arXiv subjects

Chuan Wu

Publications and source records attributed to Chuan Wu.

At least 91 records · Page 5Linked to original sources

arXiv2022

GADGET: Online Resource Optimization for Scheduling Ring-All-Reduce Learning Jobs

Fueled by advances in distributed deep learning (DDL), recent years have witnessed a rapidly growing demand for resource-intensive distributed/parallel computing to process DDL computing jobs. To resolve network communication bottleneck and load balancing issues in distributed computing, the so-called ``ring-all-reduce'' decentralized architecture has been increasingly adopted to remove the need for dedicated parameter servers. To date, however, there remains a lack of theoretical understanding on how to design resource optimization algorithms for efficiently scheduling ring-all-reduce DDL jobs in computing clusters. This motivates us to fill this gap by proposing a series of new resource scheduling designs for ring-all-reduce DDL jobs. Our contributions in this paper are three-fold: i) We propose a new resource scheduling analytical model for ring-all-reduce deep learning, which covers a wide range of objectives in DDL performance optimization (e.g., excessive training avoidance, energy efficiency, fairness); ii) Based on the proposed performance analytical model, we develop an efficient resource scheduling algorithm called GADGET (greedy ring-all-reduce distributed graph embedding technique), which enjoys a provable strong performance guarantee; iii) We conduct extensive trace-driven experiments to demonstrate the effectiveness of the GADGET approach and its superiority over the state of the art.

cs.DC↗

arXiv2021

Large-scale Machine Learning Cluster Scheduling via Multi-agent Graph Reinforcement Learning

Efficient scheduling of distributed deep learning (DL) jobs in large GPU clusters is crucial for resource efficiency and job performance. While server sharing among jobs improves resource utilization, interference among co-located DL jobs occurs due to resource contention. Interference-aware job placement has been studied, with white-box approaches based on explicit interference modeling and black-box schedulers with reinforcement learning. In today's clusters containing thousands of GPU servers, running a single scheduler to manage all arrival jobs in a timely and effective manner is challenging, due to the large workload scale. We adopt multiple schedulers in a large-scale cluster/data center, and propose a multi-agent reinforcement learning (MARL) scheduling framework to cooperatively learn fine-grained job placement policies, towards the objective of minimizing job completion time (JCT). To achieve topology-aware placements, our proposed framework uses hierarchical graph neural networks to encode the data center topology and server architecture. In view of a common lack of precise reward samples corresponding to different placements, a job interference model is further devised to predict interference levels in face of various co-locations, for training of the MARL schedulers. Testbed and trace-driven evaluations show that our scheduler framework outperforms representative scheduling schemes by more than 20% in terms of average JCT, and is adaptive to various machine learning cluster topologies.

cs.DC↗

arXiv2021

BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing

Graph neural networks (GNNs) have extended the success of deep neural networks (DNNs) to non-Euclidean graph data, achieving ground-breaking performance on various tasks such as node classification and graph property prediction. Nonetheless, existing systems are inefficient to train large graphs with billions of nodes and edges with GPUs. The main bottlenecks are the process of preparing data for GPUs - subgraph sampling and feature retrieving. This paper proposes BGL, a distributed GNN training system designed to address the bottlenecks with a few key ideas. First, we propose a dynamic cache engine to minimize feature retrieving traffic. By a co-design of caching policy and the order of sampling, we find a sweet spot of low overhead and high cache hit ratio. Second, we improve the graph partition algorithm to reduce cross-partition communication during subgraph sampling. Finally, careful resource isolation reduces contention between different data preprocessing stages. Extensive experiments on various GNN models and large graph datasets show that BGL significantly outperforms existing GNN training systems by 20.68x on average.

cs.LG↗

arXiv2021

Adversarial Deep Learning for Online Resource Allocation

Online algorithm is an important branch in algorithm design. Designing online algorithms with a bounded competitive ratio (in terms of worst-case performance) can be hard and usually relies on problem-specific assumptions. Inspired by adversarial training from Generative Adversarial Net (GAN) and the fact that competitive ratio of an online algorithm is based on worst-case input, we adopt deep neural networks to learn an online algorithm for a resource allocation and pricing problem from scratch, with the goal that the performance gap between offline optimum and the learned online algorithm can be minimized for worst-case input. Specifically, we leverage two neural networks as algorithm and adversary respectively and let them play a zero sum game, with the adversary being responsible for generating worst-case input while the algorithm learns the best strategy based on the input provided by the adversary. To ensure better convergence of the algorithm network (to the desired online algorithm), we propose a novel per-round update method to handle sequential decision making to break complex dependency among different rounds so that update can be done for every possible action, instead of only sampled actions. To the best of our knowledge, our work is the first using deep neural networks to design an online algorithm from the perspective of worst-case performance guarantee. Empirical studies show that our updating methods ensure convergence to Nash equilibrium and the learned algorithm outperforms state-of-the-art online algorithms under various settings.

