# resource orchestration

A computing technique for dynamically deploying and managing microservices across heterogeneous, multi-tiered compute and network resources.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## LEDGE: An Efficient Simulator for Distributed LLM Serving on Heterogeneous Edge Devices

DevFeed: [LEDGE: An Efficient Simulator for Distributed LLM Serving on Heterogeneous Edge Devices](<https://devfeed.tech/articles/ledge-an-efficient-simulator-for-distributed-llm-serving-on-heterogeneous-edge-devices-142754.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.24085>)

Author: Zihao Wang, Ye Tian, Tajana Rosing, Nalini Venkatasubramanian

Published: 2026-11-19T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [LLM Techniques](<https://devfeed.tech/topics/llm-techniques.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [qwen](<https://devfeed.tech/topics/qwen.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-devices](<https://devfeed.tech/tags/edge-devices.md>), [jetson](<https://devfeed.tech/tags/jetson.md>), [latency](<https://devfeed.tech/tags/latency.md>), [llm](<https://devfeed.tech/tags/llm.md>), [llm-serving](<https://devfeed.tech/tags/llm-serving.md>), [pipeline-parallelism](<https://devfeed.tech/tags/pipeline-parallelism.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [raspberry](<https://devfeed.tech/tags/raspberry.md>), [scaling](<https://devfeed.tech/tags/scaling.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [simulator](<https://devfeed.tech/tags/simulator.md>), [source-code](<https://devfeed.tech/tags/source-code.md>)

### AI overview

LEDGE simulates distributed LLM serving on heterogeneous edge devices without executing tensor operations. It builds event graphs for non-uniform tensor and pipeline parallelism, estimates execution costs with reusable performance models calibrated from end-to-end measurements, and models shared-link contention. On a 10-device Jetson and Raspberry Pi testbed, it covered 17 model-precision combinations across four LLM families. For Qwen deployments, request-level MAPE was below 3.6%, with 135.3x aggregate speedup over physical inference. Models calibrated on one family achieved 7.44% zero-shot end-to-end latency MAPE across three others. The authors also report simulations showing effects of shared-link contention on placement and network topology on scaling; 26,773 runs across four edge environments took 83.8 minutes.

### Source excerpt

arXiv:2611.24085v1 Announce Type: cross Abstract: Evaluating distributed large language model (LLM) serving on heterogeneous edge devices requires exploring diverse model partitions, device placements, and network conditions, making repeated physical experiments costly. An efficient simulator for edge LLM serving must support diverse, fine-grained placement configurations, capture runtime behavior, and accurately predict execution times while keeping simulation overhead low. We present LEDGE, a lightweight simulator for fast, fine-grained evaluation of distributed LLM serving on heterogeneous edge devices without executing tensor operations. LEDGE constructs event graphs supporting non-uniform tensor and pipeline parallelism, estimates execution costs using reusable performance models calibrated from end-to-end measurements, and simulates shared-link contention. We evaluate LEDGE on a heterogeneous testbed of 10 Jetson and Raspberry Pi devices, covering 17 model--precision combinations across 4 LLM families. On Qwen deployments, LEDGE achieves request-level mean absolute percentage errors below 3.6% and a 135.3 times aggregate speedup over physical inference. Performance model calibrated on one LLM family yield 7.44% zero-shot end-to-end latency MAPE across three other families. Further simulations show how shared-link contention affects placement decisions and network topology constrains scaling. Across four edge environments, LEDGE completes 26,773 simulation runs in 83.8 minutes, enabling efficient exploration of deployment strategies under diverse resource conditions and serving objectives. The source code is publicly available at https://github.com/gdfwj/LEDGE.

## IA-MoE: An Intensity-Aware Mixture of Experts for DNN Inference

DevFeed: [IA-MoE: An Intensity-Aware Mixture of Experts for DNN Inference](<https://devfeed.tech/articles/ia-moe-an-intensity-aware-mixture-of-experts-for-dnn-inference-136984.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19669>)

Author: Meiyan Zeng, Sasu Tarkoma, Xiaoli Liu

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Simulation](<https://devfeed.tech/topics/simulation.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [applications](<https://devfeed.tech/tags/applications.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [based](<https://devfeed.tech/tags/based.md>), [bottlenecks](<https://devfeed.tech/tags/bottlenecks.md>), [challenges](<https://devfeed.tech/tags/challenges.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-based](<https://devfeed.tech/tags/cloud-based.md>), [complexity](<https://devfeed.tech/tags/complexity.md>), [computing](<https://devfeed.tech/tags/computing.md>), [critical](<https://devfeed.tech/tags/critical.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [data](<https://devfeed.tech/tags/data.md>), [datasets](<https://devfeed.tech/tags/datasets.md>), [deep](<https://devfeed.tech/tags/deep.md>), [dnn](<https://devfeed.tech/tags/dnn.md>), [experts](<https://devfeed.tech/tags/experts.md>), [inference](<https://devfeed.tech/tags/inference.md>), [mixture-of-experts](<https://devfeed.tech/tags/mixture-of-experts.md>), [moe](<https://devfeed.tech/tags/moe.md>)

### AI overview

IA-MoE dynamically partitions DNN layers across resources in an edge-cloud environment to jointly optimize end-to-end latency and energy use under changing inference requests and multi-resource constraints. It frames each request's assignment of layer tasks to virtual machines as a bilateral matching problem. In enhanced CloudSimSDN simulations using AlexNet, VGG16, and ResNet50 with two public inference datasets, it outperformed baseline methods on weighted normalized latency and energy metrics. The reported evidence is simulation-based; the abstract provides no numerical results.

### Source excerpt

arXiv:2611.19669v1 Announce Type: new Abstract: With the ever-increasing parameter scale and structural complexity of deep neural network (DNN) models, traditional cloud-based inference encounters bottlenecks in performance, privacy protection, and energy efficiency. While edge computing and edge intelligence facilitate the migration of inference tasks to edge servers closer to the data source, existing end-edge-cloud collaborative inference still encounters critical challenges, such as sensitivity to heterogeneous resources, complexity in multi-objective optimization, and adaptability to dynamic workloads. To achieve the joint optimization of end-to-end latency and energy consumption under dynamic inference requests with multi-resource constraints, we propose IA-MoE, an intensity-aware Mixture of Experts (MoE) framework. By modeling the model partitioning problem as a bilateral matching problem between layer tasks and virtual machine resources for each inference request, IA-MoE dynamically partitions DNN models in an edge-cloud hybrid computing environment. We evaluate IA-MoE using three representative DNNs (i.e., AlexNet, VGG16, ResNet50) on two public inference datasets in an enhanced CloudSimSDN simulation environment. Experimental results demonstrate that IA-MoE outperforms baseline methods regarding weighted normalized metrics for end-to-end latency and energy consumption. To our knowledge, this is the first study to introduce the MoE paradigm for DNN partitioning and collaborative inference in hybrid computing environments, providing a practical solution for efficient inference in artificial intelligence applications.

## Sustaining In-Network Computation in Satellite Networks: An Architecture for Semantic Continuity

DevFeed: [Sustaining In-Network Computation in Satellite Networks: An Architecture for Semantic Continuity](<https://devfeed.tech/articles/sustaining-in-network-computation-in-satellite-networks-an-architecture-for-semantic-continuity-136983.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.19489>)

Author: Zhiyuan Ren, Wenchi Cheng

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>), [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [distributed algorithm](<https://devfeed.tech/topics/distributed-algorithm.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Network design](<https://devfeed.tech/topics/network-design.md>)

Tags: [acceptance](<https://devfeed.tech/tags/acceptance.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [computing](<https://devfeed.tech/tags/computing.md>), [context](<https://devfeed.tech/tags/context.md>), [continuation](<https://devfeed.tech/tags/continuation.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [dependency](<https://devfeed.tech/tags/dependency.md>), [execution](<https://devfeed.tech/tags/execution.md>), [images](<https://devfeed.tech/tags/images.md>), [isolation](<https://devfeed.tech/tags/isolation.md>), [linux](<https://devfeed.tech/tags/linux.md>), [logical](<https://devfeed.tech/tags/logical.md>), [network](<https://devfeed.tech/tags/network.md>), [node](<https://devfeed.tech/tags/node.md>), [providers](<https://devfeed.tech/tags/providers.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [source](<https://devfeed.tech/tags/source.md>), [wildfire](<https://devfeed.tech/tags/wildfire.md>)

### AI overview

The paper proposes keeping one logical computation continuous as satellite paths and providers change. Its architecture separates logical computation, stream context, and physical binding: workers run ready operations, while consumers track which results already contribute. In wildfire assessment, retained observations support recomputation while reachable association history remains available. In a 300-node Linux prototype, stage-wise placement met deadlines for 82.08% of 120 image positions, compared with 76.06% for single-node placement. For 50 observations under 20-node isolation, Confirmation-Coordinated Rebinding coordinated replanning, source replay, cross-attempt acceptance, and in-band confirmation, completing all observations with verified outputs using seven replays and ignoring 13 duplicate outputs. The supplied text is an abstract/excerpt, so it does not establish broader performance or deployment results.

