ALT Ternary LLMs read only a few gigabytes per decoded token, yet on dual-socket servers with four NUMA nodes the fastest ternary runtime we found uses less than half of the memory bandwidth. Placing each weight on the node of the thread that reads it is necessary but not enough: in a decode engine with node-local weights that reuses ikₗlama.cpp's kernels, threads on two generations of Xeon servers still spend 20-48% of each token waiting at synchronization points. Per-thread barrier traces show that most of the wait is for the slowest threads of each node; the barriers' own latency (microseconds each, but up to 12% of the token for the smallest model) and serial per-node work are smaller parts. On identical kernels, a Megatron-style column split beats a row-owned split without cross-node reductions in every configuration we measured, and isolating the number of barriers shows that most of its gain comes from its threads waiting less for each other and for their node leaders, and only a mi

ALT A cluster admission gate can receive a capacity record that no longer matches current schedulability. We study one exogenous binary schedulability unit under a bound on the joint probability of admission and unavailability. For a symmetric Markov state, we derive the stationary admission frontier for a fixed observation delay and apply finite likelihood allocation to a delay-tagged delivered record. A one-sided exact binomial endpoint calibrates each fitted policy, and exact enumeration confirms the per-policy confidence target in every tested parameter cell. A separate deterministic guard charges every unresolved admission against the realized allowance and bounds unavailable-state admissions at every prefix under truthful outcome feedback. An executable admission procedure composes the allocation and guard. The controlled system binds source sampling, NATS delivery, randomized admission, Kubernetes objects, and API-derived outcomes in one append-only record. Independent audit finds n

ALT Sparse attention reduces the computational cost of long-sequence video Transformers, but three-dimensional token layouts often leave selected blocks partially occupied, wasting physical tile capacity. We present KPage, a block-sparse attention scheme that packs valid 16-row K/V pages across logical-block boundaries into large tiles directly during loading, without materializing compacted tensors in high-bandwidth memory. KPage also routes Q blocks between 64x256 and 128x128 tiles to reduce query-side padding. Exact masking and K/V-owned page packing preserve the original sparse graph in forward and backward execution. Across five representative workloads, KPage achieves 1.35x-2.06x bare forward-kernel speedup over NVIDIA Block-Sparse Attention (BSA), with a geometric mean of 1.77x. It is near parity for fully occupied blocks (1.03x) but achieves 2.29x geometric-mean speedup on 3D layouts with partial blocks. KPage reduces physical query-key pair capacity from 334.1G to 177.5G while pre

ALT Learned surrogates for power-flow (PF) computation are usually motivated by speed, compared against a single CPU-based baseline. We show this comparison is fragile: across 7 open-source CPU-based solvers evaluated on identical inputs, throughput varies by roughly 2.5-3 orders of magnitude, so the choice of baseline alone can flip the conclusion. We build a benchmark spanning 9 physics-based solvers: 7 CPU-based and 2 GPU-based, including a new batched GPU solver, gumsim2grid - and 3 published ML surrogates, evaluated under a shared, independently implemented KCL-mismatch check across 15 grids from 14 to 70,000 buses. On the PF task itself, a sufficiently parallelized CPU solver or gpusim2grid matches or exceeds the throughput of the ML surrogates we test while remaining orders of magnitude more accurate; we find no evidence that these GNN-style surrogates are yet competitive for PF specifically, though prior work reports genuine gains on related but more combinatorial tasks such as opt

ALT Efficient use of large-scale supercomputing systems depends on accurate runtime estimation and walltime selection. To avoid premature termination, users often overestimate walltime, leading to poor resource utilization and reduced scheduling efficiency. This paper presents a predictor-agnostic framework for safe walltime recommendation in production HPC environments. Rather than proposing a new runtime prediction model, the primary contribution is the SAFE decision layer, which converts predicted runtimes into conservative bounded walltime recommendations suitable for deployment. Runtime prediction is therefore treated as a modular component within the proposed framework. We evaluate the approach using large-scale production job histories from the Fugaku and NREL Eagle supercomputers, comprising approximately 24M and 11M jobs, respectively. Runtime prediction performance is first evaluated using representative predictors, including linear regression, neural networks, and a neural ensem

