Searcharxiv⌕ Search

arXiv subjects

Niloofar Gholipour

Publications and source records attributed to Niloofar Gholipour.

5 recordsLinked to original sources

TAILROLL: Tail-Aware Rollout Acceleration for RL-Based LLM Post-Training

RL post-training of large language models (LLMs) is bottlenecked by rollout: heavy-tailed responses leave engines idle as each step waits for the longest response. Asynchronous training masks, but does not eliminate, idle time. We distinguish imbalance waste (uneven placement and decode contention) from straggler waste (waiting for the longest response). Tail length, not average load, often sets the step bound. We propose TAILROLL, an online scheduler integrated into an open-source RL framework. It combines stateless in-flight migration and sufficiency-gated termination. Migration moves decoding requests between engines at chunk boundaries and resumes them with one re-prefill under a freeness-based, cost-aware tail-balancing policy. Termination stops a response when its GRPO group has enough completed samples for advantage estimation, masking its partial output from the training loss. Across ten configurations, TAILROLL reduces rollout bubble ratio by 21% to 74% and cross-engine KV-cache imbalance by 44% to 79%. In end-to-end RL training, it cuts rollout time by 24% to 34%, full step time by 21% to 22% on long-CoT RL, and p95 request latency by 28% to 37%, while raising generation throughput by up to 33% with comparable or higher final reward.

cs.DC↗

Rollout Efficiency in Reinforcement Learning for Reasoning Large Language Models: A Taxonomy and Future Directions

Reasoning-oriented reinforcement learning enables large language models to solve mathematical, coding, and other multi-step tasks, but shifts a substantial portion of the training cost to rollout, where trajectories are generated for policy updates. Efficient rollout mechanisms are therefore essential to reduce this cost while maintaining the freshness, consistency, and statistical validity of training data. This survey provides a systematic taxonomy of recent research on rollout efficiency for reasoning-oriented reinforcement learning, classifying existing approaches from both mechanism and bottleneck perspectives. Based on this taxonomy, we analyze how different technique families address distinct sources of rollout inefficiency, examine opportunities and potential conflicts for combining them, identify gaps in the evaluation and reporting of efficiency gains, and discuss open challenges and future research directions.

cs.AI↗

HFX: Joint Design of Algorithms and Systems for Multi-SLO Serving and Fast Scaling

Large language model (LLM) serving faces the dual challenge of meeting strict user-specific service-level objectives (SLOs) while minimizing computational cost under dynamic, multi-task workloads. Existing approaches either rely on static scheduling policies or focus on single-task settings, limiting their applicability in real-world deployments with heterogeneous requests, variable prompt lengths, and elastic scaling requirements. We present HFX, a production LLM serving system that jointly optimizes request scheduling and elastic scaling across model replicas to satisfy diverse SLOs. HFX introduces a \textbf{scheduler} that performs proactive budget estimation and prioritization to ensure SLO compliance for both new and in-flight requests. HFX also integrates a \textbf{scaler} that supports fast device-to-device (D2D) weight transfer, reducing cold-start latency. Additionally, the system supports both colocated and disaggregated prefill/decode deployments, enabling adaptation to diverse workload patterns and cloud environments. Through extensive experiments on multi-task workloads, we demonstrate consistently higher SLO attainment, lower end-to-end latency, and lower NPU usage cost by up to 4.44$\times$, 65.82\%, and 49.81\%, respectively, compared to state-of-the-art systems. Our results highlight the effectiveness of SLO-aware scheduling and scaling in practical LLM serving, providing a robust framework for cost-efficient and SLO-compliant deployments.

cs.DC↗

TPTO: A Transformer-PPO based Task Offloading Solution for Edge Computing Environments

Emerging applications in healthcare, autonomous vehicles, and wearable assistance require interactive and low-latency data analysis services. Unfortunately, cloud-centric architectures cannot fulfill the low-latency demands of these applications, as user devices are often distant from cloud data centers. Edge computing aims to reduce the latency by enabling processing tasks to be offloaded to resources located at the network's edge. However, determining which tasks must be offloaded to edge servers to reduce the latency of application requests is not trivial, especially if the tasks present dependencies. This paper proposes a DRL approach called TPTO, which leverages Transformer Networks and PPO to offload dependent tasks of IoT applications in edge computing. We consider users with various preferences, where devices can offload computation to an edge server via wireless channels. Performance evaluation results demonstrate that under fat application graphs, TPTO is more effective than state-of-the-art methods, such as Greedy, HEFT, and MRLCO, by reducing latency by 30.24%, 29.61%, and 12.41%, respectively. In addition, TPTO presents a training time approximately 2.5 times faster than an existing DRL approach.

cs.DC↗

Recent Advances in Energy Efficient Resource Management Techniques in Cloud Computing Environments

Nowadays cloud computing adoption as a form of hosted application and services is widespread due to decreasing costs of hardware, software, and maintenance. Cloud enables access to a shared pool of virtual resources hosted in large energy-hungry data centers for diverse information and communication services with dynamic workloads. The huge energy consumption of cloud data centers results in high electricity bills as well as emission of a large amount of carbon dioxide gas. Needless to say, efficient resource management in cloud environments has become one of the most important priorities of cloud providers and consequently has increased the interest of researchers to propose novel energy saving solutions. This chapter presents a scientific and taxonomic survey of recent energy efficient cloud resource management' solutions in cloud environments. The main objective of this study is to propose a novel complete taxonomy for energy-efficient cloud resource management solutions, review recent research advancements in this area, classify the existing techniques based on our proposed taxonomy, and open up new research directions. Besides, it reviews and surveys the literature in the range of 2015 through 2021 in the subject of energy-efficient cloud resource management techniques and maps them to its proposed taxonomy, which unveils novel research directions and facilitates the conduction of future researches.

cs.DC↗