Energy demand from cloud and edge computing is rising rapidly, with AI workloads further intensifying electricity use and associated carbon emissions. In hybrid edge-cloud settings, sustainability impact depends on time- and location-varying grid Carbon Intensity (CI), site Power Usage Effectiveness (PUE), and heterogeneous hardware characteristics. Existing carbon-aware work explores solutions such as temporal elasticity, spatio-temporal workload shifting, and carbon-aware placement across distributed sites. However, these solutions do not provide a consistent and reproducible workflow for evaluating sustainability-aware scheduling policies on heterogeneous, federated edge-cloud topologies. We present EcoKube: a configurable simulation framework for the reproducible evaluation of sustainability-aware scheduling policies in heterogeneous edge-cloud environments. The framework includes an event-driven deterministic simulator, policy hooks, and a heterogeneity-aware reference policy. We evaluate the framework with synthetic batch workloads, comparing the reference policy against the default Kubernetes scheduler, KEIDS, and TOPSIS/KCSS. The contribution is architectural and experimental: EcoKube provides a reproducible way to compare sustainability-aware policies before deployment.

Large Language Models (LLMs) are increasingly deployed in scenarios demanding ultra-long context reasoning, such as agentic workflows and deep research understanding. However, long-context inference is constrained by the KV cache, a transient memory structure that grows linearly with sequence length and batch size, quickly dominating GPU memory usage. Existing memory reduction techniques, including eviction and quantization, often rely on static heuristics and suffer from degraded quality under tight budgets. In this paper, we propose ARKV, a lightweight and adaptive framework that dynamically allocates precision levels to cached tokens based on per-layer attention dynamics and token-level importance. During a short prefill phase, ARKV estimates the original quantization (OQ) ratio of each layer by computing statistical scores such as attention entropy, variance and kurtosis. During decoding, tokens are assigned to one of three states, Original (full precision), Quantization (low precision), or Eviction, according to a fast heavy-hitter scoring strategy. Our experiments on LLaMA3 and Qwen3 models across diverse long- and short-context tasks demonstrate that ARKV preserves 97

Large language models (LLMs) demonstrate remarkable capabilities, but their broad deployment is limited by significant computational resource demands, particularly energy consumption during inference. Static, one-model-fits-all inference strategies are often inefficient, as they do not exploit the diverse range of available models or adapt to varying query requirements. This paper presents GreenServ, a dynamic, context-aware routing framework that optimizes the trade-off between inference accuracy and energy efficiency. GreenServ extracts lightweight contextual features from each query, including task type, semantic cluster, and text complexity, and routes queries to the most suitable model from a heterogeneous pool, based on observed accuracy and energy usage. We employ a multi-armed bandit approach to learn adaptive routing policies online. This approach operates under partial feedback, eliminates the need for extensive offline calibration, and streamlines the integration of new models into the inference pipeline. We evaluated GreenServ across five benchmark tasks and a pool of 16 contemporary open-access LLMs. Experimental results show that GreenServ consistently outperforms static (single-model) and random baselines. In particular, compared to random routing, GreenServ achieved a 22% increase in accuracy while reducing cumulative energy consumption by 31%. Finally, we evaluated GreenServ with RouterBench, achieving an average accuracy of 71.7% with a peak accuracy of 75.7%. All artifacts are open-source and available as an anonymous repository for review purposes here: https://anonymous.4open.science/r/llm-inference-router-EBEA/README.md

Edge computing provides resources for IoT workloads at the network edge. Monitoring systems are vital for efficiently managing resources and application workloads by collecting, storing, and providing relevant information about the state of the resources. However, traditional monitoring systems have a centralized architecture for both data plane and control plane, which increases latency, creates a failure bottleneck, and faces challenges in providing quick and trustworthy data in volatile edge environments, especially where infrastructures are often built upon failure-prone, unsophisticated computing and network resources. Thus, we propose DEMon, a decentralized, self-adaptive monitoring system for edge. DEMon leverages the stochastic gossip communication protocol at its core. It develops efficient protocols for information dissemination, communication, and retrieval, avoiding a single point of failure and ensuring fast and trustworthy data access. Its decentralized control enables self-adaptive management of monitoring parameters, addressing the trade-offs between the quality of service of monitoring and resource consumption. We implement the proposed system as a lightweight and portable container-based system and evaluate it through experiments. We also present a use case demonstrating its feasibility. The results show that DEMon efficiently disseminates and retrieves the monitoring information, addressing the challenges of edge monitoring.