ALT Vehicular Edge Computing (VEC) reduces latency for computation-intensive vehicular applications by offloading tasks to roadside edge servers. Real vehicular workloads, however, consist of dependent subtasks with stringent reliability needs, while the cross-region collaborative learning used to adapt offloading policies must stay robust to heterogeneous data and secure over open wireless channels. We propose CRAFT, a Constrained Reinforcement-based Adaptive Federated Task Offloading framework that casts Directed Acyclic Graph (DAG)-structured offloading and resource allocation as a Constrained Markov Decision Process. CRAFT couples centralized-critic policy learning with Lagrangian relaxation to explicitly regulate the long-term task-completion constraint, and uses two-tier federated aggregation with proximal regularization to jointly mitigate non-Independent and Identically Distributed (non-IID) drift and stabilize constraint adaptation. Confidentiality and integrity for federated para

ALT Federated learning (FL) enables collaborative model training while preserving data privacy. However, non-independent and identically distributed (non-IID) data inevitably introduces classifier bias, leading to significant performance degradation. Existing studies primarily focus on calibrating or reconstructing the classifier after FL training to alleviate this bias. We find that the last-layer representations of the global model are often suboptimal for establishing a global decision boundary. Instead, classifiers constructed based on intermediate-layer representations can achieve superior performance. This finding provides a critical insight into overcoming the performance bottleneck inherent in existing paradigms. Nevertheless, identifying the optimal layer without auxiliary datasets or prior knowledge remains a formidable challenge. In this paper, we propose FLARE (Federated Cross-Layer Joint ClAssifier REconstruction) which rethinks classifier reconstruction from a cross-layer per

ALT Serverless computing is increasingly used to run complex DAG workflows across diverse applications, and individual function execution times have grown to support them, pushing workflow makespans to hours. Open-source serverless platforms, deployed on a pool of VMs, typically use on-demand instances, and their cost increases with the makespan. Spot instances are far cheaper, but a preemption is likely before the workflow finishes. A preemption at a fan-out node or a task feeding a fan-in node stalls every parallel branch, cascading a single preemption into a workflow-wide failure. Prior systems address this challenge by migrating high-impact tasks to reliable, non-spot resources, sacrificing spot's cost advantage. We present FaaSpot, a serverless workflow execution system that runs every workflow task on spot instances with no on-demand fallback. Our key insight is that cascade-prone positions are determined by the DAG's topology and can be identified before execution. Therefore, FaaSpo

ALT HPC facilities collect years of job scheduler records, yet much of this operational data remains underused because accessible tools for turning it into actionable insight are scarce. We present CQTwin, an open-source digital-twin toolkit for operational analytics of HPC scheduling data. CQTwin addresses this gap through two complementary modes: offline trace replay and online event-driven simulation. Built on the Python-based CQSim discrete-event simulator, CQTwin combines a fidelity-validated, high-performance C++ engine with an interactive browser-based dashboard and a Snapshot Lab, enabling facility staff, operators, and researchers to characterize workloads, compare scheduling policies, and explore what-if scenarios. As an offline twin, CQTwin replays archived workload traces under alternative policies to show how the same workload would have unfolded under different scheduling decisions. As an online twin, it consumes job events as they arrive and projects how alternative policies

ALT In high-performance computing (HPC), users must submit resource requirements for their jobs, with user-supplied walltime being a critical parameter. However, these estimates are notoriously inaccurate, as users tend to overestimate walltime to prevent the scheduler from terminating their jobs prematurely. While existing work on learning-based walltime prediction primarily focuses on maximizing prediction accuracy, such predictions are inherently risky for walltime recommendations. Specifically, underestimation leads to job termination before completion, resulting in lost progress and wasted resources. In this study, we present SafeWall, a novel recommendation method that adjusts walltime predictions by incorporating an explicit, nonnegative safety buffer to mitigate underestimation risk. The core of our approach is a risk-aware buffer that dynamically calibrates safety margins based on individual user's historical behavior. When user data is limited, the system employs statistical shri