### Source excerpt

arXiv:2611.19489v1 Announce Type: cross Abstract: Satellite computations often outlast the paths and providers initially serving them. We argue that in-network computing should organize changing execution opportunities around one continuing logical computation: workers execute ready operations, while consumers recognize which results already contribute. We present an architecture separating logical computation, stream context, and physical binding, and show in wildfire assessment how retained observations support recomputation while reachable association history stays in place. A 300-node Linux prototype examines placement, dependency assembly, and continuation. Across 120 images, stage-wise placement delivers 82.08% of positions within deadlines versus 76.06% for single-node placement, while traces separate blocked output delivery from dependencies split across providers. Across 50 observations with 20-node isolation, Confirmation-Coordinated Rebinding (CCR) coordinates replanning, source replay, cross-attempt acceptance, and in-band confirmation to complete all 50 observations with verified outputs using seven replays and 13 ignored duplicate outputs.

## Multi-Granularity Communication-Computing Resource Mapping for User Association in MEO-Assisted LEO Satellite Networks

DevFeed: [Multi-Granularity Communication-Computing Resource Mapping for User Association in MEO-Assisted LEO Satellite Networks](<https://devfeed.tech/articles/multi-granularity-communication-computing-resource-mapping-for-user-association-in-meo-assisted-leo-satellite-networks-126794.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.11208>)

Author: Yi Jing, Chunxiao Jiang, Jiawei Wang, Jiachen Sun, Yafeng Zhan

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Network Operations Center](<https://devfeed.tech/topics/network-operations-center.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [account](<https://devfeed.tech/tags/account.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [association](<https://devfeed.tech/tags/association.md>), [audits](<https://devfeed.tech/tags/audits.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [cache](<https://devfeed.tech/tags/cache.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [collect](<https://devfeed.tech/tags/collect.md>), [communication](<https://devfeed.tech/tags/communication.md>), [computing](<https://devfeed.tech/tags/computing.md>), [control](<https://devfeed.tech/tags/control.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [leo](<https://devfeed.tech/tags/leo.md>), [mapping](<https://devfeed.tech/tags/mapping.md>), [networks](<https://devfeed.tech/tags/networks.md>), [resource](<https://devfeed.tech/tags/resource.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [user](<https://devfeed.tech/tags/user.md>)

### AI overview

The paper proposes user association for MEO-assisted LEO networks that accounts for both radio links and onboard computing constraints. MEO aggregators relay time-stamped telemetry, while the NOCC maintains the global resource view and makes association decisions. A service-primitive mapping converts requests into target-LEO-dependent quantized resource vectors; the integer formulation, exact MILP reference, and capacity-aware primal-dual algorithm use the same mapped objective. In experiments with real-world bandwidth traces and a 1,584-satellite shell, the method improved realized utility by 10.64% and reduced end-to-end outage by 5.70 percentage points versus static mapping. Target quantization cut signaling by 19.84% with negligible utility loss. The abstract notes further audits of load, telemetry delay, temporal sessions, runtime, and mapping uncertainty, but does not provide their detailed results.

### Source excerpt

arXiv:2611.11208v1 Announce Type: new Abstract: Low-Earth-orbit (LEO) satellite networks are evolving into spaceborne communication-computing infrastructures. Their user association must jointly account for the radio link and the CPU, GPU, NPU, memory, cache, storage, energy, queue, and thermal states that determine executable service capacity. This paper studies a MEO-assisted LEO architecture in which medium-Earth-orbit (MEO) regional state aggregators collect and relay time-stamped telemetry, the network operation control center (NOCC) maintains the global logical resource view and makes the final association decisions, and LEO satellites provide access and onboard execution. We develop a multigranularity representation and service-primitive-driven mapping that converts a service-level request into a target-LEO-dependent quantized resource vector. The integer formulation, exact mixed-integer linear programming (MILP) reference, and a capacity-aware primal-dual algorithm share the same mapped objective. Experiments use real-world service bandwidth traces, a 1584-satellite LEO shell, multiple traffic seeds, controlled load and mapping error, delayed telemetry, and temporal sessions. Relative to static mapping, the proposed method improves realized utility by 10.64% and lowers end-to-end outage by 5.70 percentage points. Target quantization reduces signaling by 19.84% with negligible utility loss. Additional load, delay, temporal, runtime, and exact-MILP audits delimit the operating region and remaining mapping uncertainty.

## Adapting Multi-Stage Distributed Real-Time Systems Design to Vehicular Edge Computing

DevFeed: [Adapting Multi-Stage Distributed Real-Time Systems Design to Vehicular Edge Computing](<https://devfeed.tech/articles/adapting-multi-stage-distributed-real-time-systems-design-to-vehicular-edge-computing-126745.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.12370>)

Author: Niraj Kumar, Anant Gautam

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [real-time](<https://devfeed.tech/topics/real-time.md>), [scheduling](<https://devfeed.tech/topics/scheduling.md>), [distributed algorithm](<https://devfeed.tech/topics/distributed-algorithm.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>)

Tags: [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [real-time](<https://devfeed.tech/tags/real-time.md>), [stage](<https://devfeed.tech/tags/stage.md>)

### AI overview

The paper models remotely processed vehicular applications as multi-stage pipelines that access heterogeneous resources in sequence. It proposes a joint mapping and scheduling heuristic that extends optimal priority assignment to account for vehicle mobility and real-time constraints, aiming to build feasible schedules, minimize local-device energy use, and meet deadlines for offloaded jobs. In simulations, the approach consistently outperformed a baseline using the same delay-composition framework and a deadline decomposition-based approach. The supplied text reports simulation results but gives no numerical performance gains.

### Source excerpt

arXiv:2611.12370v1 Announce Type: new Abstract: Vehicular Edge Computing (VEC) aims to support computation-intensive applications in intelligent transportation systems while meeting stringent latency requirements. When processed remotely, applications require access to different resource types in a specific sequence, forming a multi-stage processing pipeline. This structure closely resembles multi-stage distributed applications executing across heterogeneous resources, making established multi-stage distributed real-time scheduling techniques well suited for modeling VEC systems. The resulting model accommodates diverse resource configurations and access patterns. Building on this model, we propose a joint mapping and scheduling approach that accounts for vehicle mobility and real-time constraints. The proposed heuristic extends optimal priority assignment for multi-stage systems to the VEC setting, to construct feasible schedules while minimizing local-device energy consumption and ensuring that offloaded jobs meet their deadlines. Extensive simulations show that the proposed approach consistently outperforms both a baseline based on the same delay-composition framework and a deadline decomposition-based approach.

