# When Is Availability-Aware Training Worth It? A Benchmark and Empirical Study of Interruption-Resilient Optimization Under Predictable Compute Schedules

arXiv: https://arxiv.org/abs/2609.22087 · HTML: https://mlsanity.com/paper/2609.22087/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 18 of its 27 references that are in the map.
- First posted on arXiv 22 Jun 2026.

## Builds on (most cited first, 18 of 18)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [SGDR: Stochastic Gradient Descent with Warm Restarts](https://mlsanity.com/paper/1608.03983/) (2016)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Do ImageNet Classifiers Generalize to ImageNet?](https://mlsanity.com/paper/1902.10811/) (2019)
- [ZeRO: Memory Optimizations Toward Training Trillion Parameter Models](https://mlsanity.com/paper/1910.02054/) (2019)
- [Averaging Weights Leads to Wider Optima and Better Generalization](https://mlsanity.com/paper/1803.05407/) (2018)
- [Cyclical Learning Rates for Training Neural Networks](https://mlsanity.com/paper/1506.01186/) (2015)
- [A Metric Learning Reality Check](https://mlsanity.com/paper/2003.08505/) (2020)
- [Are GANs Created Equal? A Large-Scale Study](https://mlsanity.com/paper/1711.10337/) (2017)
- [DiLoCo: Distributed Low-Communication Training of Language Models](https://mlsanity.com/paper/2311.08105/) (2023)
- [Asynchronous Parallel Stochastic Gradient for Nonconvex Optimization](https://mlsanity.com/paper/1506.08272/) (2015)
- [Asynchronous Stochastic Gradient Descent with Delay Compensation](https://mlsanity.com/paper/1609.08326/) (2016)
- [Bamboo: Making Preemptible Instances Resilient for Affordable Training of Large DNNs](https://mlsanity.com/paper/2204.12013/) (2022)
- [Varuna: Scalable, Low-cost Training of Massive Deep Learning Models](https://mlsanity.com/paper/2111.04007/) (2021)
- [Oobleck: Resilient Distributed Training of Large Models Using Pipeline Templates](https://mlsanity.com/paper/2309.08125/) (2023)
- [FedSpace: An Efficient Federated Learning Framework at Satellites and Ground Stations](https://mlsanity.com/paper/2202.01267/) (2022)
- [On-Board Federated Learning for Dense LEO Constellations](https://mlsanity.com/paper/2111.12769/) (2021)
- [Energy-Aware Federated Learning in Satellite Constellations](https://mlsanity.com/paper/2409.14832/) (2024)

## Abstract (arXiv)

Training under non-stationary but predictable compute availability (satellites under eclipse, duty-cycled edge devices, power-capped datacenters) is often framed as needing specialized, availability-aware optimizers. We test that premise. We release OrbitTrace, a benchmark of 50 physics-grounded availability traces from SGP4 propagation of live two-line element sets across three orbital regimes, and ask a falsifiable question: when an availability gap interrupts training, is a specialized resumption strategy worth it, or is competent checkpoint-and-resume enough? Our central finding: when optimizer state can be preserved across a gap, the gap is essentially free. A strong checkpoint baseline that restores full optimizer state and indexes its learning-rate schedule in effective (active) time matches uninterrupted training to within data-ordering noise on CIFAR-10/ResNet-18, and exactly on a GPT-2/AdamW task. Advantages previously reported for availability-aware methods, including our own three-pillar method AAT, arise almost entirely from comparison against a weak baseline that indexes its schedule on wall-clock time. Against the strong baseline, reactive adaptation provides no advantage across stationary, optimizer-state-loss, and distribution-drift regimes. We isolate one narrow regime where it helps: for large models whose optimizer state cannot be persisted across gaps and that are interrupted by frequent, short pauses, reconstructing a decayed optimizer moment recovers only ~21% of the state-loss penalty on average (and not robustly across seeds); this vanishes for eclipse-scale gaps, where the decayed moment is indistinguishable from zero. Our contributions are a benchmark, a strong reproducible baseline protocol, and a clear characterization of when interruption-resilient optimization is worth its complexity, and when it is not.