IEEE Transactions on Parallel and Distributed Systems最新文献

{"title":"VAHRM: Variation-Aware Resource Management in Heterogeneous Supercomputing Systems","authors":"Kohei Yoshida;Ryuichi Sakamoto;Kento Sato;Abhinav Bhatele;Hayato Yamaki;Hiroki Honda;Shinobu Miwa","doi":"10.1109/TPDS.2025.3577252","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3577252","url":null,"abstract":"In this article, we propose a novel resource management technique for heterogeneous supercomputing systems affected by manufacturing variability. Our proposed technique called VAHRM (Variation-Aware Heterogeneous Resource Management) takes a holistic approach to job scheduling on highly heterogeneous computing resources. VAHRM preferentially allocates energy-efficient computing resources to an energy-consuming job in a job queue, considering the impact on both the job turnaround time and the power consumption of individual resources. Furthermore, we have developed a novel approach to modeling the power consumption of computing resources that have manufacturing variability. Our approach called TSMVA (Two-Stage Modeling with Variation Awareness) enables us to generate the first variation-aware GPU power models, which can correctly estimate the power consumption of each GPU for a given job. Our experimental results show that, compared to conventional first-come-first-serve (FCFS) and state-of-the-art variation-aware scheduling algorithms, VAHRM can achieve respective improvements in system energy efficiency of up to 5.8% and 5.4% (4.5% and 4.2% on average) while reducing the average turnaround time of 21.2% and 11.9%, respectively, for various workloads obtained from a production system.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1713-1727"},"PeriodicalIF":5.6,"publicationDate":"2025-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11031465","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144524424","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Learned Performance Model With Transfer Learning Across GPUs on Tensorized Instructions","authors":"Yang Bai;Mingjun Li;Wendong Xu;Bei Yu","doi":"10.1109/TPDS.2025.3578630","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3578630","url":null,"abstract":"The training and inference efficiency of ever-larger deep neural networks highly rely on the performance of tensor operators on specific hardware accelerators. Therefore, a performance tuning framework with tensorized instruction compilation for automatic tensor generation is necessary for efficient deployment. These novel tensorized instruction, along with the emerging machine learning models, bring tremendous engineering challenges in compilation-based methods. They suffer from a large design space exploration with rough measurement accuracy and poor transferability among specialized instructions with certain hardware constraints. This paper presents a novel performance model for automatic code optimization with tensorized instruction. Central to the performance model is the assignment feature that not only clearly specifies the behaviour of instruction with computation and data movement abstraction, but also formally defines the matching problem from algorithm to tensorized instructions. Meanwhile, a simple yet efficient design with attention-inspired modules to accurately predict the performance of optimized tensor program by capturing global and long-range dependencies within a complete scheduling space. Compared with state-of-the-arts, our performance model can predict the optimal implementation of code configurations with tensorized instruction to reduce inference latency and search time by up to 1.21× and 3.41× on modern DNN benchmarks. Furthermore, with pre-trained parameters, our performance can quickly adapt to different workloads and platforms on tensorized instruction via transfer learning.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 9","pages":"1904-1919"},"PeriodicalIF":5.6,"publicationDate":"2025-06-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144646685","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Boosting Resource-Constrained Federated Learning Systems With Guessed Updates","authors":"Mohamed Yassine Boukhari;Akash Dhasade;Anne-Marie Kermarrec;Rafael Pires;Othmane Safsafi;Rishi Sharma","doi":"10.1109/TPDS.2025.3578522","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3578522","url":null,"abstract":"Federated learning (FL) enables a set of client devices to collaboratively train a model without sharing raw data. This process, though, operates under the constrained computation and communication resources of edge devices. These constraints combined with systems heterogeneity force some participating clients to perform fewer local updates than expected by the server, thus slowing down convergence. Exhaustive tuning of hyperparameters in FL, furthermore, can be resource-intensive, without which the convergence is adversely affected. In this work, we propose <sc>GeL</small>, the guess and learn algorithm. <sc>GeL</small> enables constrained edge devices to perform additional learning through guessed updates on top of gradient-based steps. These guesses are <italic>gradientless</i>, i.e., participating clients leverage them <italic>for free</i>. Our generic guessing algorithm (i) can be flexibly combined with several state-of-the-art algorithms including <sc>FedProx</small>, <sc>FedNova</small>, <sc>FedYogi</small> or <sc>ScaleFL</small>; and (ii) achieves significantly improved performance when the learning rates are not best tuned. We conduct extensive experiments and show that <sc>GeL</small> can boost empirical convergence by up to 40% in resource-constrained networks while relieving the need for exhaustive learning rate tuning.