## UniExit: A Unified Latency-Aware Early-Exit Runtime for Heterogeneous Edge AI

DevFeed: [UniExit: A Unified Latency-Aware Early-Exit Runtime for Heterogeneous Edge AI](<https://devfeed.tech/articles/uniexit-a-unified-latency-aware-early-exit-runtime-for-heterogeneous-edge-ai-123687.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.09159>)

Author: Yuhang Guan, Huanle Zhang, Xiaomin Ouyang, Zhidan Liu

Published: 2026-11-09T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [AI Integration Strategies](<https://devfeed.tech/topics/ai-integration-strategies.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [budget](<https://devfeed.tech/tags/budget.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cnn](<https://devfeed.tech/tags/cnn.md>), [concurrent](<https://devfeed.tech/tags/concurrent.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [devices](<https://devfeed.tech/tags/devices.md>), [edge](<https://devfeed.tech/tags/edge.md>), [edge-ai](<https://devfeed.tech/tags/edge-ai.md>), [edge-cloud](<https://devfeed.tech/tags/edge-cloud.md>), [energy](<https://devfeed.tech/tags/energy.md>), [energy-consumption](<https://devfeed.tech/tags/energy-consumption.md>), [exit](<https://devfeed.tech/tags/exit.md>), [fallback](<https://devfeed.tech/tags/fallback.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [implement](<https://devfeed.tech/tags/implement.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [runtime](<https://devfeed.tech/tags/runtime.md>)

### AI overview

UniExit uses a shared latency-aware routing policy for CNN blocks and Transformer layers, making budget-aware decisions to continue inference or exit early. It estimates the cost of reaching the next exit from runtime state and inference progress, with cache-based prediction reuse and a trend-aware fallback; model-specific quality mechanisms remain in place, including CNN-specific exit stabilization. The abstract reports evaluation on multiple edge and edge-cloud platforms across single-workload, concurrent heterogeneous, and edge-cloud settings. Against existing progressive inference baselines, it reports improved deadline satisfaction and latency-accuracy trade-offs, along with less redundant exit-module computation and lower task-level energy use. The supplied text does not give quantitative results or implementation details.

### Source excerpt

arXiv:2611.09159v1 Announce Type: new Abstract: Edge applications increasingly combine CNN-based perception with edge-deployable Transformer-based language models, but deploying such heterogeneous workloads on resource-constrained devices is challenging because inference latency varies with both input difficulty and runtime hardware state. Early-exit inference reduces computation by terminating at intermediate stages, yet existing methods are often model-specific and rely on static latency profiles that become stale under dynamic edge conditions. This paper presents UniExit, a unified latency-aware early-exit runtime for heterogeneous edge AI. UniExit abstracts CNN blocks and Transformer layers as ordered stages and applies a common latency-aware routing policy while retaining model-specific quality mechanisms. It predicts the cost of reaching the next exit from runtime state and inference progress to make budget-aware continue-or-exit decisions. UniExit combines runtime-state-aware latency prediction, cache-based prediction reuse, trend-aware fallback, and CNN-specific exit stabilization to provide low-overhead and robust online control. We implement UniExit on multiple edge and edge-cloud platforms and evaluate it with representative CNN and Transformer-based language-model workloads. Across single-workload, concurrent heterogeneous, and edge-cloud settings, UniExit consistently improves deadline satisfaction and latency-accuracy trade-offs over existing progressive inference baselines, while reducing redundant exit-module computation and task-level energy consumption.

## Uncertainty-Aware Resource Orchestration for Agentic AI Systems

DevFeed: [Uncertainty-Aware Resource Orchestration for Agentic AI Systems](<https://devfeed.tech/articles/uncertainty-aware-resource-orchestration-for-agentic-ai-systems-118315.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.05392>)

Author: Masoud Shokrnezhad, Tarik Taleb

Published: 2026-11-05T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>), [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-systems](<https://devfeed.tech/tags/agentic-ai-systems.md>), [agentic-orchestration](<https://devfeed.tech/tags/agentic-orchestration.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [compute](<https://devfeed.tech/tags/compute.md>), [conformal-prediction](<https://devfeed.tech/tags/conformal-prediction.md>), [coverage](<https://devfeed.tech/tags/coverage.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [demand](<https://devfeed.tech/tags/demand.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [resource](<https://devfeed.tech/tags/resource.md>), [resource-orchestration](<https://devfeed.tech/tags/resource-orchestration.md>), [uncertainty](<https://devfeed.tech/tags/uncertainty.md>)

### AI overview

Agentic AI pipelines have unpredictable compute, bandwidth, and latency demands that emerge during execution, making fixed-graph schedulers and parametric or stationary demand models a poor fit. The paper formulates dispatch-time resource reservation with probabilistic coverage and deadline requirements, then proposes Conformal Reservation for Orchestration (CRO), which uses a sliding window of execution traces and conformal prediction to provide finite-sample, distribution-free coverage guarantees. In simulations with heterogeneous, multimodal, heavy-tailed task profiles, CRO was the only evaluated method to meet target coverage across all tested risk levels; among coverage-feasible methods, it had the lowest over-reservation, roughly halving resource waste versus the Chebyshev-bound baseline.

### Source excerpt

arXiv:2611.05392v1 Announce Type: new Abstract: Agentic artificial intelligence (AI) systems powered by large language models (LLMs) execute as autonomous multi-step pipelines that dynamically invoke tools, integrate observations into memory, and iterate until a task is resolved or a resource budget is exhausted. Unlike classical workflows with fixed graphs, the aggregate compute, bandwidth, and end-to-end latency demands of such pipelines are unknown at dispatch time: they emerge at runtime from intermediate observations and stochastic token sampling, yielding heavy-tailed, multi-modal random variables. This unpredictability forces infrastructure orchestrators into a dilemma: under-provisioning risks mid-execution throttling and deadline violations, while over-provisioning wastes scarce edge or cloud capacity. Existing schedulers require a fully known pipeline graph, while edge resource-management methods rely on parametric or stationary demand models, and neither fits the unknown, non-stationary profiles of agentic pipelines. We formalize this setting as the uncertainty-aware agentic orchestration problem, where the goal is to reserve, at dispatch time, compute and bandwidth allocations that cover actual pipeline demands, and to verify that the pipeline can meet its deadline, each with a prescribed probability. We then propose Conformal Reservation for Orchestration (CRO), an online algorithm that translates a sliding window of past execution traces into dispatch-time resource allocations with finite-sample, distribution-free coverage guarantees via conformal prediction. Simulations over heterogeneous task mixes with multi-modal and heavy-tailed profiles show that CRO is the only evaluated method that simultaneously satisfies target probabilistic coverage across all risk levels while, among coverage-feasible methods, attaining the lowest over-reservation, roughly halving the resource waste of the Chebyshev-bound baseline.

## ORION: An Outcome-Driven LLM Advisor for RL-Based Multi-Domain 6G Slice Placement

DevFeed: [ORION: An Outcome-Driven LLM Advisor for RL-Based Multi-Domain 6G Slice Placement](<https://devfeed.tech/articles/orion-an-outcome-driven-llm-advisor-for-rl-based-multi-domain-6g-slice-placement-111599.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.41563>)

Author: Mahmoud S. Mayaleh, Fetia Bannour, Nour-El-Houda Yellas

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Large Language Model](<https://devfeed.tech/topics/llm.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [hierarchical agent systems](<https://devfeed.tech/topics/hierarchical-agent-systems.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>)

Tags: [6g](<https://devfeed.tech/tags/6g.md>), [admission](<https://devfeed.tech/tags/admission.md>), [but](<https://devfeed.tech/tags/but.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [episodic-memory](<https://devfeed.tech/tags/episodic-memory.md>), [layer](<https://devfeed.tech/tags/layer.md>), [llm](<https://devfeed.tech/tags/llm.md>), [memory](<https://devfeed.tech/tags/memory.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [np-hard](<https://devfeed.tech/tags/np-hard.md>), [online](<https://devfeed.tech/tags/online.md>), [planner](<https://devfeed.tech/tags/planner.md>), [rl](<https://devfeed.tech/tags/rl.md>)

### AI overview

ORION targets 6G slice placement across administrative domains, where each arriving request requires choosing function locations, cross-domain traffic paths, and admission. It combines LLM agents that turn service intents into structured requests and abstract placement plans with a cooperative multi-agent RL coordinator and per-domain placement agents. Its main contribution is an episodic memory that records network conditions, proposed plans, and verified feasibility outcomes, using both successes and labelled failures to adapt planning. In a controlled evaluation on held-out topology families, memory improves admission by 4.6-11.5 percentage points over the same planner without memory; importance-based retention accounts for most of the gain over a first-in-first-out store. The abstract also reports that no fixed planning strategy leads across all topology regimes.

