Minlan Yu

Zhiying Xu, Francis Y. Yan, Rachee Singh et al.

The rapid expansion of global cloud wide-area networks (WANs) has posed a challenge for commercial optimization engines to efficiently solve network traffic engineering (TE) problems at scale. Existing acceleration strategies decompose TE optimization into concurrent subproblems but realize limited parallelism due to an inherent tradeoff between run time and allocation performance. We present Teal, a learning-based TE algorithm that leverages the parallel processing power of GPUs to accelerate TE control. First, Teal designs a flow-centric graph neural network (GNN) to capture WAN connectivity and network flows, learning flow features as inputs to downstream allocation. Second, to reduce the problem scale and make learning tractable, Teal employs a multi-agent reinforcement learning (RL) algorithm to independently allocate each traffic demand while optimizing a central TE objective. Finally, Teal fine-tunes allocations with ADMM (Alternating Direction Method of Multipliers), a highly parallelizable optimization algorithm for reducing constraint violations such as overutilized links. We evaluate Teal using traffic matrices from Microsoft's WAN. On a large WAN topology with >1,700 nodes, Teal generates near-optimal flow allocations while running several orders of magnitude faster than the production optimization engine. Compared with other TE acceleration schemes, Teal satisfies 6--32% more traffic demand and yields 197--625x speedups.

Minghao Li, Ran Ben Basat, Shay Vargaftik et al.

Deep neural networks (DNNs) are the de facto standard for essential use cases, such as image classification, computer vision, and natural language processing. As DNNs and datasets get larger, they require distributed training on increasingly larger clusters. A main bottleneck is the resulting communication overhead where workers exchange model updates (i.e., gradients) on a per-round basis. To address this bottleneck and accelerate training, a widely-deployed approach is compression. However, previous deployments often apply bi-directional compression schemes by simply using a uni-directional gradient compression scheme in each direction. This results in significant computational overheads at the parameter server and increased compression error, leading to longer training and lower accuracy. We introduce Tensor Homomorphic Compression (THC), a novel bi-directional compression framework that enables the direct aggregation of compressed values and thus eliminating the aforementioned computational overheads. Moreover, THC is compatible with in-network aggregation (INA), which allows for further acceleration. Our evaluation shows that training representative vision and language models with THC reaches target accuracy by 1.40x to 1.47x faster using INA and 1.28x to 1.33x faster using a software PS compared with state-of-the-art systems.

Weiqi Feng, Yangrui Chen, Shaoyu Wang et al.

Multimodal large language models (MLLMs) have extended the success of large language models (LLMs) to multiple data types, such as image, text and audio, achieving significant performance in various domains, including multimodal translation, visual question answering and content generation. Nonetheless, existing systems are inefficient to train MLLMs due to substantial GPU bubbles caused by the heterogeneous modality models and complex data dependencies in 3D parallelism. This paper proposes Optimus, a distributed MLLM training system that reduces end-to-end MLLM training time. Optimus is based on our principled analysis that scheduling the encoder computation within the LLM bubbles can reduce bubbles in MLLM training. To make scheduling encoder computation possible for all GPUs, Optimus searches the separate parallel plans for encoder and LLM, and adopts a bubble scheduling algorithm to enable exploiting LLM bubbles without breaking the original data dependencies in the MLLM model architecture. We further decompose encoder layer computation into a series of kernels, and analyze the common bubble pattern of 3D parallelism to carefully optimize the sub-millisecond bubble scheduling, minimizing the overall training time. Our experiments in a production cluster show that Optimus accelerates MLLM training by 20.5%-21.3% with ViT-22B and GPT-175B model over 3072 GPUs compared to baselines.

Minghao Li, Alicia Golden, Samuel Hsia et al.

The rapid scaling of large language model training requires distributing GPU resources across multiple data center buildings and regions. We refer to such paradigm as "scale-across" training. As infrastructure expands, the system design space becomes increasingly intricate, encompassing new model architectures, hardware heterogeneity, and evolving communication patterns. Drawing from Meta's production experience, we highlight the complexities of deploying training jobs across a few data centers housing hundreds of thousands of GPUs. To accelerate exploration of the large design space and to enable efficient training for frontier model development, we conduct in-depth characterization of three key design dimensions: parallelism placement, parallelism scheduling, and network layer technologies. We then propose ScaleAcross Explorer, an optimizer that considers the interplay of design dimensions and holistically optimizes scale-across training. Testbed experiments and simulations demonstrate up to 64.62% training speedups over production configuration and up to 37.59% training speedups over the state-of-the-art baseline across a wide range of design points.

