Ensuring the highest training throughput to maximize resource efficiency, while maintaining fairness among users, is critical for deep learning (DL) training in heterogeneous GPU clusters. However, current DL schedulers provide only limited fairness properties and suboptimal training throughput, impeding tenants from effectively leveraging heterogeneous resources. The underlying design challenge stems from inherent conflicts between efficiency and fairness properties.
In this paper, we introduce OEF, a new resource allocation framework specifically developed for achieving optimal resource efficiency and ensuring diverse fairness properties in heterogeneous GPU clusters. By integrating resource efficiency and fairness within a global optimization framework, OEF is capable of providing users with maximized overall efficiency, as well as various guarantees of fairness, in both cooperative and non-cooperative environments. We have implemented OEF in a cluster resource manager and conducted large-scale experiments, showing that OEF can improve the overall training throughput by up to 32% while improving fairness compared to state-of-the-art heterogeneity-aware schedulers.
Deep Neural Network (DNN) models have continuously been growing in size in order to improve the accuracy and quality of the models. Moreover, for training of large DNN models, the use of heterogeneous GPUs is inevitable due to the short release cycle of new GPU architectures. In this paper, we investigate how to enable training of large DNN models on a heterogeneous GPU cluster that possibly includes whimpy GPUs that, as a standalone, could not be used for training. We present a DNN training system, HetPipe (Heterogeneous Pipeline), that integrates pipelined model parallelism (PMP) with data p…
The scaling up of deep neural networks has been demonstrated to be effective in improving model quality, but also encompasses several training challenges in terms of training efficiency, programmability, and resource adaptability. We present Whale, a general and efficient distributed training framework for giant models. To support various parallel strategies and their hybrids, Whale generalizes the programming interface by defining two new primitives in the form of model annotations, allowing for incorporating user hints. The Whale runtime utilizes those annotations and performs graph optimiza…
The Preconditioned Conjugate Gradient (PCG) method is widely used for solving linear systems of equations with sparse matrices. A recent version of PCG, Pipelined PCG, eliminates the dependencies in the computations of the PCG algorithm so that the non-dependent computations can be overlapped with communication. In this paper, we propose three methods for efficient execution of the Pipelined PCG algorithm on GPU accelerated heterogeneous architectures. The first two methods achieve task-parallelism using asynchronous executions of different tasks on CPU cores and GPU. The third method achieves…
HETHUB: A Distributed Training System with Heterogeneous Cluster for Large-Scale Models
Infinigence-AI；Tsinghua University（清华大学）；Hangzhou Dianzi University（杭州电子科技大学）；China Mobile Research Institute（中国移动研究院）；Shanghai Jiao Tong University（上海交通大学）
University of California Merced（加州大学默塞德分校）；Lawrence Berkeley National Laboratory（劳伦斯伯克利国家实验室）
Motivated by extreme multi-label classification applications, we consider training deep learning models over sparse data in multi-GPU servers. The variance in the number of non-zero features across training batches and the intrinsic GPU heterogeneity combine to limit accuracy and increase the time to convergence. We address these challenges with Adaptive SGD, an adaptive elastic model averaging stochastic gradient descent algorithm for heterogeneous multi-GPUs that is characterized by dynamic scheduling, adaptive batch size scaling, and normalized model merging. Instead of statically partition…
Binarized Neural Networks (BNNs) significantly reduce the computation and memory demands with binarized weights and activations compared to full-precision NNs. Executing a layer in a BNN on different devices of a heterogeneous multiprocessor platform consisting of CPU and GPU can affect the inference performance, i.e., accuracy and latency. Usually, a heterogeneous HW platform consisting of a CPU and a GPU is available to execute the BNN workloads. However, to use the heterogeneous HW effectively, it is necessary to find an efficient strategy for BNN workload mapping. In this work, we propose …
Matrix Factorization (MF) has been widely applied in machine learning and data mining. A large number of algorithms have been studied to factorize matrices. Among them, stochastic gradient descent (SGD) is a commonly used method. Heterogeneous systems with multi-core CPUs and GPUs have become more and more promising recently due to the prevalence of GPUs in general-purpose data-parallel applications. Due to the large computational cost of MF, we aim to improve the efficiency of SGD-based MF computation by utilizing the massive parallel processing power of heterogeneous multiprocessors. The mai…
This documentation is designed for beginners in Graphics Processing Unit (GPU)-programming and who want to get familiar with OpenACC and OpenMP offloading models. Here we present an overview of these two programming models as well as of the GPU-architectures. Specifically, we provide some insights into the functionality of these models and perform experiments involving different directives and discuss their performance. This is achieved through the use of a mini-application based on solving numerically the Laplace equation. Such experiments reveal the benefit of the use of GPU, which in our ca…
Energy conservation of large data centers for high-performance computing workloads, such as deep learning with big data, is of critical significance, where cutting down a few percent of electricity translates into million-dollar savings. This work studies energy conservation on emerging CPU-GPU hybrid clusters through dynamic voltage and frequency scaling (DVFS). We aim at minimizing the total energy consumption of processing a batch of offline tasks or a sequence of real-time tasks under deadline constraints. We derive a fast and accurate analytical model to compute the appropriate voltage/fr…
Graph neural networks (GNN) have shown great success in learning from graph-structured data. They are widely used in various applications, such as recommendation, fraud detection, and search. In these domains, the graphs are typically large and heterogeneous, containing many millions or billions of vertices and edges of different types. To tackle this challenge, we develop DistDGLv2, a system that extends DistDGL for training GNNs on massive heterogeneous graphs in a mini-batch fashion, using distributed hybrid CPU/GPU training. DistDGLv2 places graph data in distributed CPU memory and perform…