ALT While quantum technologies are rapidly evolving, near-term applications heavily rely on hybrid quantum-classical paradigms. Among these, the classical shadows framework has gained a lot of traction, not only for quantum information science, but also for the development and verification of quantum hardware. It allows to efficiently classically estimate many properties of quantum states from randomized measurements alone, a task that is otherwise intractable due to the destructive nature of measurements in quantum systems. There are, however, many properties of interest that are, due to their very problem formulation, not efficiently estimable. In this work, we identify a computational kernel occurring in estimators of polynomial functions, a class of problems that are in general computationally difficult to solve. We show that this structure allows a row-wise partition that, together with a custom SIMD vectorization, enables massive parallelization and can reduce the computational cost

ALT Currently many Byzantine fault-tolerant (BFT) protocols use directed acyclic graphs (DAGs) to let replicas propose blocks of requests in parallel for improved throughput. Yet, the existing DAG-based designs mainly consider the atomic-broadcast layer: They typically do not describe an efficient client interaction layer, their garbage collection is either unspecified or relies on trusted external storage, and the lowest-latency designs rely on timing assumptions that may be violated in geo-distributed deployments. In this paper, we propose the Pangolin protocol: It introduces a novel low-latency DAG-based atomic broadcast layer and completes it into an efficient state-machine replication (SMR) framework. To preserve system throughput, clients issue requests to a single replica and verify replies efficiently via compact execution proofs. Furthermore, Pangolin introduces a new garbage-collection scheme for DAGs which bounds memory usage even under full asynchrony. Pangolin leverages truste

ALT Training foundation models at the frontier typically requires large-scale dedicated and centralized clusters with fast interconnects - infrastructure that is prohibitively expensive and generally accessible only to well-resourced organizations. Decentralized training over pooled, geographically distributed GPUs offers an alternative, but introduces substantial heterogeneity in compute, memory, power, and network performance. Existing distributed training systems are primarily designed for clusters with fast, low-latency interconnects and typically use all available GPUs when configuring parallel training. In geo-distributed environments, however, slower devices and weak network links can become bottlenecks, while optimizing solely for throughput overlooks significant energy cost. We present GeoTrain, an automatic parallelism system that jointly optimizes training performance and energy consumption for heterogeneous geo-distributed GPU environments. GeoTrain simultaneously selects GPUs,

ALT Large language models (LLM) are increasingly integrated into pervasive applications spanning heterogeneous cloud-edge resources. These heterogeneous resources differ in computational capacity, energy efficiency, network conditions, and deployed model quality, creating trade-offs among response quality, monetary cost, energy consumption, and batch completion time when routing inference requests. Existing LLM routers typically make per-query routing decisions based primarily on model quality and token cost, whereas cloud-edge schedulers often overlook quality differences among deployed models. Consequently, these objectives remain insufficiently coordinated. We present CEMA-Router, an infrastructure-aware batch router for LLM inference in pervasive cloud-edge systems. CEMA-Router formulates batch routing as a multi-objective search problem that jointly minimizes model-quality loss, monetary cost, energy consumption, and straggler-dominated batch completion time. To efficiently explore th

ALT Efficient resource management is a challenge in multi-tenant cloud environments as workloads compete for shared resources such as last-level cache and memory bandwidth, whose congestion can change dynamically as workload composition and resource demands evolve. Existing resource management approaches often rely on static resource requirements or prior workload profiling, limiting their ability to respond to changing resource contention as required in the cloud. In this paper, we leverage workload buoyancy, a runtime framework for capturing workload sensitivity to individual resources and application-level performance, to enable adaptive scheduling and resource management. We propose a holistic approach that incorporates buoyancy into workload scheduling, node-local resource allocation, and workload rescheduling throughout the workload lifecycle. We implement our approach as a Kubernetes-native prototype and evaluate it on a multi-node cluster using diverse cloud workloads under varying

