Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00069
Younghyun Cho, J. Demmel, Jacob King, X. Li, Yang Liu, Hengrui Luo
This paper presents GPTuneCrowd, a crowd-based autotuning framework for tuning high-performance computing applications. GPTuneCrowd collects performance data from various users using a user-friendly tuner interface. GPTuneCrowd then presents novel autotuning techniques, based on transfer learning and parameter sensitivity analysis, to maximize tuning quality using collected data from the crowd. This paper shows several real-world case studies of GPTuneCrowd. Our evaluation shows that GPTuneCrowd’s transfer learning improves the tuned performance of ScaLAPACK’s PDGEQRF by 1.57x and a plasma fusion code NIMROD by 2.97x, over a non-transfer learning autotuner. We use GPTuneCrowd’s sensitivity analysis to reduce the search space of SuperLU_DIST and Hypre. Tuning on the reduced search space achieves 1.17x and 1.35x better tuned performance of SuperLU_DIST and Hypre, respectively, compared to the original search space.
{"title":"Harnessing the Crowd for Autotuning High-Performance Computing Applications","authors":"Younghyun Cho, J. Demmel, Jacob King, X. Li, Yang Liu, Hengrui Luo","doi":"10.1109/IPDPS54959.2023.00069","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00069","url":null,"abstract":"This paper presents GPTuneCrowd, a crowd-based autotuning framework for tuning high-performance computing applications. GPTuneCrowd collects performance data from various users using a user-friendly tuner interface. GPTuneCrowd then presents novel autotuning techniques, based on transfer learning and parameter sensitivity analysis, to maximize tuning quality using collected data from the crowd. This paper shows several real-world case studies of GPTuneCrowd. Our evaluation shows that GPTuneCrowd’s transfer learning improves the tuned performance of ScaLAPACK’s PDGEQRF by 1.57x and a plasma fusion code NIMROD by 2.97x, over a non-transfer learning autotuner. We use GPTuneCrowd’s sensitivity analysis to reduce the search space of SuperLU_DIST and Hypre. Tuning on the reduced search space achieves 1.17x and 1.35x better tuned performance of SuperLU_DIST and Hypre, respectively, compared to the original search space.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127218390","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00078
C. Hellwig, Fabian Czappa, Martin Michel, R. Bertrand, F. Wolf
In recent years, the number of artificial objects in Earth orbit has increased rapidly due to lower launch costs and new applications for satellites. More and more governments and private companies are discovering space for their own purposes. Private companies are using space as a new business field, launching thousands of satellites into orbit to offer services like worldwide Internet access. Consequently, the probability of collisions and, thus, the degradation of the orbital environment is rapidly increasing. To avoid devastating collisions at an early stage, efficient algorithms are required to identify satellites approaching each other. Traditional deterministic filter-based conjunction detection algorithms compare each satellite to every other satellite and pass them through a chain of orbital filters. Unfortunately, this leads to a runtime complexity of O(n2). In this paper, we propose two alternative approaches that rely on spatial data structures and thus allow us to exploit modern hardware’s parallelism efficiently. Firstly, we introduce a purely grid-based variant that relies on non-blocking atomic hash maps to identify conjunctions. Secondly, we present a hybrid method that combines this approach with traditional filter chains. Both implementations make it possible to identify conjunctions in a large population with millions of satellites with high precision in a comparatively short time. While the grid-based variant is characterized by lower memory consumption, the hybrid variant is faster if enough memory is available.