### Source excerpt

arXiv:2610.41563v1 Announce Type: new Abstract: Deploying 6G network slices across administrative domains requires an online decision for each arriving request: which domain hosts each virtual network function, how traffic crosses domain boundaries, and whether the slice is admitted. This joint decision problem is NP-hard under continuous arrivals. Experiments on a parametrized family of multi-domain topologies show that no fixed planning strategy dominates across all regimes: the performance ranking reverses between topology classes, and feasible placements exist for most rejected requests but remain undiscovered. These findings identify the plan layer as the decisive performance lever. We introduce ORION, a hierarchical orchestrator in which two LLM agents translate service intents into structured requests and generate abstract placement plans. A cooperative multi-agent RL coordinator selects the cross-domain partition, treating the plan as a persistent soft prior, while per-domain agents place the functions. ORION main contribution lies at the plan layer: an outcome-driven episodic memory that records, for each completed attempt, the network-condition profile, the proposed plan, and its verified feasibility outcome. The memory feeds verified successes and labelled failures back to the planner, thereby closing the adaptation loop using deployment feasibility rather than a scalar reward assigned to the planner's own proposals. To isolate the effect of the plan quality, we evaluate the planner on held-out topology families while keeping the downstream selection and execution mechanisms fixed. Under this controlled setting, the proposed memory improves admission by $4.6$--$11.5$ percentage points over the same planner without memory, while importance-based retention accounts for most of the improvement over a first-in-first-out store.

## Building an AI Agent Orchestrator for Decentralized GPU Compute

DevFeed: [Building an AI Agent Orchestrator for Decentralized GPU Compute](<https://devfeed.tech/articles/building-an-ai-agent-orchestrator-for-decentralized-gpu-compute-110840.md>)

Original publisher: [Read original article](<https://hackernoon.com/building-an-ai-agent-orchestrator-for-decentralized-gpu-compute?source=rss>)

Author: Haris Ahmad

Published: 2026-11-01T15:07:15Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [DePIN](<https://devfeed.tech/topics/depin.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>)

Tags: [agencies](<https://devfeed.tech/tags/agencies.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agent-identities](<https://devfeed.tech/tags/agent-identities.md>), [agent-orchestration](<https://devfeed.tech/tags/agent-orchestration.md>), [agent-workflows](<https://devfeed.tech/tags/agent-workflows.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-automation](<https://devfeed.tech/tags/ai-automation.md>), [ai-compute](<https://devfeed.tech/tags/ai-compute.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [arweave](<https://devfeed.tech/tags/arweave.md>), [building](<https://devfeed.tech/tags/building.md>), [compute](<https://devfeed.tech/tags/compute.md>), [decentralize-ai](<https://devfeed.tech/tags/decentralize-ai.md>), [decentralize-ai-hackathon](<https://devfeed.tech/tags/decentralize-ai-hackathon.md>), [decentralized](<https://devfeed.tech/tags/decentralized.md>), [developer-tools](<https://devfeed.tech/tags/developer-tools.md>), [execution](<https://devfeed.tech/tags/execution.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-cloud](<https://devfeed.tech/tags/gpu-cloud.md>), [gpu-compute](<https://devfeed.tech/tags/gpu-compute.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [nosana](<https://devfeed.tech/tags/nosana.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [routing](<https://devfeed.tech/tags/routing.md>), [stage](<https://devfeed.tech/tags/stage.md>)

### AI overview

The author proposes a design-stage orchestrator that would route AI jobs to decentralized GPUs, starting with Nosana, to give small builders an alternative to hyperscaler compute. A planned FastAPI gateway would accept jobs; an open-source LLM planner would translate constraints such as GPU memory and latency into a container manifest; a Nosana submitter would launch and monitor the job; and planned Arweave logging would preserve prompts, model references, and execution metadata as auditable receipts. The proposed planner selects the cheapest node meeting constraints. The author says the build is still ahead and identifies open questions around sustained multi-tenant spot-market behavior, reconciling token payments with client subscriptions, and monitoring or attesting heterogeneous nodes. These are plans and unresolved issues, not demonstrated results.

### Source excerpt

A concept-stage AI orchestrator for routing workloads across decentralized GPUs with Nosana and planned Arweave execution receipts.

## Aceso: Latency-Aware Geo-Distribution of Microservices for Cost-Effective Carbon Reduction

DevFeed: [Aceso: Latency-Aware Geo-Distribution of Microservices for Cost-Effective Carbon Reduction](<https://devfeed.tech/articles/aceso-latency-aware-geo-distribution-of-microservices-for-cost-effective-carbon-reduction-107552.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2603.10768>)

Author: Georgia Christofidi, Francisco \\'Alvarez-Terribas, Ioannis Roumpos, Nicolas Kourtellis, Jesus Oma\\~na Iglesias, Thaleia Dimitra Doudali

Published: 2026-10-30T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Green Software](<https://devfeed.tech/topics/green-software.md>), [Microservices](<https://devfeed.tech/topics/microservices.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [performance-engineering](<https://devfeed.tech/topics/performance-engineering.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [carbon](<https://devfeed.tech/tags/carbon.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-computing](<https://devfeed.tech/tags/cloud-computing.md>), [cost](<https://devfeed.tech/tags/cost.md>), [country](<https://devfeed.tech/tags/country.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [data-sovereignty](<https://devfeed.tech/tags/data-sovereignty.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [distribution](<https://devfeed.tech/tags/distribution.md>), [efficiency](<https://devfeed.tech/tags/efficiency.md>), [emissions](<https://devfeed.tech/tags/emissions.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [geo](<https://devfeed.tech/tags/geo.md>), [global](<https://devfeed.tech/tags/global.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [latency](<https://devfeed.tech/tags/latency.md>), [microservices](<https://devfeed.tech/tags/microservices.md>), [mitigation](<https://devfeed.tech/tags/mitigation.md>), [reduction](<https://devfeed.tech/tags/reduction.md>)

### AI overview

Aceso places microservices across geographically constrained regions to balance carbon emissions, operational cost, and latency. Its scalable optimization prunes the search space and adapts placement to changing workloads and carbon intensity. In a real-world deployment, compared with a static deployment within one country, it reduced emissions by up to 37.4% and operational cost by up to 3.36% on average across days while maintaining latency targets. The supplied abstract does not detail the deployment or evaluation methodology.

### Source excerpt

arXiv:2603.10768v3 Announce Type: replace Abstract: Microservices are a dominant architecture in cloud computing, offering scalability and modularity, but also posing complex deployment challenges. As data centers contribute significantly to global carbon emissions, carbon-aware scheduling is a promising mitigation strategy. However, most existing solutions target batch, HPC, or serverless workloads and assume global-scale infrastructure. Such an assumption does not hold for many regional small to medium-sized enterprises (SMEs) with microservice applications, due to geographic, cost, and data sovereignty constraints. In this paper, we present Aceso, an Adaptive Carbon- and Efficiency-aware placement for microservices that considers carbon, cost, and latency constraints. Aceso dynamically places microservices across geographically constrained regions using a scalable optimization with insight-based search space pruning techniques. Evaluation on a real-world deployment shows that Aceso adapts to real-time changes in workload and carbon intensity and reduces carbon emissions by up to 37.4% and operational cost by up to 3.36%, on average across days, compared to a static deployment within a single country, while maintaining the application's latency targets. In this way, Aceso enables carbon- and cost-aware microservice deployment for latency-sensitive applications in regionally limited infrastructures.

