracking drone trajectories. To evaluate the scalability of LiSWARM, we conduct a thorough analysis to benchmark the system's performance with a swarm consisting of 15,000 drones. The results demonstrate the potential to leverage LiSWARM for other applications, such as battlefield operations, errant drone detection, and securing sensitive areas such as airports and prisons.

Interdisciplinary workflows are evolving to accom-modate the growing resource diversity and parallelism in modern computing systems. The necessity of integrating various components, including multi-scale simulations and Artificial Intelligence and Machine Learning (AI/ML) with ensemble methods, has made the workflows increasingly complex and challenging to manage using traditional High performance computing (HPC) infrastructure. Cloud computing provides capabilities such as container orchestration, automation, and elasticity to manage the growing heterogeneity, scale, and complexity of HPC systems and workflows. Converged computing, a growing movement that integrates HPC and cloud technologies into a seamless environ-ment, can provide a means to bridge the gap between needs and capabilities. In particular, ensemble-based HPC workflows can benefit from the potential efficiency improvements afforded by these capabilities. While MPI - based (Message Passing Interface) workflows have been demonstrated to scale using cloud-native orchestration in Kubernetes, there is little work on understanding the combined impact of autoscaling and elasticity on MPI-based workflows. In this study, we explore the cost and performance of elasticity applied to ensembles of MPI-based HPC simulations using the Flux Operator. We propose a workload-driven autoscaling algorithm that outperforms CPU utilization-based autoscaling for MPI-based ensembles. The efficiency gains afforded by elastic, autoscaled approaches for MPI - based ensembles are described. We demonstrate that the workload-driven algorithm can reduce ensemble completion time by up to$4.7\times$in comparison with CPU utilization-based autoscaling.

The growing necessity for enhanced processing capabilities in edge devices with limited resources has led us to develop effective methods for improving high-performance computing (HPC) applications. In this paper, we introduce LASP (Lightweight Autotuning of Scientific Application Parameters), a novel strategy designed to address the parameter search space challenge in edge devices. Our strategy employs a multi-armed bandit (MAB) technique focused on online exploration and exploitation. Notably, LASP takes a dynamic approach, adapting seamlessly to changing environments. We tested LASP with four HPC applications: Lulesh, Kripke, Clomp, and Hypre. Its lightweight nature makes it particularly well-suited for resource-constrained edge devices. By employing the MAB framework to efficiently navigate the search space, we achieved significant performance improvements while adhering to the stringent computational limits of edge devices. Our experimental results demonstrate the effectiveness of LASP in optimizing parameter search on edge devices.

Significant power consumption is one of the major challenges for current and future high-performance computing (HPC) systems. All the while, HPC systems generally remain power underutilized, making them a great candidate for applying power oversubscription to reclaim unused capacity. However, an oversubscribed HPC system may occasionally get overloaded. In this paper, we propose MPR (Market-based Power Reduction), a scalable market-based approach where users actively participate in reducing the HPC system’s power consumption to mitigate overloads. In MPR, HPC users bid to supply, in exchange for incentives, the resource reduction required for handling the overloads. Using several real-world trace-based simulations, we extensively evaluate MPR and show that, by participating in MPR, users always receive more rewards than the cost of performance loss. At the same time, the HPC manager enjoys orders of magnitude more resource gain than her incentive payoff to the users. We also demonstrate the real-world effectiveness of MPR on a prototype system.

Server-level power monitoring in data centers can significantly contribute to its efficient management. Nevertheless, due to the cost of a dedicated power meter for each server, most data center power management only focuses on UPS or cluster-level power monitoring. In this paper, we propose a low-cost novel power monitoring approach that uses only one sensor to extract power consumption information of all servers. We utilize the conducted electromagnetic interference (EMI) of server power supplies to measure their power consumption from non-intrusive single-point voltage measurements. We present a theoretical characterization of conducted EMI generation in server power supply and its propagation through the data center power network. Using a set of ten commercial-grade servers (six Dell PowerEdge and four Lenovo ThinkSystem), we demonstrate that our approach can estimate each server's power consumption with less than ~7% mean absolute error.