{"title":"Satellite Collision Detection using Spatial Data Structures","authors":"C. Hellwig, Fabian Czappa, Martin Michel, R. Bertrand, F. Wolf","doi":"10.1109/IPDPS54959.2023.00078","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00078","url":null,"abstract":"In recent years, the number of artificial objects in Earth orbit has increased rapidly due to lower launch costs and new applications for satellites. More and more governments and private companies are discovering space for their own purposes. Private companies are using space as a new business field, launching thousands of satellites into orbit to offer services like worldwide Internet access. Consequently, the probability of collisions and, thus, the degradation of the orbital environment is rapidly increasing. To avoid devastating collisions at an early stage, efficient algorithms are required to identify satellites approaching each other. Traditional deterministic filter-based conjunction detection algorithms compare each satellite to every other satellite and pass them through a chain of orbital filters. Unfortunately, this leads to a runtime complexity of O(n2). In this paper, we propose two alternative approaches that rely on spatial data structures and thus allow us to exploit modern hardware’s parallelism efficiently. Firstly, we introduce a purely grid-based variant that relies on non-blocking atomic hash maps to identify conjunctions. Secondly, we present a hybrid method that combines this approach with traditional filter chains. Both implementations make it possible to identify conjunctions in a large population with millions of satellites with high precision in a comparatively short time. While the grid-based variant is characterized by lower memory consumption, the hybrid variant is faster if enough memory is available.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129819452","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00063
William Ladd, Christopher W Jensen, M. Vardhan, Jeff Ames, J. Hammond, E. Draeger, A. Randles
Cloud computing resources are becoming an increasingly attractive option for simulation workflows but require users to assess a wider variety of hardware options and associated costs than required by traditional in-house hardware or fixed allocations at leadership computing facilities. The pay-as-you-go model used by cloud providers gives users the opportunity to make more nuanced cost-benefit decisions at runtime by choosing hardware that best matches a given workload, but creates the risk of suboptimal allocation strategies or inadvertent cost overruns. In this work, we propose the use of an iteratively-refined performance model to optimize cloud simulation campaigns against overall cost, throughput, or maximum time to solution. Hemodynamic simulations represent an excellent use case for these assessments, as the relative costs and dominant terms in the performance model can vary widely with hardware, numerical parameters and physics models. Performance and scaling behavior of hemodynamic simulations on multiple cloud services as well as a traditional compute cluster are collected and evaluated, and an initial performance model is proposed along with a strategy for dynamically refining it with additional experimental data.
{"title":"Optimizing Cloud Computing Resource Usage for Hemodynamic Simulation","authors":"William Ladd, Christopher W Jensen, M. Vardhan, Jeff Ames, J. Hammond, E. Draeger, A. Randles","doi":"10.1109/IPDPS54959.2023.00063","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00063","url":null,"abstract":"Cloud computing resources are becoming an increasingly attractive option for simulation workflows but require users to assess a wider variety of hardware options and associated costs than required by traditional in-house hardware or fixed allocations at leadership computing facilities. The pay-as-you-go model used by cloud providers gives users the opportunity to make more nuanced cost-benefit decisions at runtime by choosing hardware that best matches a given workload, but creates the risk of suboptimal allocation strategies or inadvertent cost overruns. In this work, we propose the use of an iteratively-refined performance model to optimize cloud simulation campaigns against overall cost, throughput, or maximum time to solution. Hemodynamic simulations represent an excellent use case for these assessments, as the relative costs and dominant terms in the performance model can vary widely with hardware, numerical parameters and physics models. Performance and scaling behavior of hemodynamic simulations on multiple cloud services as well as a traditional compute cluster are collected and evaluated, and an initial performance model is proposed along with a strategy for dynamically refining it with additional experimental data.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130813785","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00066
J. H. Davis, Justin Shafner, Daniel Nichols, N. Grube, P. Martin, A. Bhatele
Accurate modeling of turbulent hypersonic flows has tremendous scientific and commercial value, and applies to atmospheric flight, supersonic combustion, materials discovery and climate prediction. In this paper, we describe our experiences in extending the capabilities of and modernizing CRoCCo, an MPI-based, CPU-only compressible computational fluid dynamics code. We extend CRoCCo to support block-structured adaptive mesh refinement using a highly-scalable AMR library, AMReX, and add support for a fully curvilinear solver. We also port the computational kernels in CRoCCo to GPUs to enable scaling on modern exascale systems. We present our techniques for overcoming performance challenges and evaluate the updated code, CRoCCo v2.0, on the Summit system, demonstrating a 6× to 44× speedup over the CPU-only version.
