Search arXiv⌕ Search
Home/Archive/arXiv

arXiv · 2604.09615

Calibrating Microgrid Simulations for Energy-Aware Computing Systems

Marvin Steinke
View original source ↗Read source document ↗Export metadata ↓

Abstract

The surge for computing resource demand is increasing global electricity consumption in data centers which is expected to exceed 1000 TWh by 2026, mainly attributable to adoption of new AI technologies. Carbon-aware computing strategies can mitigate their environmental impact by aligning power consumption with the production of low-carbon renewable energy, but they face challenges due to the scarcity of development environments. Existing solutions either rely on costly and complex physical system architectures that are difficult to integrate and maintain or on full simulations that, while more economical, often lack realism by ignoring system overheads, and real-time node power consumption and resource fluctuations. This thesis remediates these issues by proposing a self-calibrating energy-aware software testbed that uses the Software-in-the-Loop co-simulation framework Vessim to integrate renewable energy production simulators, while including real computing nodes. The application-level power consumption of these are first approximated by the Kepler framework and then calibrated within Vessim's microgrid simulation using an external socket power meter as a definitive measurement source on the system-level. The evaluation of the testbed with GPU and CPU intensive workloads reveal fairly accurate power approximation of the whole computing node by the Kepler framework, with an average regression coefficient of 1.01 and R^2 values of 0.95, though certain machine learning workloads showed higher deviation. The average static y-intercept of the regression line of ~5.23 W indicate inaccuracies in the idle power approximation. Calibration of dynamic per-process power consumption improved accuracy for GPU workloads by ~50%, while CPU workloads saw a modest improvement of ~3.5%.

Explore related subjects

cs.DC

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Marvin Steinke. 2026-03-14. Calibrating Microgrid Simulations for Energy-Aware Computing Systems. https://arxiv.org/abs/2604.09615

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

Weave: Fine-Grained Dynamic SM Scheduling in an MoE Megakernel for Compute-Communication Overlap

Mixture-of-Experts (MoE) inference under expert parallelism (EP) turns each MoE layer into a distributed computation with costly dispatch and combine communication. State-of-the-art systems reduce this cost through communication-computation overlap, splitting the GPU's SMs for communication and computation respectively. However, this approach still leaves GPU resources wasted along two dimensions. Spatially, the best SM split is determined by each layer's routing result and varies across layers and GPUs, so fixed policies mismatch the workload and waste either NVLink bandwidth or compute throughput. Temporally, complex MoE data dependencies introduce bubbles that leave SMs idle. We present Weave, to our knowledge the first MoE overlap system that performs fine-grained dynamic SM scheduling - deciding per layer and per GPU by routing results at runtime. Once routing completes, each layer's communication and computation volumes become known; Weave exploits this predictability through a lightweight cost model running inside the persistent megakernel: a spatial scheduler partitions SMs into communication workers and computation workers to match the communication/computation throughput ratio, and a temporal scheduler coordinates the two worker groups to minimize SM idleness. On 4x H100 SXM GPUs across six mainstream MoE models, Weave achieves a 2.89x geometric-mean MoE-layer speedup and a 1.33x geometric-mean end-to-end speedup over five state-of-the-art baselines.

cs.DC↗
arXiv2026

pytest-gpu-proof: Enabling Cloud-CPU Continuous Integration for GPU Code with Local GPU Attestation

GPU acceleration is now routine across robotics, but cloud-hosted GPU continuous integration (CI) runners are expensive, resulting in severe under-testing of GPU-accelerated code. We present pytest-gpu-proof, an open-source pytest plugin offering a practical middle ground. Tests can be run on a local machine, signed with a receipt of exactly what ran and what it produced, and integrated into standard CPU CI workflows (e.g., GitHub Actions). The tool is open source and on PyPI, and we are actively integrating it across our lab's software stack.

cs.DC↗
arXiv2026

When Fancy Eviction Fails: Rethinking Cache Replacement For LLM Prefix Reuse

Long-running LLM applications repeatedly send growing context, making prefix caching critical for reducing prefill cost. Yet prefix-cache behavior under agentic workloads remains poorly understood. We study production traces from two companies and evaluate 14 eviction algorithms across HBM-constrained and large memory-pool settings. Despite a large gap to Belady, sophisticated policies designed for traditional caches provide little benefit over LRU. The reason is structural: prefix reuse is dominated by the regular pacing of active sessions, making recency unusually predictive. Prefix caching nevertheless introduces new challenges, including heavy-tailed session footprints and highly variable miss costs as attention computation grows with sequence length. We introduce the compute-savings ratio and two offline oracles to quantify these effects. Our results show that effective prefix-cache management should retain recency as its foundation while selectively adding quick demotion for one-hit prefixes, compute-aware partial eviction for expensive misses, and capacity-dependent eviction granularity. We will release the traces and simulator to support future research.

cs.DC↗