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1666-1679"},"PeriodicalIF":5.6,"publicationDate":"2025-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144502884","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"ISACPP: Interference-Aware Scheduling Approach for Deep Learning Training Workloads Based on Co-Location Performance Prediction","authors":"Zijie Liu;Yi Cheng;Can Chen;Jun Hu;Rongguo Fu;Dengyin Zhang","doi":"10.1109/TPDS.2025.3577796","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3577796","url":null,"abstract":"Traditional exclusive cloud resource allocation for deep learning training (DLT) workloads is unsuitable for advanced GPU infrastructure, leading to resource under-utilization. Fortunately, DLT workload co-location provides a promising way to improve resource utilization. However, existing workload co-location methods fail to accurately quantify interference among DLT workloads, resulting in performance degradation. To address this problem, this article proposes an interference-aware scheduling approach for DLT workloads based on co-location performance prediction, dubbed ‘ISACPP’. ISACPP first builds an edge-fusion gated graph attention network (E-GGAT) that incorporates DL model structures, underlying GPU types, and hyper-parameter settings to predict co-location performance. Since the co-location state changes as each workload is completed, ISACPP proposes a multi-stage co-location interference quantification model derived from the predicted co-location performance to identify the GPU device with the minimum overall interference. Experimental results demonstrate that ISACPP can accurately estimate the co-location performance of DLT workloads with a maximum prediction error of 8.72%, 1.9%, and 4.4% for execution time, GPU memory consumption, and GPU utilization, respectively. Meanwhile, ISACPP can significantly shorten workload makespan by up to 34.9% compared to state-of-the-art interference-aware scheduling methods.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1591-1607"},"PeriodicalIF":5.6,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144323113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Towards Efficiency and Decentralization: A Blockchain Assisted Distributed Fuzzy-Rough Feature Selection","authors":"Lin Qiu;Xingwei Wang;Bo Yi;Kaimin Zhang;Fei Gao;Min Huang;Yanpeng Qu","doi":"10.1109/TPDS.2025.3578032","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3578032","url":null,"abstract":"Fuzzy-rough sets-based feature selection (FRFS), as an effective data pre-processing technique, has drawn significant attention with the growing prevalence of large-scale datasets. However, centralized FRFS approaches suffer from the following shortcomings: 1) low computational efficiency, 2) bottlenecks in memory and computational resources, and 3) strict limitation of collaborative implementation using non-shared datasets owned by different data providers. These limitations highlight the growing necessity of integrating FRFS into a distributed FS framework. Nevertheless, most existing distributed FS schemes are reliant on a designated central server to collect and merge the local results from all slave nodes, which may result in several challenges including single point of failure risk, lack of trust and reliability, and lack of transparency and traceability. To relieve the above issues, this paper proposes a blockchain assisted distributed FS framework, successfully implementing a distributed solution for FRFS (BDFRFS). First, this framework introduces blockchain to merge, reach consensus and publish the global results generated during each iteration of FRFS, including the currently selected feature subset with its corresponding similarity matrix and dependency degree. This not only eliminates the reliance of central server and alleviates the burden on the central server, but also enhances the credibility and traceability of the results. Additionally, the implementation of FRFS is designed within this framework, utilizing three strategies to improve the efficiency of centralized FRFS: 1) eliminating the irrelevant and redundant features prior to the executing FRFS; 2) removing redundant and unnecessary computations involved in generating the similarity matrices; and 3) enabling parallel computation of dependency degrees. Finally, the experimental results conducted on eight large-scale datasets demonstrate that the proposed framework can significantly reduce the runtime cost and improve the classification accuracy compared to centralized FRFS and several distributed FS approaches.