Weifan Jiang, Rana Shahout, Minghao Li et al.

Multi-Agent Debate (MAD) improves LLM-agent accuracy but suffers from rapid context growth, limiting scalability in larger multi-agent settings. Existing methods prune low-utility communications using prior signals, such as token-level log-likelihoods or LLM self-reported confidence. However, these signals become unreliable under hallucination, degrading the accuracy of MAD methods that rely on them. We propose SVR-MAD, a Bayesian-inspired MAD framework that treats pre-debate signals as priors and debate outcomes as posterior-style evidence for estimating agent correctness. SVR-MAD uses this evidence to incrementally construct the communication graph, prioritizing agents whose answers survive peer challenges. Experiments across multiple LLMs and benchmarks show that SVR-MAD reduces token cost by up to 61% while matching or improving accuracy relative to the most accurate competing MAD baseline.

Can Hankendi, Rana Shahout, Minlan Yu et al.

Large language model (LLM) inference has become a dominant workload in modern data centers, driving significant GPU utilization and energy consumption. While prior systems optimize throughput and latency by batching, scheduling, and parallelism, they largely treat GPU power as a static constraint rather than a controllable resource. In this paper, we present a power-aware runtime for LLM serving, PALS, that treats GPU power caps as a first-class control knob and jointly optimizes them with software parameters such as batch size. The system combines lightweight offline power-performance models with a feedback-driven controller to select configurations that satisfy throughput targets while maximizing energy efficiency. We implement PALS within an existing LLM serving framework, vLLM, demonstrating that it requires no model retraining or API changes. Across multi-GPU systems and both dense and mixture-of-experts (MoE) models, PALS improves energy efficiency by up to 26.3%, reduces QoS violations by 4x to 7x under power constraints, and tracks dynamic power budgets. These results highlight the potential of integrating power control directly into LLM inference runtimes, enabling energy-proportional and grid-interactive AI systems.

Shaoke Xi, ChonLam Lao, Boyi Jia et al.

Large language model (LLM) training today runs on clusters spanning thousands of GPUs. While this scale enables rapid model advances, developing, debugging, and performance-tuning the training framework inevitably becomes complex and costly. This is because engineers often need to reproduce production behaviors to diagnose failures or evaluate optimizations, thereby demanding frequent and even exclusive access to production-scale clusters -- which becomes increasingly hard given that the majority of GPUs are already committed to production workloads. Simulation relies on complex performance models that are difficult to maintain, and downscaled experiments often fail to capture scale-dependent behaviors. We present PrismLLM to decouple large-scale execution from the need to access large clusters, enabling engineers to run and observe ranks of interest under faithful large-scale behavior using only a few GPUs. PrismLLM constructs a high-fidelity execution graph via a slicing-based approach that captures computation, communication, and dependencies of the target scale. Then, PrismLLM performs hybrid emulation where selected ranks execute the original program while the remaining ranks are replayed as virtual participants. Experiments on large-scale LLM training workloads show that PrismLLM accurately reproduces performance and memory behavior, achieving only 0.58\% average error in iteration time and less than 0.01\% error in peak GPU memory usage. PrismLLM can emulate clusters of up to 8192 GPUs using fewer than 1\% of the physical GPUs required by the original deployment.

Rana Shahout, Hayder Tirmazi, Minlan Yu et al.