Cloud applications are increasingly moving away from monolithic services to agile microservices-based deployments. However, efficient resource management for microservices poses a significant hurdle due to the sheer number of loosely coupled and interacting components. The interdependencies between various microservices make existing cloud resource autoscaling techniques ineffective. Meanwhile, machine learning (ML) based approaches that try to capture the complex relationships in microservices require extensive training data and cause intentional SLO violations. Moreover, these ML-heavy approaches are slow in adapting to dynamically changing microservice operating environments. In this paper, we propose PEMA (Practical Efficient Microservice Autoscaling), a lightweight microservice resource manager that finds efficient resource allocation through opportunistic resource reduction. PEMA's lightweight design enables novel workload-aware and adaptive resource management. Using three prototype microservice implementations, we show that PEMA can find efficient resource allocation and save up to 33% resources compared to the commercial rule-based resource allocations.

Server-level power monitoring in data centers can significantly contribute to its efficient management. Nevertheless, due to the cost of a dedicated power meter for each server, most data center power management only focuses on UPS or cluster-level power monitoring. In this paper, we propose a low-cost novel power monitoring approach that uses only one sensor to extract power consumption information of all servers. We utilize the conducted electromagnetic interference of server power supplies to measure its power consumption from non-intrusive single-point voltage measurement. Using a pair of commercial grade Dell PowerEdge servers, we demonstrate that our approach can estimate each server's power consumption with ~3% mean absolute percentage error.

The widespread adoption of Internet of Things and latency-critical applications has fueled the burgeoning development of edge colocation data centers (a.k. a., edge colocation) — small-scale data centers in distributed locations. In an edge colocation, multiple entities/tenants house their own physical servers together, sharing the power and cooling infrastructures for cost efficiency and scalability. In this paper, we discover that the sharing of cooling systems also exposes edge colocations’ potential vulnerabilities to cooling load injection attacks (called thermal attacks) by an attacker which, if left at large, may create thermal emergencies and even trigger system outages. Importantly, thermal attacks can be launched by leveraging the emerging architecture of built-in batteries integrated with servers that can conceal the attacker’s actual server power (or cooling load). We consider both one-shot attacks (which aim at creating system outages) and repeated attacks (which aim at causing frequent thermal emergencies). For repeated attacks, we present a foresighted attack strategy which, using reinforcement learning, learns on the fly a good timing for attacks based on the battery state and benign tenants’ load. We also combine prototype experiments with simulations to validate our attacks and show that, for a small 8kW edge colocation, an attacker can potentially cause significant losses. Finally, we suggest effective countermeasures to the potential threat of thermal attacks.

With the rapid development of the Internet of Things (IoT), computational workloads are gradually moving toward the internet edge for low latency. Due to significant workload fluctuations, edge data centers built in distributed locations suffer from resource underutilization and requires capacity underprovisioning to avoid wasting capital investment. The workload fluctuations, however, also make edge data centers more suitable for battery-assisted power management to counter the performance impact due to underprovisioning. In particular, the workload fluctuations allow the battery to be frequently recharged and made available for temporary capacity boosts. But, using batteries can overload the data center cooling system which is designed with a matching capacity of the power system. In this paper, we design a novel power management solution, DeepPM, that exploits the UPS battery and cold air inside the edge data center as energy storage to boost the performance. DeepPM uses deep reinforcement learning (DRL) to learn the data center thermal behavior online in a model-free manner and uses it on-the-fly to determine power allocation for optimum latency performance without overheating the data center. Our evaluation shows that DeepPM can improve latency performance by more than 50% compared to a power capping baseline while the server inlet temperature remains within safe operating limits (e.g., 32°C).

The massive energy consumption of data centers worldwide has resulted in a large carbon footprint, raising serious concerns to sustainable IT initiatives and attracting a great amount of research attention. Nonetheless, the current efforts to date, despite encouraging, have been primarily centered around owner-operated data centers (e.g., Google data center), leaving out another major segment of data center industry-colocation data centers-much less explored. As a major hindrance to carbon efficiency desired by the operator, colocation suffers from “split incentive”: tenants may not be willing to manage their servers for carbon efficiency. In this paper, we aim at minimizing the carbon footprint of geo-distributed colocation data centers, while ensuring that the operator's cost meets a long-term budget constraint. We overcome the “split incentive” hurdle by devising a novel online carbon-aware incentive mechanism, called GreenColo, in which tenants voluntarily bid for energy reduction at self-determined prices and will receive financial rewards if their bids are accepted at runtime. Using trace based simulation we show that GreenColo results in a carbon footprint fairly close (23 versus 18 percent) to the optimal offline solution with future information, while being able to satisfy the colocation operator's long-term budget constraint. We demonstrate the effectiveness of GreenColo in practical scenarios via both simulation studies and scaled-down prototype experiments. Our results show that GreenColo can reduce the carbon footprint by up to 24 percent without incurring any additional cost for the colocation operator (compared to the no-incentive baseline case), while tenants receive financial rewards for “free” without violating service level agreement.