## Uber Moves Beyond CPU Cores with Performance-Aware Scheduling for Stateful Workloads

DevFeed: [Uber Moves Beyond CPU Cores with Performance-Aware Scheduling for Stateful Workloads](<https://devfeed.tech/articles/uber-moves-beyond-cpu-cores-with-performance-aware-scheduling-for-stateful-workloads-102114.md>)

Original publisher: [Read original article](<https://www.infoq.com/news/2026/10/uber-cpu-stateful-workloads/>)

Author: Craig Risi

Published: 2026-10-27T12:00:00Z

Content type: article

Language: en

Sources: [InfoQ](<https://devfeed.tech/sources/infoq.md>)

Topics: [scheduling](<https://devfeed.tech/topics/scheduling.md>), [Stateful](<https://devfeed.tech/topics/stateful.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Load Balancing](<https://devfeed.tech/topics/load-balancing.md>)

Tags: [1-4](<https://devfeed.tech/tags/1-4.md>), [abstraction](<https://devfeed.tech/tags/abstraction.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-ml-data-engineering](<https://devfeed.tech/tags/ai-ml-data-engineering.md>), [apache-kafka](<https://devfeed.tech/tags/apache-kafka.md>), [arm](<https://devfeed.tech/tags/arm.md>), [automation-in-the-cloud-and-management-at-scale](<https://devfeed.tech/tags/automation-in-the-cloud-and-management-at-scale.md>), [aws](<https://devfeed.tech/tags/aws.md>), [balancing](<https://devfeed.tech/tags/balancing.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [compute](<https://devfeed.tech/tags/compute.md>), [core](<https://devfeed.tech/tags/core.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cpu-time](<https://devfeed.tech/tags/cpu-time.md>), [data-centres](<https://devfeed.tech/tags/data-centres.md>), [devops](<https://devfeed.tech/tags/devops.md>), [fleet](<https://devfeed.tech/tags/fleet.md>), [kafka](<https://devfeed.tech/tags/kafka.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [make](<https://devfeed.tech/tags/make.md>), [ml-data-engineering](<https://devfeed.tech/tags/ml-data-engineering.md>), [news](<https://devfeed.tech/tags/news.md>), [performance](<https://devfeed.tech/tags/performance.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [server](<https://devfeed.tech/tags/server.md>), [state-machine](<https://devfeed.tech/tags/state-machine.md>), [stateful](<https://devfeed.tech/tags/stateful.md>), [uber](<https://devfeed.tech/tags/uber.md>), [uber-cpu-stateful-workloads](<https://devfeed.tech/tags/uber-cpu-stateful-workloads.md>), [units](<https://devfeed.tech/tags/units.md>)

### AI overview

Uber's Normalized Compute Units (NCUs) let Odin allocate comparable compute capacity across servers whose CPU performance varies substantially. Odin assigns each server type a performance factor, initially measured during hardware qualification and refined with production workload data, then uses it to translate workload requirements into different numbers of physical cores. This addresses a reliability risk for stateful workloads, which cannot readily shift requests away from a slower server; stateless services can more often use load balancing. Across about 1.4 million automatically scaled cores, Kafka's utilization gap between server types fell from roughly eight to three percentage points, and total Kafka CPU allocation dropped about 8%. Uber contrasts NCUs with Kubernetes CPU requests in cores or millicores, which represent CPU time without expressing relative processor performance.

### Source excerpt

Uber has introduced Normalized Compute Units (NCUs) to make CPU allocation more consistent across its highly heterogeneous server fleet. By Craig Risi

## How Virgin Media O2 cut compute costs 30% by giving dbt a memory

DevFeed: [How Virgin Media O2 cut compute costs 30% by giving dbt a memory](<https://devfeed.tech/articles/how-virgin-media-o2-cut-compute-costs-30-by-giving-dbt-a-memory-101845.md>)

Original publisher: [Read original article](<https://www.getdbt.com/blog/virgin-media-o2-dbt-state-compute-costs>)

Author: Daniel Poppy

Published: 2026-10-26T14:04:00Z

Content type: news

Language: en

Sources: [dbt Labs](<https://devfeed.tech/sources/dbt-labs.md>)

Topics: [cost-optimization](<https://devfeed.tech/topics/cost-optimization.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Tool](<https://devfeed.tech/topics/tool.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [adopting](<https://devfeed.tech/tags/adopting.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-workflows](<https://devfeed.tech/tags/ai-workflows.md>), [analytics](<https://devfeed.tech/tags/analytics.md>), [analytics-engineering](<https://devfeed.tech/tags/analytics-engineering.md>), [bigquery](<https://devfeed.tech/tags/bigquery.md>), [ci-cd](<https://devfeed.tech/tags/ci-cd.md>), [ci-cd-pipeline](<https://devfeed.tech/tags/ci-cd-pipeline.md>), [compute](<https://devfeed.tech/tags/compute.md>), [compute-costs](<https://devfeed.tech/tags/compute-costs.md>), [costs](<https://devfeed.tech/tags/costs.md>), [dbt](<https://devfeed.tech/tags/dbt.md>), [dependency](<https://devfeed.tech/tags/dependency.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [learn](<https://devfeed.tech/tags/learn.md>), [model](<https://devfeed.tech/tags/model.md>), [scale](<https://devfeed.tech/tags/scale.md>), [state](<https://devfeed.tech/tags/state.md>), [times](<https://devfeed.tech/tags/times.md>), [v2](<https://devfeed.tech/tags/v2.md>)

### AI overview

Virgin Media O2 reports that migrating its BigQuery dbt platform to dbt v2 with dbt State cut compute costs by about 30% and model run times by about 35% across jobs. Its platform serves 300+ users across 100+ projects and 5,500+ models; previously, even small changes could trigger redundant full rebuilds. In a 1,424-model test, dbt State reused 32 unchanged models, reducing runtime from about 1 hour 53 minutes to 1 hour 13 minutes and BigQuery consumption from roughly 4,500 to 3,400 slots per hour. The team says migration required addressing accumulated technical debt and outdated packages. Full production rollout still depends on aligning dbt v2 with custom FinOps configurations that route work between on-demand and reservation pricing.

### Source excerpt

Virgin Media O2 cut dbt compute costs 30% and model run times 35% by adopting dbt v2 and dbt State at scale.

## Migration as a Steady-State Execution Primitive: A Runtime for Virtual Stationarity in Low-Earth Orbit

DevFeed: [Migration as a Steady-State Execution Primitive: A Runtime for Virtual Stationarity in Low-Earth Orbit](<https://devfeed.tech/articles/migration-as-a-steady-state-execution-primitive-a-runtime-for-virtual-stationarity-in-low-earth-orbit-100175.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.32720>)

Author: Daniel Kleebinder, Alireza Furutanpey, Stefan Nastic

Published: 2026-10-26T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>), [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [virtualization](<https://devfeed.tech/topics/virtualization.md>)

Tags: [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [downtime](<https://devfeed.tech/tags/downtime.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [latency](<https://devfeed.tech/tags/latency.md>), [live-migration](<https://devfeed.tech/tags/live-migration.md>), [low-earth-orbit](<https://devfeed.tech/tags/low-earth-orbit.md>), [membrane](<https://devfeed.tech/tags/membrane.md>), [memory](<https://devfeed.tech/tags/memory.md>), [migration](<https://devfeed.tech/tags/migration.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [satellite](<https://devfeed.tech/tags/satellite.md>), [satellites](<https://devfeed.tech/tags/satellites.md>), [simulation](<https://devfeed.tech/tags/simulation.md>), [state](<https://devfeed.tech/tags/state.md>)

### AI overview

The proposed Virtual Stationarity Runtime (VSR) treats migration as continuous execution across successive LEO satellites, addressing handovers that traditional mechanisms designed for individual relocations or replication do not target. Its Migration Membrane combines Capability Virtualization and Continuous State Projection to rebind host resources and reconstruct evolving execution on each successor. The authors model latency and availability objectives, migration overhead, and state-replay convergence. On heterogeneous physical hosts, they report 95th-percentile downtime below 6 ms over 300 recurrent handovers, migration time effectively invariant to memory-write rate, and no cumulative growth in migration time or continuation state. A 1,600-satellite simulation completed migrations within orbital handover windows.

### Source excerpt

arXiv:2610.32720v1 Announce Type: new Abstract: Low Earth Orbit (LEO) satellite mega-constellations enable orbital edge computing for latency-critical geocentric applications, but satellite motion compresses line-of-sight communication windows to mere minutes. To maintain logical geographic proximity, a paradigm termed Virtual Stationarity (VS), applications must continuously migrate across passing satellites in the constellation. Traditional live migration mechanisms optimize individually initiated relocations or replication, but not the continuous handover between successive hosts that orbital motion requires. We reformulate live migration as a steady-state execution model and introduce a novel virtual stationarity runtime paradigm. We formalize the viability of continuous migration through a lightweight model capturing latency and availability service-level objectives (SLOs), continuous migration overhead, and the convergence bound of state replay. To satisfy these constraints, we derive a Virtual Stationarity Runtime (VSR) framework whose Migration Membrane combines Capability Virtualization and Continuous State Projection to rebind host resources and reconstruct evolving execution on a successor. Evaluation on physical heterogeneous hosts shows 95th-percentile downtime below 6 milliseconds across 300 recurrent handovers, migration time effectively invariant to memory-write rate, and no cumulative growth in migration time or continuation state. A 1,600-satellite simulation further confirms completion within orbital handover windows.

