分布式、并行与集群计算

We introduce GCAS, a natural generalization of the well-known compare-and-swap (CAS) object. Intuitively, GCAS just replaces the fixed equality test of CAS with a parametrized comparator chosen from $\{<, =, >\}$. To showcase the utility of…

We prove that given any $\alpha$-approximation LOCAL algorithm for Minimum Dominating Set (MDS) on planar graphs, we can construct an $f(g)$-round $(3\alpha+1)$-approximation LOCAL algorithm for MDS on graphs embeddable in a given Euler…

Delivering AI-generated content (AIGC) services fundamentally relies on the reasoning capabilities of generative AI (GenAI) models. Chain-of-Thought (CoT) enhances such reasoning by guiding models through intermediate steps, while…

Kubernetes clusters generate rich operational events during pod lifecycle transitions, yet the platform's native event retention model discards the most diagnostically valuable context. The LastTerminationState field, which records a…

Discovering atom-level phenomena requires molecular dynamics (MD) simulations with ab initio accuracy. Machine learning interatomic potentials (MLIPs) enable stable, high-accuracy MD simulations, and their models exhibit scaling-law trends…

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…

Multimodal Large Language Models (MLLMs) have achieved remarkable advances by integrating text, image, and audio understanding within a unified architecture. However, existing distributed training frameworks remain fundamentally data-blind:…

Global cryptocurrencies are unbacked and have high transaction cost incurred by global consensus. In contrast, grassroots cryptocurrencies are backed by the goods and services of their issuers -- any person, natural or legal -- and have no…

Large Language Model (LLM) serving faces a fundamental tension between stringent latency Service Level Objectives (SLOs) and limited GPU memory capacity. When high request rates exhaust the KV cache budget, existing LLM inference systems…

Cloud applications often insert a caching lay\-er in front of a database in order to reduce I/O latency and improve throughput. One complication occurs when a client fetches some data from one cache node, then migrates to another (e.g., due…

As large-scale language models continue to scale up in both size and context length, the memory and communication cost of key-value (KV) cache storage has become a major bottleneck in multi-GPU and multi-node inference. While MoE-based…

Pipeline parallelism is a key technique for scaling large-model training, but modern workloads exhibit runtime variability in computation and communication. Existing pipeline systems typically consume static, profiled, or adaptively…

With the wide adoption of Multimodal Models (MMs) in real-world scenarios, it is significant to efficiently train emerging MMs that exhibit increasingly complex module architectures. For MM deployment, existing works allocate a GPU to only…

Population protocols are a model of distributed computing where $n$ agents, each a simple finite-state machine, interact in pairs to solve a common task against a (adversarial) interaction scheduler. This model was intensively studied in…

Compound AI applications, which compose calls to ML models using a general-purpose programming language like Python, are widely used for a variety of user-facing tasks, from software engineering to enterprise automation, making their…

In-Network Collective (INC) acceleration holds immense potential for optimizing AI training and inference; however, its cross-layer nature has historically hindered investment and adoption within the open Ethernet ecosystem. To bridge this…

Sparse matrix-vector multiplication (SpMV) is crucial in computational science, engineering, and machine learning. Despite substantial efforts to improve SpMV performance on GPUs through various techniques, issues related to data locality,…

This paper presents a novel method to compare two numbers in Residue Number System (RNS) using an additional modulus, which is often already available because it is required in modular computations and digital signal processing scaling.Our…

Continuous cloud service performance benchmarking is essential for detecting performance bugs early before deploying them to production. However, detecting performance regressions using application benchmarks, which usually treat the system…

Uninterrupted system availability is a critical requirement for enterprise operations, yet traditional high-availability clusters suffer from limitations such as single points of failure and inefficient resource allocation. This paper…