Maliciously-injected power load, a.k.a. power attack, has recently surfaced as a new egregious attack vector for dangerously compromising the data center availability. This paper focuses on the emerging threat of power attacks in a multi-tenant colocation data center, an important type of data center where multiple tenants house their own servers and share the power distribution system. Concretely, we discover a novel physical side channel --- a voltage side channel --- which leaks the benign tenants' power usage information at runtime and helps an attacker precisely time its power attacks. The key idea we exploit is that, due to the Ohm's Law, the high-frequency switching operation (40~100kHz) of the power factor correction circuit universally built in today's server power supply units creates voltage ripples in the data center power lines. Importantly, without overlapping the grid voltage in the frequency domain, the voltage ripple signals can be easily sensed by the attacker to track the benign tenants' runtime power usage and precisely time its power attacks. We evaluate the timing accuracy of the voltage side channel in a real data center prototype, demonstrating that the attacker can extract benign tenants' power pattern with a great accuracy (correlation coefficient = 0.90+) and utilize 64% of all the attack opportunities without launching attacks randomly or consecutively. Finally, we highlight a few possible defense strategies and extend our study to more complex three-phase power distribution systems used in large multi-tenant data centers.

Despite the common practice of oversubscription, power capacity is largely under-utilized in data centers. A significant factor driving this under-utilization is fluctuation of the aggregate power demand, resulting in unused “spot (power) capacity”. In this paper, we tap into spot capacity for improving power infrastructure utilization in multi-tenant data centers, an important but under-explored type of data center where multiple tenants house their own physical servers. We propose a novel market, called SpotDC, to allocate spot capacity to tenants on demand. Specifically, SpotDC extracts tenants' racklevel spot capacity demand through an elastic demand function, based on which the operator sets the market price for spot capacity allocation. We evaluate SpotDC using both testbed experiments and simulations, demonstrating that SpotDC improves power infrastructure utilization and creates a “win-win” situation: the data center operator increases its profit (by nearly 10%), while tenants improve their performance (by 1.2-1.8x on average compared to the no spot capacity case, yet at a marginal cost).

Installation of backup power supply plays a vital role in maintaining communication services which can save billions of dollars as well as human lives during natural disasters. Due to the higher capital and operational expense compared to public power, pooling and sharing the backup power supplies can be an economical solution since the backup power capacity can be sized based on the aggregate demand of co-located operators. However, how to pool and share the backup power at multi-operator cellular sites in a fair manner should be considered due to the limited capacity and high user demands. In this paper, we adopt the Nash Bargaining Solution (NBS) of a bargaining problem which can guarantee the fairness of backup power sharing and design a decentralized algorithm approach with limited information exchange among the operators. Our simulation demonstrates that the sharing the backup power reduces the average delay and requires less BS power consumption than the non-sharing approach, especially for high traffic load scenarios. In addition, we also extend the formulation with respect to admission control for very high traffic demand cases.