Proliferation of cloud-based latency-sensitive workloads requires infrastructures tuned to their workload-specific latency constraints. Today, they shape the cloud from a generalized computing platform to diverse workload-specific cloud environments. As the demand for latency-sensitive workloads increases, cloud service providers continue to scale their infrastructure, adversely increasing the carbon footprint and challenging climate-crisis-driven net-zero emission goals. Due to performance-oriented rigid deployment patterns of latency-optimizations, reducing its carbon footprint is challenging. Therefore, efficient techniques that exploit application specific opportunities are needed in that. To this end, we present a detailed taxonomy of recent literature on carbon-aware resource management in latency-sensitive cloud computing environments. Using the taxonomy, we analyze existing works discussing their optimization aspects, identify the gaps, and highlight future research directions.

LLM inference exhibits substantial variability across queries and execution phases, yet inference configurations are often applied uniformly. We present a measurement-driven characterization of workload heterogeneity and energy-performance behavior of LLM inference under GPU dynamic voltage and frequency scaling (DVFS). We evaluate five decoder-only LLMs (1B-32B parameters) across four NLP benchmarks using a controlled offline setup. We show that lightweight semantic features predict inference difficulty better than input length, with 44.5% of queries achieving comparable quality across model sizes. At the hardware level, the decode phase dominates inference time (77-91%) and is largely insensitive to GPU frequency. Consequently, reducing GPU frequency from 2842 MHz to 180 MHz achieves an average of 42% energy savings with only a 1-6% latency increase. We further provide a use case with an upper-bound analysis of the potential benefits of combining workload-aware model selection with phase-aware DVFS, motivating future energyefficient LLM inference systems.

Load shifting across geographic regions to chase intermittent renewable energy availability is commonly used in reducing cloud infrastructure carbon footprint. However, it often omits low-latency applications due to high latency variances of wide area networks (WAN) that interconnect regions. This paper addresses accommodating low-latency applications into load shifting by minimizing their shifting across the WAN. We propose a technique using a hardware-software co-design approach. At the hardware level, we conduct server load matching over renewables supply peaks and valleys by deep idling physical cores in two otherwise identical server pools, with one enabling simultaneous multi-threading (SMT) in CPUs. In return, we achieve a static set of logical cores amidst energy supply dynamics, reducing the probability of workload shifting. At the software level, we efficiently chase the static set of cores for low-latency applications within regions while prioritizing best-effort applications to accommodate shifting requirements across WANs. Our approach exploits the lower performance compromise of SMT cores due to their hardware multi-threading. We implement the proposed technique with OpenStack and CPU idle states and evaluate its performance on a real experimental testbed with Azure VM arrival traces. Results show an 80

Mobile devices offload latency-sensitive application tasks to edge servers to satisfy applications' Quality of Service (QoS) deadlines. Consequently, ensuring reliable offloading without QoS violations is challenging in distributed and unreliable edge environments with diverse resource and reliability levels. We propose FRESCO, a fast and reliable edge offloading framework that utilizes a blockchain-based reputation system, which enhances the reliability of offloading in the distributed edge. The distributed reputation system tracks the historical performance of edge servers, while blockchain through a consensus mechanism ensures that sensitive reputation information is secured against tampering. However, blockchain consensus typically has high latency, and therefore we employ a Hybrid Smart Contract (HSC) as a reputation state manager that automatically computes and stores reputation securely on-chain (i.e., on the blockchain) while allowing fast offloading decisions off-chain (i.e., outside of blockchain). The offloading decision engine uses a reputation score from HSC to derive fast offloading decisions, which are based on Satisfiability Modulo Theory (SMT). The SMT can formally guarantee a feasible solution that is valuable for latency-sensitive applications that require high reliability. With a combination of an on-chain HSC reputation state manager and an off-chain SMT decision engine, FRESCO offloads tasks to reliable servers without being hindered by blockchain consensus. In our experiment, FRESCO reduces response time by up to 7.86 times and saves energy by up to 5.4% compared to all baselines while minimizing QoS violations to 0.4% and achieving an average decision time of just 5.05 milliseconds.