{"title":"Porting a Computational Fluid Dynamics Code with AMR to Large-scale GPU Platforms","authors":"J. H. Davis, Justin Shafner, Daniel Nichols, N. Grube, P. Martin, A. Bhatele","doi":"10.1109/IPDPS54959.2023.00066","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00066","url":null,"abstract":"Accurate modeling of turbulent hypersonic flows has tremendous scientific and commercial value, and applies to atmospheric flight, supersonic combustion, materials discovery and climate prediction. In this paper, we describe our experiences in extending the capabilities of and modernizing CRoCCo, an MPI-based, CPU-only compressible computational fluid dynamics code. We extend CRoCCo to support block-structured adaptive mesh refinement using a highly-scalable AMR library, AMReX, and add support for a fully curvilinear solver. We also port the computational kernels in CRoCCo to GPUs to enable scaling on modern exascale systems. We present our techniques for overcoming performance challenges and evaluate the updated code, CRoCCo v2.0, on the Summit system, demonstrating a 6× to 44× speedup over the CPU-only version.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130949397","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Federated Learning (FL) emerges as a distributed machine learning paradigm without end-user data transmission, effectively avoiding privacy leakage. Participating devices in FL are usually bandwidth-constrained, and the uplink is much slower than the downlink in wireless networks, which causes a severe uplink communication bottleneck. A prominent direction to alleviate this problem is federated dropout, which drops fractional weights of local models. However, existing federated dropout studies focus on random or ordered dropout and lack theoretical support, resulting in unguaranteed performance. In this paper, we propose Federated learning with Bayesian Inference-based Adaptive Dropout (FedBIAD), which regards weight rows of local models as probability distributions and adaptively drops partial weight rows based on importance indicators correlated with the trend of local training loss. By applying FedBIAD, each client adaptively selects a high-quality dropping pattern with accurate approximations and only transmits parameters of non-dropped weight rows to mitigate uplink costs while improving accuracy. Theoretical analysis demonstrates that the convergence rate of the average generalization error of FedBIAD is minimax optimal up to a squared logarithmic factor. Extensive experiments on image classification and next-word prediction show that compared with status quo approaches, FedBIAD provides 2× uplink reduction with an accuracy increase of up to 2.41% even on non-Independent and Identically Distributed (non-IID) data, which brings up to 72% decrease in training time.
联邦学习(FL)作为一种不需要终端用户数据传输的分布式机器学习范式而出现，有效地避免了隐私泄露。FL的参与设备通常受带宽限制，且上行速度比无线网络中的下行速度慢得多，造成了严重的上行通信瓶颈。缓解这一问题的一个重要方向是联邦dropout，它降低了局部模型的分数权重。然而，现有的联邦退学研究多集中于随机退学或有序退学，缺乏理论支持，导致性能得不到保证。本文提出了基于贝叶斯推理的自适应Dropout (FedBIAD)联邦学习方法，该方法将局部模型的权重行视为概率分布，并根据与局部训练损失趋势相关的重要指标自适应地丢弃部分权重行。通过应用FedBIAD，每个客户端自适应地选择具有精确近似的高质量丢弃模式，并且只传输未丢弃的权值行参数，从而在降低上行成本的同时提高准确性。理论分析表明，FedBIAD的平均泛化误差收敛速度在一个对数因子的平方范围内是极小极大最优的。大量的图像分类和下一词预测实验表明，与现有方法相比，FedBIAD在非独立同分布(non-Independent and Identically Distributed, non-IID)数据上的上行链路减少了2倍，准确率提高了2.41%，训练时间减少了72%。
{"title":"FedBIAD: Communication-Efficient and Accuracy-Guaranteed Federated Learning with Bayesian Inference-Based Adaptive Dropout","authors":"Jingjing Xue, Min Liu, Sheng Sun, Yuwei Wang, Hui Jiang, Xue Jiang","doi":"10.1109/IPDPS54959.2023.00056","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00056","url":null,"abstract":"Federated Learning (FL) emerges as a distributed machine learning paradigm without end-user data transmission, effectively avoiding privacy leakage. Participating devices in FL are usually bandwidth-constrained, and the uplink is much slower than the downlink in wireless networks, which causes a severe uplink communication bottleneck. A prominent direction to alleviate this problem is federated dropout, which drops fractional weights of local models. However, existing federated dropout studies focus on random or ordered dropout and lack theoretical support, resulting in unguaranteed performance. In this paper, we propose Federated learning with Bayesian Inference-based Adaptive Dropout (FedBIAD), which regards weight rows of local models as probability distributions and adaptively drops partial weight rows based on importance indicators correlated with the trend of local training loss. By applying FedBIAD, each client adaptively selects a high-quality dropping pattern with accurate approximations and only transmits parameters of non-dropped weight rows to mitigate uplink costs while improving accuracy. Theoretical analysis demonstrates that the convergence rate of the average generalization error of FedBIAD is minimax optimal up to a squared logarithmic factor. Extensive experiments on image classification and next-word prediction show that compared with status quo approaches, FedBIAD provides 2× uplink reduction with an accuracy increase of up to 2.41% even on non-Independent and Identically Distributed (non-IID) data, which brings up to 72% decrease in training time.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129357731","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00091