We introduce Orla, a library for constructing and running LLM-based agentic systems. Modern agentic applications consist of workflows that combine multiple LLM inference steps, tool calls, and heterogeneous infrastructure. Today, developers typically build these systems by manually composing orchestration code with LLM serving engines and tool execution logic. Orla provides a general abstraction that separates request execution from workflow-level policy. It acts as a serving layer above existing LLM inference engines: developers define workflows composed of stages, while Orla manages how those stages are mapped, executed, and coordinated across models and backends. It provides agent-level control through three mechanisms: a stage mapper, which assigns each stage to an appropriate model and backend; a workflow orchestrator, which schedules stages and manages their resources and context; and a memory manager, which manages inference state such as the KV cache across workflow boundaries. We demonstrate Orla with a customer support workflow that exercises many of its capabilities. We evaluate Orla on two datasets, showing that stage mapping improves latency and cost compared to a single-model vLLM baseline, while workflow-level cache management reduces time-to-first-token.

Subir Majumder, Minlan Yu, Le Xie

Artificial intelligence (AI) is driving rapid growth in electricity demand, yet the grid-facing power dynamics of AI data centers remain poorly understood. Here we show that, in shared-GPU systems, the composition of batch and inference workloads decouples aggregate power variability from short-horizon ramping. As the inference share rises, variability becomes U-shaped, whereas ramping becomes hump-shaped, particularly under higher loading. The magnitude and turning points of these patterns also depend on system loading. Using a trace-calibrated framework linking workload arrivals, queueing, scheduling, and GPU power, we show that the underlying mechanism is asymmetric. At intermediate workload mixes, queued batch jobs fill capacity left idle by fluctuating inference demand, reducing aggregate power variability. However, short-horizon ramping remains elevated because inference-side fluctuations propagate more directly into realized power. AI data centers should therefore be understood as dynamic systems whose workload composition shapes their grid impact.

Noman Bashir, Rob Sherwood, Le Xie et al.

For over a century, the electric grid has relied on a single statistical assumption: \emph{load diversity}, the principle that the uncorrelated demands of millions of small consumers produce a smooth, predictable aggregate. AI training data centers break that assumption. A single hyperscale training campus can draw power comparable to a mid-sized city, driven by one tightly synchronized job whose demand swings by hundreds of megawatts in seconds. This paper argues that the resulting entanglement of compute and power infrastructure requires a shift from implicit coexistence to explicit co-development between the historically decoupled data center and electric power industries. We introduce the distinct design principles, operational philosophies, and economic incentives of each sector, and show why their cultural and technical misalignment makes coordination difficult. We identify key research directions, from joint capacity planning, multi-timescale control, a compute--power protocol stack, to market innovation, that must be pursued to power the future of AI sustainably and reliably.

Xuanlin Jiang, Yang Zhou, Shiyi Cao et al.

Online LLM inference powers many exciting applications such as intelligent chatbots and autonomous agents. Modern LLM inference engines widely rely on request batching to improve inference throughput, aiming to make it cost-efficient when running on expensive GPU accelerators. However, the limited GPU memory has largely limited the batch size achieved in practice, leaving significant GPU compute resources wasted. We present NEO, an online LLM inference system that offloads part of attention compute and KV cache states from the GPU to the local host CPU, effectively increasing the GPU batch size and thus inference throughput. To this end, NEO proposes asymmetric GPU-CPU pipelining and load-aware scheduling to balance GPU and CPU loads and fully utilize their compute and memory resources. We evaluate NEO on a wide range of workloads (i.e., code generation, text summarization), GPUs (i.e., T4, A10G, H100), and LLM models (i.e., 7B, 8B, 70B). NEO achieves up to 7.5$\times$, 26%, and 14% higher throughput compared to GPU-only approach on T4, A10G, and H100 GPUs, respectively, while maintaining the same latency; with more powerful CPUs, NEO achieves up to 79.3% throughput gain on A10G GPU.

Yangtao Deng, Xiang Shi, Zhuo Jiang et al.

Large-scale distributed model training requires simultaneous training on up to thousands of machines. Faulty machine detection is critical when an unexpected fault occurs in a machine. From our experience, a training task can encounter two faults per day on average, possibly leading to a halt for hours. To address the drawbacks of the time-consuming and labor-intensive manual scrutiny, we propose Minder, an automatic faulty machine detector for distributed training tasks. The key idea of Minder is to automatically and efficiently detect faulty distinctive monitoring metric patterns, which could last for a period before the entire training task comes to a halt. Minder has been deployed in our production environment for over one year, monitoring daily distributed training tasks where each involves up to thousands of machines. In our real-world fault detection scenarios, Minder can accurately and efficiently react to faults within 3.6 seconds on average, with a precision of 0.904 and F1-score of 0.893.