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1762-1778"},"PeriodicalIF":5.6,"publicationDate":"2025-06-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144536658","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Everything Distributed and Asynchronous: A Practical System for Key Management Service","authors":"Zhaoyang Xie;Haibin Zhang;Sisi Duan;Chao Liu;Shengli Liu;Xuanji Meng;Yong Yu;Fangguo Zhang;Boxin Zhao;Liehuang Zhu;Tianqing Zhu","doi":"10.1109/TPDS.2025.3577038","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3577038","url":null,"abstract":"A key management service (KMS) is vital to modern mission-critical systems. At the core of KMS are the key generation process and the key refresh process. In this paper, we design and implement a purely asynchronous system for completely distributed KMS supporting traditional applications such as threshold cryptosystems and multiparty computation (MPC) as well as emerging blockchains and Web3 applications. In this system, we have built a number of new asynchronous distributed key generation (ADKG) protocols and their corresponding asynchronous distributed key refresh (ADKR) protocols. We have demonstrated that our ADKG and ADKR protocols in the standard model outperform existing ones of the same kind, while our protocols in the random oracle model (ROM) are more efficient than other protocols with small and medium-sized networks.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 9","pages":"1841-1856"},"PeriodicalIF":5.6,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144657462","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SSS-DIMM: Removing Redundant Data Movement in Trusted DIMM-Based Near-Memory-Processing Kernel Offloading via Secure Space Sharing","authors":"Weiyi Sun;Jianfeng Zhu;Mingyu Gao;Zhaoshi Li;Shaojun Wei;Leibo Liu","doi":"10.1109/TPDS.2025.3576438","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3576438","url":null,"abstract":"DIMM-based Near-Memory-Processing (NMP) kernel offloading enables a program to execute in computation-enabled DIMM buffer chips, bypassing the bandwidth-constrained CPU main memory bus for high performance. Yet, it also enables programs to access memory without restrictions and protection from CPU, resulting in potential security hazards. To protect general NMP kernel offloading even with malicious privileged software, a heterogeneous TEE is required. However, for architectural design simplification, the conventional heterogeneous TEE design isolates host CPU process from NMP kernel’s memory and vice versa, such that CPU TEE and trusted NMP driver can protect CPU processes and NMP kernels in complete separation. Such isolation results in redundant input/output data movement between the two isolated memory spaces, with half of the movement performed by host CPU. Worsened by limited CPU memory bandwidth, we identify that such redundancy severely bottlenecks the performance of many potential NMP applications. To overcome this bottleneck, we propose to abandon isolation and share the NMP kernel memory with its host CPU process. Based on this idea, we design <underline>SSS-DIMM</u>, an efficient TEE for DIMM-based NMP kernel offloading that removes the redundant data movement via <underline>S</u>ecure <underline>S</u>pace <underline>S</u>haring. SSS-DIMM resolves the two security challenges faced by memory sharing: to provide consistent security guarantees on CPU processes and NMP kernels with CPU TEE and the NMP driver for both memory ownership (allocation) and views (mapping), and to ensure that cryptography metadata be securely shared and synchronized between CPU and NMP unit. Our evaluation shows that SSS-DIMM maintains both security and high performance.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1810-1827"},"PeriodicalIF":5.6,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144536659","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"RHINO: An Efficient Serverless Container System for Small-Scale HPC Applications","authors":"He Zhu;Mingyu Li;Haihang You","doi":"10.1109/TPDS.2025.3576584","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3576584","url":null,"abstract":"Serverless computing, characterized by its pay-as-you-go and auto-scaling features, offers a promising alternative for High Performance Computing (HPC) applications, as traditional HPC clusters often face long waiting times and resources over/under-provisioning. However, current serverless platforms struggle to support HPC applications due to restricted inter-function communication and high coupling runtime. To address these issues, we introduce RHINO, which offers end-to-end support for the development and deployment of serverless HPC. Using the Two-Step Adaptive Build strategy, the HPC code is packaged into lightweight, scalable functions. The Rhino Function Execution Model decouples HPC applications from the underlying infrastructures. The Auto-scaling Engine dynamically scales cloud resources and schedules tasks based on performance and cost requirements. We deploy RHINO on AWS Fargate and evaluate it on both benchmarks and real-world workloads. Experimental results show that, when compared to the traditional VM clusters, RHINO can achieve a performance improvement of 10% –30% for small-scale applications and more than 40% cost reduction.