Lu Zhang, C. Li, Xinkai Wang, Weiqi Feng, Zheng Yu, Quan Chen, Jingwen Leng, Minyi Guo, Pu Yang, Shang Yue
Emerging cloud-native development models raise new challenges for managing server performance and power at microsecond scale. Compared with traditional cloud workloads, serverless functions exhibit unprecedented heterogeneity, variability, and dynamicity. Designing cloud-native power management schemes for serverless functions requires significant engineering effort. Current solutions remain sub-optimal since their orchestration process is often one-sided, lacking a systematic view. A key obstacle to truly efficient function deployment is the fundamental wide abstraction gap between the upper-layer request scheduling and the low-level hardware execution.In this work, we show that the optimal operating point (OOP) for energy efficiency cannot be attained without synthesizing the multi-dimensional attributes of functions. We present FIRST, a novel mechanism that enables servers to better orchestrate serverless functions. The key feature of FIRST is that it leverages a lightweight Internal Representation and meta-Scheduling (IRS) layer for collecting the maximum potential revenue from the servers. Specifically, FIRST follows a pipeline-style workflow. Its frontend components aim to analyze functions from different angles and expose their key features to the system. Meanwhile, its backend components are able to make informed function assignment decisions to avoid OOP divergence. We further demonstrate the way to create extensions based on FIRST to enable versatile cloud-native power management. In total, our design constitutes a flexible management layer that supports power-aware function deployment. We show that FIRST could allow 94% functions to be processed under the OOP, which brings up to 24% energy efficiency improvements.
{"title":"FIRST: Exploiting the Multi-Dimensional Attributes of Functions for Power-Aware Serverless Computing","authors":"Lu Zhang, C. Li, Xinkai Wang, Weiqi Feng, Zheng Yu, Quan Chen, Jingwen Leng, Minyi Guo, Pu Yang, Shang Yue","doi":"10.1109/IPDPS54959.2023.00091","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00091","url":null,"abstract":"Emerging cloud-native development models raise new challenges for managing server performance and power at microsecond scale. Compared with traditional cloud workloads, serverless functions exhibit unprecedented heterogeneity, variability, and dynamicity. Designing cloud-native power management schemes for serverless functions requires significant engineering effort. Current solutions remain sub-optimal since their orchestration process is often one-sided, lacking a systematic view. A key obstacle to truly efficient function deployment is the fundamental wide abstraction gap between the upper-layer request scheduling and the low-level hardware execution.In this work, we show that the optimal operating point (OOP) for energy efficiency cannot be attained without synthesizing the multi-dimensional attributes of functions. We present FIRST, a novel mechanism that enables servers to better orchestrate serverless functions. The key feature of FIRST is that it leverages a lightweight Internal Representation and meta-Scheduling (IRS) layer for collecting the maximum potential revenue from the servers. Specifically, FIRST follows a pipeline-style workflow. Its frontend components aim to analyze functions from different angles and expose their key features to the system. Meanwhile, its backend components are able to make informed function assignment decisions to avoid OOP divergence. We further demonstrate the way to create extensions based on FIRST to enable versatile cloud-native power management. In total, our design constitutes a flexible management layer that supports power-aware function deployment. We show that FIRST could allow 94% functions to be processed under the OOP, which brings up to 24% energy efficiency improvements.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125484185","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00088
Yifeng Tang, Cho-Li Wang
The popularity of Artificial Intelligence (AI) motivates novel domain-specific hardware named AI processors. With a design trade-off, the AI processors feature incredible computation power for matrix multiplications and activations, while some leave other operations less powerful, e.g., scalar operations and vectorized comparisons & selections. For k-nearest neighbors (k-NN) algorithm, consisting of distance computation phase and k-selection phase, while the former is naturally accelerated, the previous efficient k-selection becomes problematic. Moreover, limited memory forces k-NN to adopt a mini-batch manner with tiling technique. As the distance computation’s results are the k-selection’s inputs, the former’s tiling shape determines that of the latter. Since the two phases execute on separate hardware units requiring different performance analyses, whether the former’s tiling strategies