Rana Shahout, Cong Liang, Shiji Xin et al.

Augmented Large Language Models (LLMs) enhance the capabilities of standalone LLMs by integrating external data sources through API calls. In interactive LLM applications, efficient scheduling is crucial for maintaining low request completion times, directly impacting user engagement. However, these augmentations introduce scheduling challenges due to the need to manage limited memory for cached information (KV caches). As a result, traditional size-based scheduling algorithms, such as Shortest Job First (SJF), become less effective at minimizing completion times. Existing work focuses only on handling requests during API calls by preserving, discarding, or swapping memory without considering how to schedule requests with API calls. In this paper, we propose LAMPS, a novel LLM inference framework for augmented LLMs. LAMPS minimizes request completion time through a unified scheduling approach that considers the total length of requests and their handling strategies during API calls. Recognizing that LLM inference is memory-bound, our approach ranks requests based on their consumption of memory over time, which depends on both the output sizes and how a request is managed during its API calls. To implement our scheduling, LAMPS predicts the strategy that minimizes memory waste of a request during its API calls, aligning with but improving upon existing approaches. We also propose starvation prevention techniques and optimizations to mitigate the overhead of our scheduling. We implement LAMPS on top of vLLM and evaluate its performance against baseline LLM inference systems, demonstrating improvements in end-to-end latency by 27%-85% and reductions in TTFT by 4%-96% compared to the existing augmented-LLM system, with even greater gains over vLLM.

Benjamin C. Lee, David Brooks, Arthur van Benthem et al.

Computing is at a moment of profound opportunity. Emerging applications -- such as capable artificial intelligence, immersive virtual realities, and pervasive sensor systems -- drive unprecedented demand for computer. Despite recent advances toward net zero carbon emissions, the computing industry's gross energy usage continues to rise at an alarming rate, outpacing the growth of new energy installations and renewable energy deployments. A shift towards sustainability is needed to spark a transformation in how computer systems are manufactured, allocated, and consumed. Carbon Connect envisions coordinated research thrusts that produce design and management strategies for sustainable, next-generation computer systems. These strategies must flatten and then reverse growth trajectories for computing power and carbon for society's most rapidly growing applications such as artificial intelligence and virtual spaces. We will require accurate models for carbon accounting in computing technology. For embodied carbon, we must re-think conventional design strategies -- over-provisioned monolithic servers, frequent hardware refresh cycles, custom silicon -- and adopt life-cycle design strategies that more effectively reduce, reuse and recycle hardware at scale. For operational carbon, we must not only embrace renewable energy but also design systems to use that energy more efficiently. Finally, new hardware design and management strategies must be cognizant of economic policy and regulatory landscape, aligning private initiatives with societal goals. Many of these broader goals will require computer scientists to develop deep, enduring collaborations with researchers in economics, law, and industrial ecology to spark change in broader practice.

Yangtao Deng, Lei Zhang, Qinlong Wang et al.

Reliability is essential for ensuring efficiency in LLM training. However, many real-world reliability issues remain difficult to resolve, resulting in wasted resources and degraded model performance. Unfortunately, today's collective communication libraries operate as black boxes, hiding critical information needed for effective root cause analysis. We propose Mycroft, a lightweight distributed tracing and root cause analysis system designed to address previously hidden reliability issues in collective communication. Mycroft's key idea is to trace collective communication states and leverage internal control and data dependencies to resolve reliability problems in LLM training. Mycroft has been deployed at ByteDance for over six months to debug collective communication related issues at runtime. It detected anomalies within 15 seconds in 90% of cases and identified the root cause within 20 seconds in 60% of cases. We also conducted extensive fault injection experiments to demonstrate Mycroft's capability and efficiency.

Jiawei Zhou, Woojeong Kim, Zhiying Xu et al.