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1560-1573"},"PeriodicalIF":5.6,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144323114","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Featherlight Stateful WebAssembly for Serverless Inference Workflows","authors":"Xingguo Pang;Liu Liu;Yanze Zhang;Zhuofu Chen;Zhijun Ding;Dazhao Cheng;Xiaobo Zhou","doi":"10.1109/TPDS.2025.3575753","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3575753","url":null,"abstract":"In serverless inference, complex prediction tasks are executed as workflows, relying on efficient state transfer across multiple functions. Serverless platforms typically deploy each function in a separate stateless container, depending on external processes for state management, which often results in suboptimal system utilization and increased latency. We introduce WasmFlow, a novel framework designed for serverless inference that ensures low latency and high throughput. This is achieved through process-level virtualization using WebAssembly. WasmFlow operates functions on a per-thread basis within compact WebAssembly modules, significantly reducing startup times and memory usage. The framework has two key features. (1) Efficient Memory Sharing: WasmFlow facilitates direct and rapid state transfer between functions using threads within the WebAssembly runtime. This is enabled through lightweight, lock-free, zero-copy intra-process communication, complemented by effective inter-process RPC. (2) System Optimizations: We further optimize WasmFlow with an advanced synchronization technique between functions, an affinity-aware workflow scheduler, and adaptive request batching. Implemented and integrated within the Kubernetes ecosystem, WasmFlow’s performance was evaluated using synthetic workloads and real-world Azure traces, including typical serverless workflows and ML models. Our results demonstrate that WasmFlow dramatically outperforms existing serverless frameworks. It reduces P90 end-to-end latency by 74x and 78x, increases function density by 1.7x and 223x compared to Faasm and SPRIGHT, and improves system throughput by 12.3x and 8.8x over Knative and WasmEdge, respectively.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1651-1665"},"PeriodicalIF":5.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144481927","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"OpenSN: An Open Source Library for Emulating LEO Satellite Networks","authors":"Wenhao Lu;Zhiyuan Wang;Hefan Zhang;Shan Zhang;Hongbin Luo","doi":"10.1109/TPDS.2025.3575920","DOIUrl":"https://doi.org/10.1109/TPDS.2025.3575920","url":null,"abstract":"Low-earth-orbit (LEO) satellite constellations (e.g., Starlink) are becoming a necessary component of future Internet. There have been increasing studies on LEO satellite networking. It is a crucial problem how to evaluate these studies in a systematic and reproducible manner. In this paper, we present OpenSN, i.e., an open source library for emulating large-scale satellite network (SN). Different from Mininet-based SN emulators (e.g., LeoEM), OpenSN adopts container-based virtualization, thus allows for running distributed routing software on each node, and can achieve horizontal scalability via flexible multi-machine extension. Compared to other container-based SN emulators (e.g., StarryNet), OpenSN streamlines the interaction with Docker command line interface and significantly reduces unnecessary operations of creating virtual links. These modifications improve emulation efficiency and vertical scalability on a single machine. Furthermore, OpenSN separates user-defined configuration from container network management via a Key-Value Database that records the necessary information for SN emulation. Such a separation architecture enhances the function extensibility. To sum up, OpenSN exhibits advantages in efficiency, scalability, and extensibility, thus is a valuable open source library that empowers research on LEO satellite networking. Experiment results show that OpenSN constructs mega-constellations 5X-10X faster than StarryNet, and updates link state 2X-4X faster than LeoEM. We also verify the scalability of OpenSN by successfully emulating the five-shell Starlink constellation with a total of 4408 satellites.","PeriodicalId":13257,"journal":{"name":"IEEE Transactions on Parallel and Distributed Systems","volume":"36 8","pages":"1574-1590"},"PeriodicalIF":5.6,"publicationDate":"2025-06-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144323115","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}