LLM inference exhibits substantial variability across queries and execution phases, yet inference configurations are often applied uniformly. We present a measurement-driven characterization of workload heterogeneity and energy-performance behavior of LLM inference under GPU dynamic voltage and frequency scaling (DVFS). We evaluate five decoder-only LLMs (1B-32B parameters) across four NLP benchmarks using a controlled offline setup. We show that lightweight semantic features predict inference difficulty better than input length, with 44.5

Broad adoption of Large Language Models (LLM) demands rapid expansions of cloud LLM inference clusters, leading to the accumulation of embodied carbon-the emissions from manufacturing and supplying IT assets-that mostly concentrate on inference server CPU. This paper delves into the challenges of sustainable growth of cloud LLM inference, emphasizing extended amortization of CPU embodied over an increased lifespan. Given the reliability risks of silicon aging, we propose an aging-aware CPU core management technique to delay CPU aging effects, allowing the cluster operator to safely increase CPU life. Our technique exploits CPU underutilization patterns that we uncover in cloud LLM inference by halting aging in unused cores and even-outing aging in active cores via selective deep idling and aging-aware inference task allocation. Through extensive simulations using real-world Azure inference traces and an extended LLM cluster simulator from Microsoft, we show superior performance of our technique over existing methods with an estimated 37.67% reduction in yearly embodied carbon emissions through p99 performance of managing CPU aging effects, a 77% reduction in CPU underutilization, and less than 10% impact to the inference service quality.

Large Language Models (LLMs) are becoming integral to daily life, showcasing their vast potential across various Natural Language Processing (NLP) tasks. Beyond NLP, LLMs are increasingly used in software development tasks, such as code completion, modification, bug fixing, and code translation. Software engineers widely use tools like GitHub Copilot and Amazon Q, streamlining workflows and automating tasks with high accuracy. While the resource and energy intensity of LLM training is often highlighted, inference can be even more resourceintensive over time, as it's a continuous process with a high number of invocations. Therefore, developing resource-efficient alternatives for LLM inference is crucial for sustainability. This work proposes GREEN-CODE, a framework for energy-aware code generation in LLMs. GREEN-CODE performs dynamic early exit during LLM inference. We train a Reinforcement Learning (RL) agent that learns to balance the trade-offs between accuracy, latency, and energy consumption. Our approach is evaluated on two open-source LLMs, Llama 3.2 3B and OPT 2.7 B, using the JavaCorpus and PY150 datasets. Results show that our method reduces the energy consumption between 2350 % on average for code generation tasks without significantly affecting accuracy.

Edge computing has emerged as a pivotal technology, offering significant advantages such as low latency, enhanced data security, and reduced reliance on centralized cloud infrastructure. These benefits are crucial for applications requiring real-time data processing or strict security measures. Despite these advantages, edge devices operating within edge clusters are often underutilized. This inefficiency is mainly due to the absence of a holistic performance profiling mechanism which can help dynamically adjust the desired system configuration for a given workload. Since edge computing environments involve a complex interplay between CPU frequency, power consumption, and application performance, a deeper understanding of these correlations is essential. By uncovering these relationships, it becomes possible to make informed decisions that enhance both computational efficiency and energy savings. To address this gap, this paper evaluates the power consumption and performance characteristics of a single processing node within an edge cluster using a synthetic microbenchmark by varying the workload size and CPU frequency. The results show how an optimal measure can lead to optimized usage of edge resources, given both performance and power consumption.