## InferSlice: A Reference Architecture for SLO-driven LLM Inference over a Telco Wide-Area Network

DevFeed: [InferSlice: A Reference Architecture for SLO-driven LLM Inference over a Telco Wide-Area Network](<https://devfeed.tech/articles/inferslice-a-reference-architecture-for-slo-driven-llm-inference-over-a-telco-wide-area-network-100030.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.32638>)

Author: Juan Parra-Ullauri, Anthony Chiu, Hari Madhukumar, Zhaozhou Wu, Anderson Bravalheri, Alex Healing, Andy Corston-Petrie, Dimitra Simeonidou, Nektaria Efthymiou

Published: 2026-10-26T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [compliance](<https://devfeed.tech/tags/compliance.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [disaggregation](<https://devfeed.tech/tags/disaggregation.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [gpu-partitioning](<https://devfeed.tech/tags/gpu-partitioning.md>), [latency](<https://devfeed.tech/tags/latency.md>), [llm-inference](<https://devfeed.tech/tags/llm-inference.md>), [locality](<https://devfeed.tech/tags/locality.md>), [memory](<https://devfeed.tech/tags/memory.md>), [observability](<https://devfeed.tech/tags/observability.md>), [parallelism](<https://devfeed.tech/tags/parallelism.md>), [pipeline-parallelism](<https://devfeed.tech/tags/pipeline-parallelism.md>), [reference-architecture](<https://devfeed.tech/tags/reference-architecture.md>), [replication](<https://devfeed.tech/tags/replication.md>), [scale-out](<https://devfeed.tech/tags/scale-out.md>), [slo](<https://devfeed.tech/tags/slo.md>), [sovereignty](<https://devfeed.tech/tags/sovereignty.md>), [streaming](<https://devfeed.tech/tags/streaming.md>), [telco](<https://devfeed.tech/tags/telco.md>), [wide-area-network](<https://devfeed.tech/tags/wide-area-network.md>)

### AI overview

InferSlice is a reference architecture for SLO-driven LLM inference across telco infrastructure, combining network slicing, GPU partitioning, elastic serving and observability. Evaluated on the JOINER UK wide-area network, it reports distinct latency-throughput regimes across service tiers and faster scale-out with weight streaming. Replication is insensitive to placement but constrained by memory; pipeline parallelism works across near- and mid-cross-site hops, while tensor parallelism and prefill/decode disaggregation fail beyond a single subnet hop. Client-edge bandwidth is also evaluated as a bottleneck. These findings describe a general-purpose telco network and highlight limits in extending tightly coupled inference techniques across WAN sites.

### Source excerpt

arXiv:2610.32638v1 Announce Type: cross Abstract: LLM inference is increasingly deployed as a core service for enterprise and interactive applications, but sovereignty, compliance, latency and resource-locality requirements can limit the suitability of centralised cloud deployments. Telco operators offer geographically distributed compute and network infrastructure for addressing these constraints. However, existing LLM serving frameworks are designed for tightly coupled data centres and provide limited support for coordinating geographically distributed and heterogeneous resources. We present InferSlice, a reference architecture for Service-Level Objective (SLO)-driven LLM inference over a telco infrastructure, integrating network slicing, GPU partitioning, elastic serving and observability. Led by BT in collaboration with the University of Bristol and NVIDIA, and realised on the JOINER UK national-scale wide-area network (WAN), InferSlice is evaluated on service differentiation across tiers, elastic scale-out, client-edge bandwidth as a bottleneck, and the reach of tensor, pipeline, data parallelism, and prefill/decode disaggregation over the WAN. Our results show that, on a general-purpose telco network, tiers yield distinct latency-throughput regimes, and weight streaming shortens scale-out time. Replication is placement-insensitive but memory-bound; pipeline parallelism tolerates near- and mid-cross-site hops, while tensor parallelism and disaggregation fail beyond a single subnet hop. We share practical insights from operating SLO-driven inference on a telco network, lessons learned and open questions, and hope to encourage further research on distributed and heterogeneous LLM inference systems.

## CoreWeave Adds Skypilot Support for Effortless Multi-Cloud AI Orchestration

DevFeed: [CoreWeave Adds Skypilot Support for Effortless Multi-Cloud AI Orchestration](<https://devfeed.tech/articles/coreweave-adds-skypilot-support-for-effortless-multi-cloud-ai-orchestration-96097.md>)

Original publisher: [Read original article](<https://wf.coreweave.com/blog/coreweave-adds-skypilot-support-for-effortless-multi-cloud-ai-orchestration>)

Author: CoreWeave

Published: 2026-10-22T16:27:39Z

Content type: release

Language: en

Sources: [Coreweave Blog Posts](<https://devfeed.tech/sources/coreweave-blog-posts.md>)

Topics: [skypilot](<https://devfeed.tech/topics/skypilot.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [NCCL](<https://devfeed.tech/topics/nccl.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-orchestration](<https://devfeed.tech/tags/ai-orchestration.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cloud-orchestration](<https://devfeed.tech/tags/cloud-orchestration.md>), [content-type](<https://devfeed.tech/tags/content-type.md>), [coreweave](<https://devfeed.tech/tags/coreweave.md>), [ease](<https://devfeed.tech/tags/ease.md>), [job](<https://devfeed.tech/tags/job.md>), [launch](<https://devfeed.tech/tags/launch.md>), [ml](<https://devfeed.tech/tags/ml.md>), [multi-cloud](<https://devfeed.tech/tags/multi-cloud.md>), [orchestration](<https://devfeed.tech/tags/orchestration.md>), [skypilot](<https://devfeed.tech/tags/skypilot.md>), [speed](<https://devfeed.tech/tags/speed.md>), [support](<https://devfeed.tech/tags/support.md>), [teams](<https://devfeed.tech/tags/teams.md>), [visibility](<https://devfeed.tech/tags/visibility.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

SkyPilot adds CoreWeave as a backend, letting teams launch training, development, and serving jobs on CoreWeave Kubernetes clusters through SkyPilot YAML or CLI. The article describes resource selection, job submission, retries, log streaming, and autoscaling, alongside existing CoreWeave orchestration options SUNK and Kueue. Setup includes following SkyPilot's CoreWeave installation steps, optionally enabling InfiniBand with `network_tier: best` and mounting CoreWeave Object Storage with the `cw://` scheme. The article cites an NCCL example reporting 3.6 Tb/s interconnect bandwidth and storage throughput up to 7 GB/s per GPU; these are presented as example results and stated capabilities.

### Source excerpt

Launch any AI job, anywhere, with minimal friction. CoreWeave adds SkyPilot support for multi-cloud orchestration, unlocking greater ease, speed, and visibility for ML teams.

## How to Sync Thousands of Dynamic Kubernetes Custom Resources Across Secure One-Way Network

DevFeed: [How to Sync Thousands of Dynamic Kubernetes Custom Resources Across Secure One-Way Network](<https://devfeed.tech/articles/how-to-sync-thousands-of-dynamic-kubernetes-custom-resources-across-secure-one-way-network-95219.md>)

Original publisher: [Read original article](<https://blog.devops.dev/how-to-sync-thousands-of-dynamic-kubernetes-custom-resources-across-secure-one-way-networks-cb851eeb1c64?source=rss----33f8b2d9a328---4>)

Author: Techie Crow

Published: 2026-10-22T08:54:22Z

Content type: article

Language: en

Sources: [DevOps.dev](<https://devfeed.tech/sources/devops-dev.md>)

Topics: [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Cloud Native Operations](<https://devfeed.tech/topics/cloud-native-operations.md>)

Tags: [apis](<https://devfeed.tech/tags/apis.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [devops](<https://devfeed.tech/tags/devops.md>), [gitops](<https://devfeed.tech/tags/gitops.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [network](<https://devfeed.tech/tags/network.md>), [security](<https://devfeed.tech/tags/security.md>), [sync](<https://devfeed.tech/tags/sync.md>)

### AI overview

The article frames a Kubernetes sync problem: a production engine dynamically creates hundreds or thousands of Custom Resources daily and registers them in a central Main Registry through a custom Proxy API. Ephemeral testing and host clusters must fetch configurations on demand, while security rules allow only one-way communication from target clusters to the Proxy over Private Links. The excerpt says traditional approaches face bottlenecks, and that the article examines three approaches before an optimized, stateless design; it does not provide their details or tradeoffs.