ALT This paper studies a Deadline-constrained task Offloading and resource Allocation Problem (DOAP) in a backhaul-network-enabled Mobile Edge Computing (MEC) system with both bandwidth and computational resource constraints. The goal is to maximize the total inference accuracy improvement across all tasks. We formulate this problem, denoted as P, as an Integer Linear Programming (ILP) problem and show it is NP-hard via a reduction from the maximum-weight three-dimensional matching problem. To address P, we develop 𝙵𝚊𝚜𝚝𝚂𝙰, a linear-time ¹⁄₁₀-approximation algorithm based on the local-ratio technique. To the best of our knowledge, 𝙵𝚊𝚜𝚝𝚂𝙰 provides the first constant-factor performance guarantee for backhaul-network-enabled DOAPs. In addition, we introduce a fault detection and recovery framework that enables robust MEC operation under backhaul link and edge server failures. Experiments using profiled image classification workloads and real-world taxi traces show that 𝙵𝚊𝚜𝚝𝚂𝙰

ALT Flash floods in Hanoi submerge roads within minutes of heavy rain onset, yet public flood advisories remain retrospective. Hydrodynamic models could supply forward-looking depth guidance, but they are typically run offline for historical events, and prior studies have not evaluated whether physics-based simulation, spatial serving, and vehicle-aware routing can operate together within a fixed refresh cycle using a cloud solver and a single commodity server. This paper evaluates an end-to-end prototype of an automated flood digital twin designed to meet a 15-minute refresh cycle under data scarcity. The architecture organizes execution into decoupled, event-driven workflows that chain weather-gated 3Di simulations through hydraulic restart states, serve flood depths through a dual-track spatial layer, and publish road-edge depths via standardized output contracts. Evaluated on a 57-run storm replay in the My Dinh and Tay Mo sector of Hanoi driven by satellite rainfall estimates, every c

ALT Real-time streaming video question answering with vision-language models must answer queries that arrive at any time with low latency while sustaining the video-processing rate required by the application. Existing systems construct key-value (KV) caches for the entire video as it arrives, although a query typically uses only a few of the preceding frames as evidence. This highlights a fundamental trade-off: performing more computation as video arrives reduces query-time computation but can increase processing lag, whereas deferring computation can increase the video-processing rate but requires additional query-time computation for retrieved content. We present StreamAhead, an algorithm-system co-design that exploits this trade-off through adaptive precomputation: dynamically controlling how much computation to complete for each video segment before a query arrives. StreamAhead encodes video segments independently and lets each remain at a different representation level, from inexpens

ALT Optical Circuit Switches (OCS) make inter-pod wiring in AI clusters programmable, posing design questions on wiring choice, symmetry departure, and reconfigurations. We present a unified information-geometric framework to address these issues.A wiring is a point in a probability simplex over circuits within a degree polytope. Departure from a symmetric Fat-Tree prior q is measured by DKL(⋅ ‖ q). The cost U is the smoothed DAG critical-path length, whose gradient equals path marginals of a Gibbs distribution via forward–backward recursion.For free energy Fβ₁ = U + β₁⁻¹ DKL(⋅ ‖ q), we prove well-posedness, monotonic dependence on β₁, adaptation bounds via Jeffreys divergence, and a symmetry principle for static optimality. An entropic proximal-gradient method updates multiplicatively with a |V|-dim dual KL projection, achieving linear convergence.For time-varying traffic, a threshold policy yields regret O(√ΓC V₁ T) against the switching-free optimum. Traffic changes are detected by a mu

ALT Repeatedly sampling short sequence windows from a human pangenome requires an access representation as well as compression. panvram stores each assembly as a reference-relative block archive and pools these archives on a GPU. For 558 haplotype assemblies, the default q4k representation occupies 16.642 GB of persistent pools including the decoded reference; a packed-reference, compact-directory representation occupies 11.634 GB. These quantities are not peak application memory. In the C1 experiment, default and fully compact forms delivered 743,271 and 736,762 windows/s; a separate public Colab run delivered 741,931 and 680,760 windows/s. All are batch API measurements and must not be pooled as repeated trials of an identical environment. Full-decoding checks compared reconstructed FASTA checksums with the manifest. On the CPU, the M6 shuffled-window workload at the same cohort cardinality reported 26 requests/s for AGC and 145,222 requests/s for refrel3 q4k. This comparison concerns on