Cloud platforms commonly exploit workload temporal flexibility to reduce their carbon emissions. They suspend/resume workload execution for when and where the energy is greenest. However, increasingly prevalent delay-intolerant real-time workloads challenge this approach. To this end, we present a framework to harvest green renewable energy for real-time workloads in cloud systems. We use Renewables-driven cores in servers to dynamically switch CPU cores between real-time and low-power profiles, matching renewable energy availability. We then develop a VM Execution Model to guarantee that running VMs are allocated with cores in the real-time power profile. If such cores are insufficient, we conduct criticality-aware VM evictions as needed. Furthermore, we develop a VM Packing Algorithm to utilize available cores across the servers. We introduce the Green Cores concept in our algorithm to convert renewable energy usage into a server inventory attribute. Based on this, we jointly optimize for renewable energy utilization and reduction of VM eviction incidents. We implement a prototype of our framework in OpenStack as openstack-gc. Using an experimental openstack-gc cloud and a large-scale simulation testbed, we expose our framework to VMs running RTEval, a real-time evaluation program, and a 14-day Azure VM arrival trace. Our results show: i) a 6.52x reduction in coefficient of variation of real-time latency over an existing workload temporal flexibility-based solution, and ii) a joint 79.64% reduction in eviction incidents with a 34.83% increase in energy harvest over the state-of-the-art packing algorithms. We open source openstack-gc at https://github.com/tharindu-b-hewage/openstack-gc.

Federated Learning (FL) provides a privacy-preserving mechanism for distributed training of machine learning models on networked devices (e.g., mobile devices, IoT edge nodes). It enables Artificial Intelligence (AI) at the edge by creating models without sharing actual data across the network. Existing research typically focuses on generic aspects of non-IID data and heterogeneity in client's system characteristics, but they often neglect the issue of insufficient data for model development, which can arise from uneven class label distribution and highly variable data volumes across edge nodes. In this work, we propose FLIGAN, a novel approach to address the issue of data incompleteness in FL. First, we leverage Generative Adversarial Networks (GANs) to adeptly capture complex data distributions and generate synthetic data that closely resemble real-world data. Then, we use synthetic data to enhance the robustness and completeness of datasets across nodes. Our methodology adheres to FL's privacy requirements by generating synthetic data in a federated manner without sharing the actual data in the process. We incorporate techniques such as classwise sampling and node grouping, designed to improve the federated GAN's performance, enabling the creation of high-quality synthetic datasets and facilitating efficient FL training. Empirical results from our experiments demonstrate that FLIGAN significantly improves model accuracy, especially in scenarios with high class imbalances, achieving up to a 20% increase in model accuracy over traditional FL baselines.

Cloud computing has become a critical infrastructure for modern society, like electric power grids and roads. As the backbone of the modern economy, it offers subscription-based computing services anytime, anywhere, on a pay-as-you-go basis. Its use is growing exponentially with the continued development of new classes of applications driven by a huge number of emerging networked devices. However, the success of Cloud computing has created a new global energy challenge, as it comes at the cost of vast energy usage. Currently, data centres hosting Cloud services world-wide consume more energy than most countries. Globally, by 2025, they are projected to consume 20% of global electricity and emit up to 5.5% of the world's carbon emissions. In addition, a significant part of the energy consumed is transformed into heat which leads to operational problems, including a reduction in system reliability and the life expectancy of devices, and escalation in cooling requirements. Therefore, for the future generations of Cloud computing to address the environmental and operational consequences of such significant energy usage, they must become energy-efficient and environmentally sustainable while continuing to deliver high-quality services. In this article, we propose a vision for learning-centric approach for the integrated management of new generation Cloud computing environments to reduce their energy consumption and carbon footprint while delivering service quality guarantees. In this article, we identify the dimensions and key issues of integrated resource management and our envisioned approaches to address them. We present a conceptual architecture for energy-efficient new generation Clouds and early results on the integrated management of resources and workloads that evidence its potential benefits towards energy efficiency and sustainability.