Understanding the traffic dynamics in networks is a core capability for automated systems to monitor and analyze networking behaviors, reducing expensive human efforts and economic risks through tasks such as traffic classification, congestion prediction, and attack detection. However, it is still challenging to accurately model network traffic with machine learning approaches in an efficient and broadly applicable manner. Task-specific models trained from scratch are used for different networking applications, which limits the efficiency of model development and generalization of model deployment. Furthermore, while networking data is abundant, high-quality task-specific labels are often insufficient for training individual models. Large-scale self-supervised learning on unlabeled data provides a natural pathway for tackling these challenges. We propose to pre-train a general-purpose machine learning model to capture traffic dynamics with only traffic data from NetFlow records, with the goal of fine-tuning for different downstream tasks with small amount of labels. Our presented NetFlowGen framework goes beyond a proof-of-concept for network traffic pre-training and addresses specific challenges such as unifying network feature representations, learning from large unlabeled traffic data volume, and testing on real downstream tasks in DDoS attack detection. Experiments demonstrate promising results of our pre-training framework on capturing traffic dynamics and adapting to different networking tasks.

ChonLam Lao, Minlan Yu, Aditya Akella et al.

Large-scale ML training jobs are frequently interrupted by hardware and software anomalies, failures, and management events. Existing solutions like checkpointing or runtime reconfiguration suffer from long downtimes, degraded performance, or undesired changes to training strategies. We present TrainMover, a resilient runtime that leverages standby machines to handle interruptions with minimal downtime and zero memory overhead. To achieve these goals, TrainMover introduces two key techniques: two-phase, delta-based communication group setups and communication-free sandboxed shadow iterations. Our evaluation shows that TrainMover consistently achieves second-level downtime across all evaluated models during migration, maintaining 99\% training efficiency during periodic 10-minute rebalancing. We also demonstrate the effectiveness of TrainMover in handling various interruptions.

Minghao Li, Dmitrii Avdiukhin, Rana Shahout et al.

Federated Learning (FL) trains deep models across edge devices without centralizing raw data, preserving user privacy. However, client heterogeneity slows down convergence and limits global model accuracy. Clustered FL (CFL) mitigates this by grouping clients with similar representations and training a separate model for each cluster. In practice, client data evolves over time, a phenomenon we refer to as data drift, which breaks cluster homogeneity and degrades performance. Data drift can take different forms depending on whether changes occur in the output values, the input features, or the relationship between them. We propose FIELDING, a CFL framework for handling diverse types of data drift with low overhead. FIELDING detects drift at individual clients and performs selective re-clustering to balance cluster quality and model performance, while remaining robust to malicious clients and varying levels of heterogeneity. Experiments show that FIELDING improves final model accuracy by 1.9-5.9% and achieves target accuracy 1.16x-2.23x faster than existing state-of-the-art CFL methods.

Sherman Wong, Zhenting Qi, Zhaodong Wang et al.

Real-world software engineering tasks require coding agents that can operate on massive repositories, sustain long-horizon sessions, and reliably coordinate complex toolchains at test time. Existing research-grade coding agents offer transparency but struggle when scaled to heavier, production-level workloads, while production-grade systems achieve strong practical performance but provide limited extensibility, interpretability, and controllability. We introduce the Confucius Code Agent (CCA), a software engineering agent that can operate at large-scale codebases. CCA is built on top of the Confucius SDK, an agent development platform structured around three complementary perspectives: Agent Experience (AX), User Experience (UX), and Developer Experience (DX). The SDK supports a unified orchestrator with advanced context management for long-context reasoning, a persistent note-taking system for cross-session continual learning, and a modular extension system for reliable tool use. In addition, we introduce a meta-agent that automates the construction, evaluation, and refinement of agents through a build-test-improve cycle, enabling rapid agent development on new tasks and tool stacks. Instantiated on the Confucius SDK using the meta-agent, CCA demonstrates strong performance on real-world software engineering tasks. On SWE-Bench-Pro, CCA achieves a Resolve@1 of 59%, exceeding prior research baselines as well as commercial results, under identical repositories, model backends, and tool access.

Min Si, Pavan Balaji, Yongzhou Chen et al.

