William Won

William Won, Taekyung Heo, Saeed Rashidi et al.

As deep learning models and input data are scaling at an unprecedented rate, it is inevitable to move towards distributed training platforms to fit the model and increase training throughput. State-of-the-art approaches and techniques, such as wafer-scale nodes, multi-dimensional network topologies, disaggregated memory systems, and parallelization strategies, have been actively adopted by emerging distributed training systems. This results in a complex SW/HW co-design stack of distributed training, necessitating a modeling/simulation infrastructure for design-space exploration. In this paper, we extend the open-source ASTRA-sim infrastructure and endow it with the capabilities to model state-of-the-art and emerging distributed training models and platforms. More specifically, (i) we enable ASTRA-sim to support arbitrary model parallelization strategies via a graph-based training-loop implementation, (ii) we implement a parameterizable multi-dimensional heterogeneous topology generation infrastructure with analytical performance estimates enabling simulating target systems at scale, and (iii) we enhance the memory system modeling to support accurate modeling of in-network collective communication and disaggregated memory systems. With such capabilities, we run comprehensive case studies targeting emerging distributed models and platforms. This infrastructure lets system designers swiftly traverse the complex co-design stack and give meaningful insights when designing and deploying distributed training platforms at scale.

Hanjiang Wu, Abhimanyu Rajeshkumar Bambhaniya, Sarbartha Banerjee et al.

Modern large language model (LLM) inference has progressively disaggregated to keep pace with growing model sizes and tight TTFT and TPOT service-level objectives: from chunked-prefill aggregation, to prefill-decode (P/D) disaggregation, and most recently to operator-level Attention-FFN Disaggregation (AFD). This trend is especially important for mixture-of-experts (MoE) models, where memory-bound attention, compute-intensive expert FFNs, and MoE dispatch/combine communication create distinct resource demands. AFD further exposes this heterogeneity by placing attention and MoE-FFN execution on separate GPU groups. Each level of disaggregation deepens the scheduling design space across workload characteristics, resource allocation, and interconnect topology, raising the central question: when does each level actually pay off? We systematically characterize this trade-off for MoE inference across realistic workloads spanning input/output sequence lengths, prefix-KV reuse, and per-user latency constraints. Using chunked-prefill and P/D disaggregation as baselines, we study the benefits and limits of AFD at scale through a framework that fuses on-device kernel measurements with high-fidelity network simulation. Under strict TTFT/TPOT SLOs, AFD sustains around 4k tokens/s of system throughput on DeepSeek-V3.2 across chat, coding, and agentic-coding workloads, where non-AFD deployments are infeasible. We distill concrete takeaways for jointly optimizing throughput and interactivity, including how to partition attention and FFN across GPUs as a function of workload and model architecture, providing design principles for current rack- and cluster-scale deployments as well as future disaggregated AI infrastructure.

William Won, Midhilesh Elavazhagan, Sudarshan Srinivasan et al.

The surge of artificial intelligence, particularly large language models, has driven the rapid development of large-scale machine learning clusters. Executing distributed models on these clusters is often constrained by communication overhead, making efficient utilization of available network resources crucial. As a result, the routing algorithm employed for collective communications (i.e., collective algorithms) plays a pivotal role in determining overall performance. Unfortunately, existing collective communication libraries for distributed machine learning are limited by a fixed set of basic collective algorithms. This limitation hinders communication optimization, especially in modern clusters with heterogeneous and asymmetric topologies. Furthermore, manually designing collective algorithms for all possible combinations of network topologies and collective patterns requires heavy engineering and validation efforts. To address these challenges, this paper presents TACOS, an autonomous synthesizer capable of automatically generating topology-aware collective algorithms tailored to specific collective patterns and network topologies. TACOS is highly flexible, synthesizing an All-Reduce algorithm for a heterogeneous 128-NPU system in just 1.08 seconds, while achieving up to a 4.27x performance improvement over state-of-the-art synthesizers. Additionally, TACOS demonstrates better scalability with polynomial synthesis times, in contrast to NP-hard approaches which only scale to systems with tens of NPUs. TACOS can synthesize for 40K NPUs in just 2.52 hours.