benefit the latter and entire k-NN is doubtful.To address the new challenges brought by the AI processors, this paper proposes SelB-k-NN (Selection-Bitonic-k-NN), a mini-batch algorithm inspired by selection sort and bitonic k-selection. SelB-k-NN avoids the expansion of the weakly-supported operations on the huge scale of datasets. To apply SelB-k-NN to various AI processors, we propose two algorithms to reduce the hardware support requirements. Since the matrix multiplication operates data with the specifically-designed memory hierarchy which k-selection does not share, the tiling shape of the former cannot guarantee the best execution of the latter and vice versa. By quantifying the runtime workload variations of k-selection, we formulate an optimization problem to search for the optimal tiling shapes of both phases with an offline pruning method, which reduces the search space in the preprocessing stage. Evaluations show that on Huawei Ascend 310 AI processor, SelB-k-NN achieves 2.01× speedup of the bitonic k-selection, 23.93× of the heap approach, 78.52× of the CPU approach. For mini-batch SelB-k-NN, the optimal tiling shapes for two phases respectively achieve 1.48× acceleration compared with the matrix multiplication tiling shapes and 1.14× with the k-selection tiling shapes, with 72.80% of the search space pruned.
{"title":"SelB-k-NN: A Mini-Batch K-Nearest Neighbors Algorithm on AI Processors","authors":"Yifeng Tang, Cho-Li Wang","doi":"10.1109/IPDPS54959.2023.00088","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00088","url":null,"abstract":"The popularity of Artificial Intelligence (AI) motivates novel domain-specific hardware named AI processors. With a design trade-off, the AI processors feature incredible computation power for matrix multiplications and activations, while some leave other operations less powerful, e.g., scalar operations and vectorized comparisons & selections. For k-nearest neighbors (k-NN) algorithm, consisting of distance computation phase and k-selection phase, while the former is naturally accelerated, the previous efficient k-selection becomes problematic. Moreover, limited memory forces k-NN to adopt a mini-batch manner with tiling technique. As the distance computation’s results are the k-selection’s inputs, the former’s tiling shape determines that of the latter. Since the two phases execute on separate hardware units requiring different performance analyses, whether the former’s tiling strategies benefit the latter and entire k-NN is doubtful.To address the new challenges brought by the AI processors, this paper proposes SelB-k-NN (Selection-Bitonic-k-NN), a mini-batch algorithm inspired by selection sort and bitonic k-selection. SelB-k-NN avoids the expansion of the weakly-supported operations on the huge scale of datasets. To apply SelB-k-NN to various AI processors, we propose two algorithms to reduce the hardware support requirements. Since the matrix multiplication operates data with the specifically-designed memory hierarchy which k-selection does not share, the tiling shape of the former cannot guarantee the best execution of the latter and vice versa. By quantifying the runtime workload variations of k-selection, we formulate an optimization problem to search for the optimal tiling shapes of both phases with an offline pruning method, which reduces the search space in the preprocessing stage. Evaluations show that on Huawei Ascend 310 AI processor, SelB-k-NN achieves 2.01× speedup of the bitonic k-selection, 23.93× of the heap approach, 78.52× of the CPU approach. For mini-batch SelB-k-NN, the optimal tiling shapes for two phases respectively achieve 1.48× acceleration compared with the matrix multiplication tiling shapes and 1.14× with the k-selection tiling shapes, with 72.80% of the search space pruned.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"113 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117286368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00019
Haodi Lu, Haikun Liu, Chencheng Ye, Xiaofei Liao, Fubing Mao, Yu Zhang, Hai Jin
Modern storage systems typically replicate data on multiple servers to provide high reliability and availability. However, most commercially-deployed datastores often fail to offer low latency, high throughput, and strong consistency at the same time. This paper presents Whale, a Remote Direct Memory Access (RDMA) based primary-backup replication system for in-memory datastores. Whale achieves both low latency and strong consistency by decoupling metadata multicasting from data replication for all backup nodes, and using an optimistic commitment mechanism to respond to client write requests earlier. Whale achieves high throughput by propagating writes from the primary node to backup nodes asynchronously via RDMA-optimized chain replication. To further reduce the cost of data replication, we design a log-structured datastore to fully exploit the advantages of one-sided RDMA and Persistent Memory (PM). We implement Whale on a cluster equipped with PM and InfiniBand RDMA networks. Experimental results show that Whale achieves much higher throughput and lower latency than state-of-the-art replication protocols.