### Source excerpt

Managing custom APIs inside Kubernetes usually involves a predictable lifecycle. A developer writes a Custom Resource Definition (CRD)... Continue reading on DevOps.dev "

## EP-NCO: Latency-Aware Service Placement using Neural Combinatorial Optimisers for Edge-Cloud Systems

DevFeed: [EP-NCO: Latency-Aware Service Placement using Neural Combinatorial Optimisers for Edge-Cloud Systems](<https://devfeed.tech/articles/ep-nco-latency-aware-service-placement-using-neural-combinatorial-optimisers-for-edge-cloud-systems-95149.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.25553>)

Author: Kimia Abedpour, Mohammadsadeq Garshasbi Herabad, Zheng Li, Javid Taheri

Published: 2026-10-22T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>)

Topics: [Combinatorial optimization](<https://devfeed.tech/topics/combinatorial-optimization.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [Internet of things](<https://devfeed.tech/topics/iot.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [ablation](<https://devfeed.tech/tags/ablation.md>), [aggregate](<https://devfeed.tech/tags/aggregate.md>), [algorithm](<https://devfeed.tech/tags/algorithm.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [communication](<https://devfeed.tech/tags/communication.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [delay](<https://devfeed.tech/tags/delay.md>), [dependency](<https://devfeed.tech/tags/dependency.md>), [edge](<https://devfeed.tech/tags/edge.md>), [execution](<https://devfeed.tech/tags/execution.md>), [framework](<https://devfeed.tech/tags/framework.md>), [graph](<https://devfeed.tech/tags/graph.md>), [inference](<https://devfeed.tech/tags/inference.md>), [latency](<https://devfeed.tech/tags/latency.md>), [neural](<https://devfeed.tech/tags/neural.md>), [service](<https://devfeed.tech/tags/service.md>), [systems](<https://devfeed.tech/tags/systems.md>)

### AI overview

EP-NCO uses a graph-based neural combinatorial optimiser to place interdependent IoT microservices across edge-cloud platforms, accounting for execution latency, direct-link communication delay, and approximate bandwidth sharing among dependency flows. On held-out graph sizes from the same synthetic framework, the Large-trained variant with hard decoding had the lowest mean aggregate latency cost at every scale: 6.0%-12.5% below GA, 14.4%-20.8% below PSO, and 1.2%-9.4% below the strongest reinforcement-learning ablation. On Small instances, its mean cost was 10.5% below the best Gurobi solutions found and close to Gurobi's mean lower bound. EP-NCO inference took 0.88-1.03 seconds per instance versus 85-97 seconds for GA and PSO; the reported one-off training cost is recovered after roughly 1,000 placements.

### Source excerpt

arXiv:2606.25553v2 Announce Type: replace Abstract: Latency-sensitive Internet of Things (IoT) applications are increasingly deployed as interdependent microservices across heterogeneous edge-cloud infrastructures. Placing them under resource and network constraints is an NP-hard combinatorial optimisation problem. This study proposes EP-NCO, a graph-based neural combinatorial optimisation framework for service placement in compute continuum platforms. The objective accounts for execution latency, direct-link communication delay, and an approximate sharing of bandwidth among dependency flows that coexist in the resulting placement. Across held-out graph sizes generated by the same synthetic framework, the EP-NCO variant trained on Large instances with hard decoding achieved the lowest mean aggregate latency cost at every scale. It was 6.0%-12.5% lower than a genetic algorithm (GA), 14.4%-20.8% lower than particle swarm optimisation (PSO), and 1.2%-9.4% lower than the strongest reinforcement learning ablation. On Small instances, its mean cost was 10.5% below the best Gurobi solutions found and close to Gurobi's mean lower bound. The observed inference time of the EP-NCO variants was 0.88-1.03 s per instance, compared with 85-97 s for GA and PSO. As a result, the one-off training cost of the EP-NCO variant trained on Large instances with hard decoding is recovered after roughly 1,000 placements.

## StarMind: Challenges and Opportunities in Orbital Computing

DevFeed: [StarMind: Challenges and Opportunities in Orbital Computing](<https://devfeed.tech/articles/starmind-challenges-and-opportunities-in-orbital-computing-93849.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.26916>)

Author: Zisheng Gong, Zhaoyu Liu, Houtianfu Wang, O. Tansel Baydas, Liu Cao, Xia Xu, Ozgur B. Akan

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Networking](<https://devfeed.tech/sources/arxiv-networking.md>)

Topics: [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Big data infrastructure](<https://devfeed.tech/topics/big-data-infrastructure.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [case-study](<https://devfeed.tech/tags/case-study.md>), [challenges](<https://devfeed.tech/tags/challenges.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cs-ni](<https://devfeed.tech/tags/cs-ni.md>), [cs-sy](<https://devfeed.tech/tags/cs-sy.md>), [distributed](<https://devfeed.tech/tags/distributed.md>), [e2e](<https://devfeed.tech/tags/e2e.md>), [eess-sy](<https://devfeed.tech/tags/eess-sy.md>), [federated](<https://devfeed.tech/tags/federated.md>), [framework](<https://devfeed.tech/tags/framework.md>), [gateway](<https://devfeed.tech/tags/gateway.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [interfaces](<https://devfeed.tech/tags/interfaces.md>), [large](<https://devfeed.tech/tags/large.md>), [latency](<https://devfeed.tech/tags/latency.md>), [layers](<https://devfeed.tech/tags/layers.md>), [links](<https://devfeed.tech/tags/links.md>), [mission](<https://devfeed.tech/tags/mission.md>), [model](<https://devfeed.tech/tags/model.md>)

### AI overview

StarMind proposes a five-layer framework and six standardizable interfaces for coordinating distributed computing across heterogeneous orbital systems. In a representative SAR/InSAR case study, resource-aware task placement and gateway selection reduce end-to-end latency to 41 minutes, compared with 61 minutes for localized onboard processing and 146 minutes for ground processing. The improvement comes from reducing contact waits and urgent Earth downlinks, not from faster orbital processors. The approach is more effective with large source datasets, compact outputs, and sufficient optical inter-satellite link capacity; otherwise, conventional processing may be preferable. The article also identifies joint task partitioning and cross-domain orchestration, hierarchical federated intelligence, and large model-enabled intelligence as research opportunities. This overview reflects the supplied abstract.

### Source excerpt

arXiv:2610.26916v1 Announce Type: new Abstract: Space-based artificial intelligence (AI) is moving beyond isolated onboard processing toward constellation-scale computing over optical inter-satellite links (OISLs). Yet current infrastructures remain fragmented, making end-to-end (E2E) coordination across heterogeneous and multi-vendor systems difficult. This article examines the StarMind vision and develops a reference framework for distributed orbital computing. It organizes the system into five layers and defines six standardizable interfaces that link mission intent to resource conditions and trustworthy result delivery. The framework also clarifies the operational constraints that determine whether orbital cooperation is worthwhile. In a representative synthetic aperture radar/interferometric SAR (SAR/InSAR) case study, the framework reduces E2E latency to 41-min, compared with 61-min for localized onboard computing and 146-min for ground-based processing. This gain does not rely on faster processors in orbit. Instead, it comes from reducing contact waiting and urgent Earth downlink traffic through resource-aware task placement and gateway selection. Feasibility analysis shows that the proposed framework is more effective when source datasets are large, output products are compact, and optical inter-satellite link capacity is sufficient. When these conditions are absent, conventional processing may remain preferable. Finally, we discuss research opportunities in joint task partitioning and cross-domain orchestration, hierarchical federated intelligence, and large model-enabled intelligence to advance an interoperable space-ground computing continuum.