The increasing scale of large language models (LLMs) necessitates highly efficient collective communication frameworks, particularly as training workloads extend to hundreds of thousands of GPUs. Traditional communication methods face significant throughput and latency limitations at this scale, hindering both the development and deployment of state-of-the-art models. This paper presents the NCCLX collective communication framework, developed at Meta, engineered to optimize performance across the full LLM lifecycle, from the synchronous demands of large-scale training to the low-latency requirements of inference. The framework is designed to support complex workloads on clusters exceeding 100,000 GPUs, ensuring reliable, high-throughput, and low-latency data exchange. Empirical evaluation on the Llama4 model demonstrates substantial improvements in communication efficiency. This research contributes a robust solution for enabling the next generation of LLMs to operate at unprecedented scales.

Rana Shahout, Colin Cai, Yilun Du et al.

Mixture-of-Experts (MoE) models can scale parameter capacity by routing each token to a subset of experts through a learned gate function. While conditional routing reduces training costs, it shifts the burden on inference memory: expert parameters and activations consume memory, limiting the number of experts per device. As tokens are routed, some experts become overloaded while others are underutilized. Because experts are mapped to GPUs, this imbalance translates directly into degraded system performance in terms of latency, throughput, and cost. We present LASER, a plug-and-play, inference-time routing algorithm that balances load while preserving accuracy. LASER adapts to the shape of the gate's score distribution. When scores provide a clear preference, it routes to the strongest experts; when scores are more uniform, it broadens the set of viable experts and routes to the least-loaded among them. Because LASER relies only on gate scores from a trained model, it integrates directly into existing MoE inference pipelines without retraining or finetuning. We evaluate LASER on Mixtral-8x7B and DeepSeek-MoE-16b-chat across four datasets (ARC-Easy, ARC-Challenge, MMLU, and GSM8K). LASER improves load balancing, translating into lower latency and higher throughput, while keeping the accuracy changes negligible.

Weifan Jiang, Rana Shahout, Yilun Du et al.

Large Language Models (LLMs) increasingly rely on inference-time reasoning algorithms such as chain-of-thought and multi-branch reasoning to improve accuracy on complex tasks. These methods, however, substantially increase token usage and per-request latency. Prior work has largely focused on reducing token usage, often at the expense of accuracy, while overlooking other latency factors. We present DUCHESS, an LLM serving system that reduces cost and latency without sacrificing accuracy through intra-request branch orchestration guided by predictions. DUCHESS employs a lightweight linear probing model over LLM layer activations to estimate branch correctness, and its orchestration policy decides whether to terminate, duplicate, or continue a branch. When handling multiple requests, DUCHESS further reduces latency by prioritizing easier reasoning tasks when complexity can be estimated from the prompt. Experiments on three reasoning benchmarks show that DUCHESS consistently improves the token-accuracy Pareto frontier, reducing token usage by 42-63% at matched accuracy compared to self-consistency. In serving with vLLM, DUCHESS reduces mean, median, and tail latencies by 57-81%, 58-85%, and 52-84% with First-Come-First-Served scheduling, and achieves additional gains under difficulty-aware scheduling at higher request rates.

Samuel Lin, Jiawei Zhou, Minlan Yu

Traditional approaches to network management have been accessible only to a handful of highly-trained network operators with significant expert knowledge. This creates barriers for lay users to easily manage their networks without resorting to experts. With recent development of powerful large language models (LLMs) for language comprehension, we design a system to make network management accessible to a broader audience of non-experts by allowing users to converse with networks in natural language. To effectively leverage advancements in LLMs, we propose an agentic framework that uses an intermediate representation to streamline configuration across diverse vendor equipment, retrieves the network state from memory in real-time, and provides an interface for external feedback. We also conduct pilot studies to collect real user data of natural language utterances for network control, and present a visualization interface to facilitate dialogue-driven user interaction and enable large-scale data collection for future development. Preliminary experiments validate the effectiveness of our proposed system components with LLM integration on both synthetic and real user utterances. Through our data collection and visualization efforts, we pave the way for more effective use of LLMs and democratize network control for everyday users.

Xingyu Xiang, Raj Joshi, Yuhan Liu et al.