As the critical infrastructure for supporting Internet and cloud computing services, massive geo-distributed data centers are notorious for their huge electricity appetites and carbon footprints. Nonetheless, a lesser-known fact is that data centers are also “thirsty”: to operate data centers, millions of gallons of water are required for cooling and electricity production. The existing water-saving techniques primarily focus on improved “engineering” (e.g., upgrading to air economizer cooling, diverting recycled/sea water instead of potable water) and do not apply to all data centers due to high upfront capital costs and/or location restrictions. In this paper, we propose a software-based approach towards water conservation by exploiting the inherent spatio-temporal diversity of water efficiency across geo-distributed data centers. Specifically, we propose a batch job scheduling algorithm, called WACE (minimization of WAter, Carbon and Electricity cost), which dynamically adjusts geographic load balancing and resource provisioning to minimize the water consumption along with carbon emission and electricity cost while satisfying average delay performance requirement. WACE can be implemented online without foreseeing the far future information and yields a total cost (incorporating electricity cost, water consumption and carbon emission) that is provably close to the optimal algorithm with lookahead information. Finally, we validate WACE through a trace-based simulation study and show that WACE outperforms state-of-the-art benchmarks: 25 percent water saving while incurring an acceptable delay increase. We also extend WACE to joint scheduling of batch workloads and delay-sensitive interactive workloads for further water footprint reduction in geo-distributed data centers.

Keeping wireless base stations operating continually and providing uninterrupted communications services can save billions of dollars as well as human lives during natural disasters and/or electricity outages. Toward this end, wireless operators need to install backup power supplies whose capacity is sufficient to support their peak power demand, thus incurring a significant capital expense. Hence, pooling together backup power supplies and sharing it among co-located wireless operators can effectively reduce the capital expense, as the backup power capacity can be sized based on the aggregate demand of co-located operators instead of individual demand. Turning this vision into reality, however, faces a new challenge: how to fairly share the backup power supply? In this paper, we propose fair sharing of backup power supply by multiple wireless operators based on the Nash bargaining solution (NBS). In addition, we integrate our analysis with multiple time slots for emergency cases in which the study the backup energy sharing based on model predictive control and NBS subject to an energy capacity constraint regarding future service availability. Our simulations demonstrate that sharing backup power/energy improves the communications service quality with lower cost and consumes less base station power than the non-sharing approach.

The power capacity of multi-tenant data centers is typically oversubscribed in order to increase the utilization of expensive power infrastructure. This practice can create dangerous situations and compromise data center availability if the designed power capacity is exceeded. This paper demonstrates that current safeguards are vulnerable to well-timed power attacks launched by malicious tenants (i.e., attackers). Further, we demonstrate that there is a physical side channel --- a thermal side channel due to hot air recirculation --- that contains information about the benign tenants' runtime power usage and can enable a malicious tenant to time power attacks effectively. In particular, we design a state-augmented Kalman filter to extract this information from the side channel and guide an attacker to use its maximum power at moments that coincide with the benign tenants' high power demand, thus overloading the shared power capacity. Our experimental results show that an attacker can capture 54% of all attack opportunities, significantly compromising the data center availability. Finally, we discuss a set of possible defense strategies to safeguard the data center infrastructure against power attacks.

Spreading across many parts of the world and presently hard striking California, extended droughts could even potentially threaten reliable electricity production and local water supplies, both of which are critical for data center operation. While numerous efforts have been dedicated to reducing data centers' energy consumption, the enormity of data centers' water footprints is largely neglected and, if still left unchecked, may handicap service availability during droughts. In this paper, we propose a water-aware workload management algorithm, called WATCH (WATer-constrained workload sCHeduling in data centers), which caps data centers' long-term water consumption by exploiting spatio-temporal diversities of water efficiency and dynamically dispatching workloads among distributed data centers. We demonstrate the effectiveness of WATCH both analytically and empirically using simulations: based on only online information, WATCH can result in a provably-low operational cost while successfully capping water consumption under a desired level. Our results also show that WATCH can cut water consumption by 20 percent while only incurring a negligible cost increase even compared to state-of-the-art cost-minimizing but water-oblivious solution. Sensitivity studies are conducted to validate WATCH under various settings.

Power oversubscription in data centers may occasionally trigger an emergency when the aggregate power demand exceeds the capacity. Handling such an emergency requires a graceful power capping solution that minimizes the performance loss. In this paper, we study power capping in a multi-tenant data center where the operator supplies power to multiple tenants that manage their own servers. Unlike owner-operated data centers, the operator lacks control over tenants' servers. To address this challenge, we propose a novel market mechanism based on supply function bidding, called COOP, to financially incentivize and coordinate tenants' power reduction for minimizing total performance loss (quantified in performance cost) while satisfying multiple power capping constraints. We build a prototype to show that COOP is efficient in terms of minimizing the total performance cost, even compared to the ideal but infeasible case that assumes the operator has full control over tenants' servers. We also demonstrate that COOP is "win-win", increasing the operator's profit (through oversubscription) and reducing tenants' cost (through financial compensation for their power reduction during emergencies).