## New compute plans for memory-intensive applications

DevFeed: [New compute plans for memory-intensive applications](<https://devfeed.tech/articles/new-compute-plans-for-memory-intensive-applications-81906.md>)

Original publisher: [Read original article](<https://render.com/blog/new-compute-plans-for-memory-intensive-applications>)

Published: 2026-08-28T03:19:00Z

Content type: release

Language: en

Sources: [Render](<https://devfeed.tech/sources/render.md>)

Topics: [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [compute](<https://devfeed.tech/tags/compute.md>), [costs](<https://devfeed.tech/tags/costs.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [efficiently](<https://devfeed.tech/tags/efficiently.md>), [memory](<https://devfeed.tech/tags/memory.md>), [memory-pressure](<https://devfeed.tech/tags/memory-pressure.md>)

### AI overview

Render introduces more than 15 memory-optimized compute plans, allowing services to choose memory independently of CPU. The release also adds a 12-CPU tier with up to 96 GB of RAM and explains how to use service metrics to select plans for memory- or CPU-bound workloads.

### Source excerpt

Today, we're adding over 15 new memory-optimized compute plans for Render services. These plans are designed specifically to help you run agents and other memory-intensive workloads efficiently.

## Scaling Kubernetes at Mercado Libre with Karpenter and GitOps

DevFeed: [Scaling Kubernetes at Mercado Libre with Karpenter and GitOps](<https://devfeed.tech/articles/scaling-kubernetes-at-mercado-libre-with-karpenter-and-gitops-22556.md>)

Original publisher: [Read original article](<https://medium.com/mercadolibre-tech/scaling-kubernetes-at-mercado-libre-with-karpenter-and-gitops-2c792a7403c5?source=rss----5011f85401f0---4>)

Author: Juliano Marcos Martins

Published: 2025-10-28T13:10:37Z

Content type: article

Language: en

Sources: [Mercado Libre Tech](<https://devfeed.tech/sources/mercado-libre-tech.md>)

Topics: [Kubernetes](<https://devfeed.tech/topics/kubernetes.md>), [GitOps](<https://devfeed.tech/topics/gitops.md>), [Deployment & Scaling](<https://devfeed.tech/topics/deployment-scaling.md>), [Cloud Native Operations](<https://devfeed.tech/topics/cloud-native-operations.md>), [cloud cost management](<https://devfeed.tech/topics/cloud-cost-management.md>), [internal developer platform automation](<https://devfeed.tech/topics/internal-developer-platform-automation.md>), [platform engineering automation](<https://devfeed.tech/topics/platform-engineering-automation.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>), [internal developer platform](<https://devfeed.tech/topics/internal-developer-platform.md>), [Cloud Architecture](<https://devfeed.tech/topics/cloud-architecture.md>), [Microservices](<https://devfeed.tech/topics/microservices.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>)

Tags: [automation](<https://devfeed.tech/tags/automation.md>), [cloud-native](<https://devfeed.tech/tags/cloud-native.md>), [cloud-native-operations](<https://devfeed.tech/tags/cloud-native-operations.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [compute](<https://devfeed.tech/tags/compute.md>), [compute-resources](<https://devfeed.tech/tags/compute-resources.md>), [container](<https://devfeed.tech/tags/container.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [cost](<https://devfeed.tech/tags/cost.md>), [cost-optimization](<https://devfeed.tech/tags/cost-optimization.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [declarative](<https://devfeed.tech/tags/declarative.md>), [delivery](<https://devfeed.tech/tags/delivery.md>), [demand](<https://devfeed.tech/tags/demand.md>), [deployments](<https://devfeed.tech/tags/deployments.md>), [devops](<https://devfeed.tech/tags/devops.md>), [gitops](<https://devfeed.tech/tags/gitops.md>), [karpenter](<https://devfeed.tech/tags/karpenter.md>), [kubernetes](<https://devfeed.tech/tags/kubernetes.md>), [mercado](<https://devfeed.tech/tags/mercado.md>), [nodepool](<https://devfeed.tech/tags/nodepool.md>), [scaling](<https://devfeed.tech/tags/scaling.md>)

### AI overview

Mercado Libre describes using Karpenter to provision Kubernetes nodes on demand, evaluating pod constraints and selecting node pools, capacity types, and zones; batching and bin packing aim to improve utilization. The company reports average pod scheduling fell to about 50 seconds. It organizes capacity through NodePools and NodeClasses, with Spot limits varying by workload priority and topology constraints distributing pods across zones. GitOps with FluxCD and custom pipelines reconciles Git-defined state, while pull request reviews, validation, and staged rollouts support auditability and rollback. Its Automatrix platform generates cluster provisioning and upgrade workflows and integrates corporate standards with GitOps. The account presents these mechanisms as ways to scale operations while managing cost, resilience, and governance.

### Source excerpt

Mercado Libre, Latin America's largest digital ecosystem, runs one of the most demanding and scalable infrastructures in the Southern Hemisphere. Our operation at scale 35,000 active microservices Over 20,000 technology professionals Approximately 30,000 daily deployments Around 120,000 pull requests per day Traffic peaks of 15 million requests per second In this fast-paced environment, efficient workload orchestration across multiple cloud providers is crucial. Kubernetes has become our standard for container management and is the foundation of our multi-cloud strategy, powering our internal compute engine. You can see more details here. This article explores how we've used Karpenter and GitOps to evolve our ecosystem, achieving automated provisioning, declarative governance, and large-scale cloud-native operations. The challenge of scale With tens of thousands of microservices, ticket-driven workflows and manual changes quickly became bottlenecks, slowing delivery and increasing operational risk. Even with the evolution of Infrastructure as Code (IaC), new challenges emerged: Explosive diversity of templates and pipelines State drift and lack of standardization Overloaded Platform/DevOps teams with repetitive requests In response, we evolved toward declarative, auditable, and fully automated practices, guided by GitOps, continuous automation, and internal infrastructure abstractions. Karpenter: What it is and why it changed the game Karpenter is an open-source, Kubernetes-native provisioning engine designed to deliver automatic scalability, cost optimization, and high flexibility in managing a cluster's data plane (compute nodes). How it works at a high level Monitors the Kubernetes scheduler for pending pods caused by insufficient compute resources. Analyzes each workload's requirements (CPU, memory, architecture, affinities, labels, tolerations, and more). Provisions, scales, or terminates nodes on demand, automatically selecting the ideal instance type, profile

## Multi-service templates

DevFeed: [Multi-service templates](<https://devfeed.tech/articles/multi-service-templates-95621.md>)

Original publisher: [Read original article](<https://blog.railway.com/p/multi-service-templates>)

Author: Faraz Patankar

Published: 2022-03-08T00:00:00Z

Content type: release

Language: en

Sources: [Railway](<https://devfeed.tech/sources/railway.md>)

Topics: [self-service](<https://devfeed.tech/topics/self-service.md>), [internal developer platform automation](<https://devfeed.tech/topics/internal-developer-platform-automation.md>), [resource orchestration](<https://devfeed.tech/topics/resource-orchestration.md>)

Tags: [cloud](<https://devfeed.tech/tags/cloud.md>), [configuration](<https://devfeed.tech/tags/configuration.md>), [deploy](<https://devfeed.tech/tags/deploy.md>), [environment-variables](<https://devfeed.tech/tags/environment-variables.md>), [github](<https://devfeed.tech/tags/github.md>), [manage](<https://devfeed.tech/tags/manage.md>), [management](<https://devfeed.tech/tags/management.md>), [monorepo](<https://devfeed.tech/tags/monorepo.md>), [railway](<https://devfeed.tech/tags/railway.md>), [service](<https://devfeed.tech/tags/service.md>), [templates](<https://devfeed.tech/tags/templates.md>)

### AI overview

Railway's multi-service templates let users deploy related services and databases within one project, replacing the need for a separate project per service. Template creators can add services from GitHub URLs and define environment variables plus defaults, root directories, start commands, and healthcheck paths. For monorepos, service setup is configured once and deployment creates one GitHub repo using the creator's service configurations. Existing projects can generate templates from their services, plugins, and environment variable keys; creators can then edit, share, or delete their templates.

### Source excerpt

An overview of the recently launched multi-service templates and everything you can do with them.