{"title":"Software-Defined, Fast and Strongly-Consistent Data Replication for RDMA-Based PM Datastores","authors":"Haodi Lu, Haikun Liu, Chencheng Ye, Xiaofei Liao, Fubing Mao, Yu Zhang, Hai Jin","doi":"10.1109/IPDPS54959.2023.00019","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00019","url":null,"abstract":"Modern storage systems typically replicate data on multiple servers to provide high reliability and availability. However, most commercially-deployed datastores often fail to offer low latency, high throughput, and strong consistency at the same time. This paper presents Whale, a Remote Direct Memory Access (RDMA) based primary-backup replication system for in-memory datastores. Whale achieves both low latency and strong consistency by decoupling metadata multicasting from data replication for all backup nodes, and using an optimistic commitment mechanism to respond to client write requests earlier. Whale achieves high throughput by propagating writes from the primary node to backup nodes asynchronously via RDMA-optimized chain replication. To further reduce the cost of data replication, we design a log-structured datastore to fully exploit the advantages of one-sided RDMA and Persistent Memory (PM). We implement Whale on a cluster equipped with PM and InfiniBand RDMA networks. Experimental results show that Whale achieves much higher throughput and lower latency than state-of-the-art replication protocols.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121540927","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/ipdps54959.2023.00006
{"title":"Message from the IPDPS 2023 Program Chairs","authors":"","doi":"10.1109/ipdps54959.2023.00006","DOIUrl":"https://doi.org/10.1109/ipdps54959.2023.00006","url":null,"abstract":"","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127699535","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-05-01DOI: 10.1109/IPDPS54959.2023.00087
P. Dutot, Yeu-Shin Fu, Nikhil Prasad, O. Sinnen
Scheduling task graphs with communication delay is a widely studied NP-hard problem. Many heuristics have been proposed, but there is no constant approximation algorithm for this classic model. In this paper, we focus on the scheduling of the important class of fork-join task graphs (describing many types of common computations) on homogeneous processors. For this sub-case, we propose a guaranteed algorithm with a $left( {1 + frac{m}{{m - 1}}} right)$-approximation factor, where m is the number of processors. The algorithm is not only the first constant approximation for an important sub-domain of the classic scheduling problem, it is also a practical algorithm that can obtain shorter makespans than known heuristics. To demonstrate this, we propose adaptations of known scheduling heuristic for the specific fork-join structure. In an extensive evaluation, we then implemented these algorithms and scheduled many fork-join graphs with up to thousands of tasks and various computation time distributions on up to hundreds of processors. Comparing the obtained results demonstrates the competitive nature of the proposed approximation algorithm.
{"title":"A Guaranteed Approximation Algorithm for Scheduling Fork-Joins with Communication Delay","authors":"P. Dutot, Yeu-Shin Fu, Nikhil Prasad, O. Sinnen","doi":"10.1109/IPDPS54959.2023.00087","DOIUrl":"https://doi.org/10.1109/IPDPS54959.2023.00087","url":null,"abstract":"Scheduling task graphs with communication delay is a widely studied NP-hard problem. Many heuristics have been proposed, but there is no constant approximation algorithm for this classic model. In this paper, we focus on the scheduling of the important class of fork-join task graphs (describing many types of common computations) on homogeneous processors. For this sub-case, we propose a guaranteed algorithm with a $left( {1 + frac{m}{{m - 1}}} right)$-approximation factor, where m is the number of processors. The algorithm is not only the first constant approximation for an important sub-domain of the classic scheduling problem, it is also a practical algorithm that can obtain shorter makespans than known heuristics. To demonstrate this, we propose adaptations of known scheduling heuristic for the specific fork-join structure. In an extensive evaluation, we then implemented these algorithms and scheduled many fork-join graphs with up to thousands of tasks and various computation time distributions on up to hundreds of processors. Comparing the obtained results demonstrates the competitive nature of the proposed approximation algorithm.","PeriodicalId":343684,"journal":{"name":"2023 IEEE International Parallel and Distributed Processing Symposium (IPDPS)","volume":"146 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116472569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：