The growing environmental and sustainability concerns have made energy efficiency a pressing issue for data center operation. Governments, as well as various organizations, are urging data centers to cap the increasing energy consumption. Naturally, achieving long term energy capping involves deciding energy usage over a long timescale (without accurately foreseeing the far future) and hence, we call this process “energy budgeting”. In this paper, we introduce an online resource management solution, called eBud (energy Budgeting), for a virtualized data center. eBud determines the number of servers, resource allocation to virtual machines and corresponding workload distribution to minimize data center operational cost while satisfying a long term energy cap. We prove that eBud achieves a close-to-minimum cost compared to the optimal offline algorithm with future information, while bounding the potential violation of energy budget constraint, in an almost arbitrarily random environment. We also perform a trace-based simulation study to complement the performance analysis. The simulation results show that eBud reduces the cost by more than$\mathrm{16}$percent (compared to state-of-the-art prediction-based algorithm) while resulting in a zero energy budget deficit. We also perform an experimental study based on RUBiS, demonstrating that in a real life scenario, eBud can achieve energy capping with a negligible increase in operational cost.

Power-hungry data centers face an urgent pressure on reducing the energy cost. The existing efforts, despite being numerous, have primarily centered around owner-operated data centers (e.g., Google), leaving another critical data center segment - colocation data center (e.g., Equinix) which rents out physical space to multiple tenants for housing their own servers - much less explored. Colocations have a major barrier to achieve cost efficiency: server power management by individual tenants is uncoordinated. This paper proposes RECO (REward for COst reduction), which shifts tenants' power management from uncoordinated to coordinated, using financial reward as a lever. RECO pays (voluntarily participating) tenants for energy reduction such that the colocation operator's overall cost is minimized. RECO incorporates the time-varying operation environment (e.g., cooling efficiency, intermittent renewables), addresses the peak power demand charge, and also proactively learns tenants' unknown responses to the offered reward. RECO includes a new feedback-based online algorithm to optimize the reward without far future offline information. We evaluate RECO using both scaled-down prototype experiments and simulations. Our results show that RECO is "win-win" and can successfully reduce the colocation operator's overall cost, by up to 27% compared to the no-incentive baseline case. Further, tenants receive financial rewards (up to 15% of their colocation costs) for "free" without violating Service Level Agreements.

Demand response programs maintain transmission stability in power grid through reducing electricity use during peak period, making grid more efficient and robust. While numerous demand response programs are currently being deployed by utility companies, we focus on emergency demand response program, which is critical to ensure reliability during emergency situations. As a key participant in such program, we consider a critical type of data center: multi-tenant colocation data center (or colocation), where multiple tenants mange their own servers in shared space but typically lack incentives to reduce energy for demand response. To enable multi-tenant data center demand response, we propose a contract-based mechanism, called Contract-DR, which offers financial incentives to tenants to shed energy during emergency situations, reducing the usage of cost-ineffective and environmentally-unfriendly diesel generation. We conduct theoretical analysis to prove the optimality of Contract-DR and also validate it through a trace-based study.

Increasingly serious concerns about the IT carbon footprints have been pushing data center operators to cap their (brown) consumption. Naturally, achieving capping involves deciding the usage over a long timescale (without foreseeing the far future) and hence, we call this process energy budgeting. The specific goal of this paper is to study budgeting for virtualized data centers from an algorithmic perspective: we develop a provably-efficient online algorithm, called eBud (energy Budgeting), which determines server CPU speed and resource allocation to virtual machines for minimizing the data center operational cost while satisfying the long-term capping constraint in an online fashion. We rigorously prove that eBud achieves a close-to-minimum cost compared to the optimal offline algorithm with future information, while bounding the potential violation of budget constraint, in an almost arbitrarily random environment. We also perform a trace-based simulation study to complement the analysis. The simulation results are consistent with our theoretical analysis and show that eBud reduces the cost by more than 60% (compared to state-of-the-art prediction-based algorithm) while resulting in a zero budget deficit.