Distributed prefix caching accelerates long-context LLM serving by reusing KV cache entries for common context prefixes. However, KV cache fetches can become a bottleneck when network bandwidth is limited. Compression mitigates the bandwidth issue, but can degrade overall performance when decompression interferes with model computation. We present ShadowServe, the first SmartNIC-accelerated, interference-free prefix caching system for LLM serving. ShadowServe separates a control plane on the host and a data plane fully offloaded to the SmartNIC, which eliminates interference to both host GPU and CPU. To overcome the SmartNIC's limited compute and memory resources, we design a chunked pipeline that parallelizes data plane operations across the SmartNIC's compute resources, and a minimal-copy memory management scheme that reduces memory pressure on the SmartNIC. Compared to state-of-the-art solutions, ShadowServe achieves up to 2.2x lower loaded time-per-output-token (TPOT), and reduces time-to-first-token (TTFT) by up to 1.38x in low-bandwidth scenarios (<= 20 Gbps), translating to up to 1.35x higher throughput.

Jinsun Yoo, ChonLam Lao, Lianjie Cao et al.

This paper lays the foundation for Genie, a testing framework that captures the impact of real hardware network behavior on ML workload performance, without requiring expensive GPUs. Genie uses CPU-initiated traffic over a hardware testbed to emulate GPU to GPU communication, and adapts the ASTRA-sim simulator to model interaction between the network and the ML workload.

Zeyu Zhang, Haiying Shen, Shay Vargaftik et al.

Disaggregated Large Language Model (LLM) inference has gained popularity as it separates the computation-intensive prefill stage from the memory-intensive decode stage, avoiding the prefill-decode interference and improving resource utilization. However, transmitting Key-Value (KV) data between the two stages can be a bottleneck, especially for long prompts. Additionally, the computation time overhead for prefill and decode is key for optimizing Job Completion Time (JCT), and KV data size can become prohibitive for long prompts and sequences. Existing KV quantization methods can alleviate the transmission bottleneck and reduce memory requirements, but they introduce significant dequantization overhead, exacerbating the computation time. We propose Homomorphic Acceleration via Compression of the KV cache (HACK) for disaggregated LLM inference. HACK eliminates the heavy KV dequantization step, and directly performs computations on quantized KV data to approximate and reduce the cost of the expensive matrix-multiplication step. Extensive trace-driven experiments show that HACK reduces JCT by up to 70.9% compared to disaggregated LLM inference baseline and by up to 52.3% compared to state-of-the-art KV quantization methods.

Shaoke Xi, Jiaqi Gao, Mengqi Liu et al.

With the growing performance requirements on networked applications, there is a new trend of offloading stateful network applications to SmartNICs to improve performance and reduce the total cost of ownership. However, offloading stateful network applications is non-trivial due to state operation complexity, state resource consumption, and the complicated relationship between traffic and state. Naively partitioning the program by state or traffic can result in a suboptimal partition plan with higher CPU usage or even packet drops. In this paper, we propose Cora, a compiler and runtime that offloads stateful network applications to SmartNIC-accelerated hosts. Cora compiler introduces an accurate performance model for each SmartNIC and employs an efficient compiling algorithm to search the offloading plan. Cora runtime can monitor traffic dynamics and adapt to minimize CPU usage. Cora is built atop Netronome Agilio and BlueField 2 SmartNICs. Our evaluation shows that for the same throughput target, Cora can propose partition plans saving up to 94.0% CPU cores, 1.9 times more than baseline solutions. Under the same resource constraint, Cora can accelerate network functions by 44.9%-82.3%. Cora runtime can adapt to traffic changes and keep CPU usage low.

Jiawei Zhou, Zhiying Xu, Alexander M. Rush et al.

Botnets are now a major source for many network attacks, such as DDoS attacks and spam. However, most traditional detection methods heavily rely on heuristically designed multi-stage detection criteria. In this paper, we consider the neural network design challenges of using modern deep learning techniques to learn policies for botnet detection automatically. To generate training data, we synthesize botnet connections with different underlying communication patterns overlaid on large-scale real networks as datasets. To capture the important hierarchical structure of centralized botnets and the fast-mixing structure for decentralized botnets, we tailor graph neural networks (GNN) to detect the properties of these structures. Experimental results show that GNNs are better able to capture botnet structure than previous non-learning methods when trained with appropriate data, and that deeper GNNs are crucial for learning difficult botnet topologies. We believe our data and studies can be useful for both the network security and graph learning communities.