In recent years, Edge AI has become more prevalent with applications across various industries, from environmental monitoring to smart city management. Edge AI facilitates the processing of Internet of Things (IoT) data and provides privacy-enabled and latency-sensitive services to application users using Machine Learning (ML) algorithms, e.g., Time Series Classification (TSC). However, existing TSC algorithms require access to full raw data and demand substantial computing resources to train and use them effectively in runtime. This makes them impractical for deployment in resource-constrained Edge environments. To address this, in this paper, we propose an Adaptive Brownian Bridge-based Symbolic Aggregation Vector Space Model (ABBA-VSM). It is a new TSC model designed for classification services on Edge. Here, we first adaptively compress the raw time series into symbolic representations, thus capturing the changing trends of data. Subsequently, we train the classification model directly on these symbols. ABBA-VSM reduces communication data between IoT and Edge devices, as well as computation cycles, in the development of resource-efficient TSC services on Edge. We evaluate our solution with extensive experiments using datasets from the UCR time series classification archive. The results demonstrate that the ABBA-VSM achieves up to 80% compression ratio and 90-100% accuracy for binary classification. Whereas, for non-binary classification, it achieves an average compression ratio of 60% and accuracy ranging from 60-80%.

HPC datacenters offer a backbone to the modern digital society. Increasingly, they run Machine Learning (ML) jobs next to generic, compute-intensive workloads, supporting science, business, and other decision-making processes. However, understanding how ML jobs impact the operation of HPC datacenters, relative to generic jobs, remains desirable but understudied. In this work, we leverage long-term operational data, collected from a national-scale production HPC datacenter, and statistically compare how ML and generic jobs can impact the performance, failures, resource utilization, and energy consumption of HPC datacenters. Our study provides key insights, e.g., ML-related power usage causes GPU nodes to run into temperature limitations, median/mean runtime and failure rates are higher for ML jobs than for generic jobs, both ML and generic jobs exhibit highly variable arrival processes and resource demands, significant amounts of energy are spent on unsuccessfully terminating jobs, and concurrent jobs tend to terminate in the same state. We open-source our cleaned-up data traces on Zenodo (https://doi. org/10.5281/zenodo.13685426), and provide our analysis toolkit as software hosted on GitHub (https://github.com/atlarge-research/2024-icpads-hpc-workload-characterization). This study offers multiple benefits for data center administrators, who can improve operational efficiency, and for researchers, who can further improve system designs, scheduling techniques, etc.

Cloud computing environment simulators enable cost-effective experimentation of novel infrastructure designs and management approaches by avoiding significant costs incurred from repetitive deployments in real Cloud platforms. However, widely used Cloud environment simulators compromise on usability due to complexities in design and configuration, along with the added overhead of programming language expertise. Existing approaches attempting to reduce this overhead, such as script-based simulators and graphical user interface (GUI) based simulators, often compromise on the extensibility of the simulator. Simulator extensibility allows for customization at a fine-grained level, thus reducing it significantly affects flexibility in creating simulations. To address these challenges, we propose an architectural framework to enable human-readable script-based simulations in existing Cloud environment simulators while minimizing the impact on simulator extensibility. We implement the proposed framework for the widely used Cloud environment simulator, the CloudSim toolkit, and compare it against state-of-the-art baselines using a practical use case. The resulting framework, called CloudSim Express, achieves extensible simulations while surpassing baselines with over a 71.43$$ 71.43 $$% reduction in code complexity and an 89.42% reduction in lines of code.

Federated Learning (FL) provides a privacy-preserving mechanism for distributed training of machine learning models on networked devices (e.g., mobile devices, IoT edge nodes). It enables Artificial Intelligence (AI) at the edge by creating models without sharing actual data across the network. Existing research typically focuses on generic aspects of non-IID data and heterogeneity in client's system characteristics, but they often neglect the issue of insufficient data for model development, which can arise from uneven class label distribution and highly variable data volumes across edge nodes. In this work, we propose FLIGAN, a novel approach to address the issue of data incompleteness in FL. First, we leverage Generative Adversarial Networks (GANs) to adeptly capture complex data distributions and generate synthetic data that closely resemble real-world data. Then, we use synthetic data to enhance the robustness and completeness of datasets across nodes. Our methodology adheres to FL's privacy requirements by generating synthetic data in a federated manner without sharing the actual data in the process. We incorporate techniques such as classwise sampling and node grouping, designed to improve the federated GAN's performance, enabling the creation of high-quality synthetic datasets and facilitating efficient FL training. Empirical results from our experiments demonstrate that FLIGAN significantly improves model accuracy, especially in scenarios with high class imbalances, achieving up to a 20% increase in model accuracy over traditional FL baselines.