Srinivas Sridharan, Andy Balogh, Bradford M. Beckmann et al.

The fast pace of artificial intelligence~(AI) innovation demands an agile methodology for observation, reproduction and optimization of distributed machine learning~(ML) workload behavior in production AI systems and enables efficient software-hardware~(SW-HW) co-design for future systems. We present Chakra, an open and portable ecosystem for performance benchmarking and co-design. The core component of Chakra is an open and interoperable graph-based representation of distributed AI/ML workloads, called Chakra execution trace~(ET). These ETs represent key operations, such as compute, memory, and communication, data and control dependencies, timing, and resource constraints. Additionally, Chakra includes a complementary set of tools and capabilities to enable the collection, analysis, generation, and adoption of Chakra ETs by a broad range of simulators, emulators, and replay tools. We present analysis of Chakra ETs collected on production AI clusters and demonstrate value via real-world case studies. Chakra has been adopted by MLCommons and has active contributions and engagement across the industry, including but not limited to NVIDIA, AMD, Meta, Keysight, HPE, and Scala, to name a few.

Saeed Rashidi, William Won, Sudarshan Srinivasan et al.

Distributed Deep Neural Network (DNN) training is a technique to reduce the training overhead by distributing the training tasks into multiple accelerators, according to a parallelization strategy. However, high-performance compute and interconnects are needed for maximum speed-up and linear scaling of the system. Wafer-scale systems are a promising technology that allows for tightly integrating high-end accelerators with high-speed wafer-scale interconnects, making it an attractive platform for distributed training. However, the wafer-scale interconnect should offer high performance and flexibility for various parallelization strategies to enable maximum optimizations for compute and memory usage. In this paper, we propose FRED, a wafer-scale interconnect that is tailored for the high-BW requirements of wafer-scale networks and can efficiently execute communication patterns of different parallelization strategies. Furthermore, FRED supports in-switch collective communication execution that reduces the network traffic by approximately 2X. Our results show that FRED can improve the average end-to-end training time of ResNet-152, Transformer-17B, GPT-3, and Transformer-1T by 1.76X, 1.87X, 1.34X, and 1.4X, respectively when compared to a baseline waferscale 2D-Mesh fabric.

Saeed Rashidi, William Won, Sudarshan Srinivasan et al.

Distributed training is a solution to reduce DNN training time by splitting the task across multiple NPUs (e.g., GPU/TPU). However, distributed training adds communication overhead between the NPUs in order to synchronize the gradients and/or activation, depending on the parallelization strategy. In next-generation platforms for training at scale, NPUs will be connected through multi-dimensional networks with diverse, heterogeneous bandwidths. This work identifies a looming challenge of keeping all network dimensions busy and maximizing the network BW within the hybrid environment if we leverage scheduling techniques for collective communication on systems today. We propose Themis, a novel collective scheduling scheme that dynamically schedules collectives (divided into chunks) to balance the communication loads across all dimensions, further improving the network BW utilization. Our results show that on average, Themis can improve the network BW utilization of the single All-Reduce by 1.72X (2.70X max), and improve the end-to-end training iteration performance of real workloads such as ResNet-152, GNMT, DLRM, and Transformer-1T by 1.49X (2.25X max), 1.30X (1.78X max), 1.30X (1.77X max), and 1.25X (1.53X max), respectively.

William Won, Saeed Rashidi, Sudarshan Srinivasan et al.

As model sizes in machine learning continue to scale, distributed training is necessary to accommodate model weights within each device and to reduce training time. However, this comes with the expense of increased communication overhead due to the exchange of gradients and activations, which become the critical bottleneck of the end-to-end training process. In this work, we motivate the design of multi-dimensional networks within machine learning systems as a cost-efficient mechanism to enhance overall network bandwidth. We also identify that optimal bandwidth allocation is pivotal for multi-dimensional networks to ensure efficient resource utilization. We introduce LIBRA, a framework specifically focused on optimizing multi-dimensional fabric architectures. Through case studies, we demonstrate the value of LIBRA, both in architecting optimized fabrics under diverse constraints and in enabling co-optimization opportunities.