cs.LG↗

arXiv2021

A Sum-of-Ratios Multi-Dimensional-Knapsack Decomposition for DNN Resource Scheduling

In recent years, to sustain the resource-intensive computational needs for training deep neural networks (DNNs), it is widely accepted that exploiting the parallelism in large-scale computing clusters is critical for the efficient deployments of DNN training jobs. However, existing resource schedulers for traditional computing clusters are not well suited for DNN training, which results in unsatisfactory job completion time performance. The limitations of these resource scheduling schemes motivate us to propose a new computing cluster resource scheduling framework that is able to leverage the special layered structure of DNN jobs and significantly improve their job completion times. Our contributions in this paper are three-fold: i) We develop a new resource scheduling analytical model by considering DNN's layered structure, which enables us to analytically formulate the resource scheduling optimization problem for DNN training in computing clusters; ii) Based on the proposed performance analytical model, we then develop an efficient resource scheduling algorithm based on the widely adopted parameter-server architecture using a sum-of-ratios multi-dimensional-knapsack decomposition (SMD) method to offer strong performance guarantee; iii) We conduct extensive numerical experiments to demonstrate the effectiveness of the proposed schedule algorithm and its superior performance over the state of the art.

cs.DC↗

arXiv2020

DAPPLE: A Pipelined Data Parallel Approach for Training Large Models

It is a challenging task to train large DNN models on sophisticated GPU platforms with diversified interconnect capabilities. Recently, pipelined training has been proposed as an effective approach for improving device utilization. However, there are still several tricky issues to address: improving computing efficiency while ensuring convergence, and reducing memory usage without incurring additional computing costs. We propose DAPPLE, a synchronous training framework which combines data parallelism and pipeline parallelism for large DNN models. It features a novel parallelization strategy planner to solve the partition and placement problems, and explores the optimal hybrid strategy of data and pipeline parallelism. We also propose a new runtime scheduling algorithm to reduce device memory usage, which is orthogonal to re-computation approach and does not come at the expense of training throughput. Experiments show that DAPPLE planner consistently outperforms strategies generated by PipeDream's planner by up to 3.23x under synchronous training scenarios, and DAPPLE runtime outperforms GPipe by 1.6x speedup of training throughput and reduces the memory consumption of 12% at the same time.

cs.DC↗

arXiv2019

Distributed Machine Learning through Heterogeneous Edge Systems

Many emerging AI applications request distributed machine learning (ML) among edge systems (e.g., IoT devices and PCs at the edge of the Internet), where data cannot be uploaded to a central venue for model training, due to their large volumes and/or security/privacy concerns. Edge devices are intrinsically heterogeneous in computing capacity, posing significant challenges to parameter synchronization for parallel training with the parameter server (PS) architecture. This paper proposes ADSP, a parameter synchronization scheme for distributed machine learning (ML) with heterogeneous edge systems. Eliminating the significant waiting time occurring with existing parameter synchronization models, the core idea of ADSP is to let faster edge devices continue training, while committing their model updates at strategically decided intervals. We design algorithms that decide time points for each worker to commit its model update, and ensure not only global model convergence but also faster convergence. Our testbed implementation and experiments show that ADSP outperforms existing parameter synchronization models significantly in terms of ML model convergence time, scalability and adaptability to large heterogeneity.

cs.DC↗

arXiv2019

Characterizing Deep Learning Training Workloads on Alibaba-PAI

Modern deep learning models have been exploited in various domains, including computer vision (CV), natural language processing (NLP), search and recommendation. In practical AI clusters, workloads training these models are run using software frameworks such as TensorFlow, Caffe, PyTorch and CNTK. One critical issue for efficiently operating practical AI clouds, is to characterize the computing and data transfer demands of these workloads, and more importantly, the training performance given the underlying software framework and hardware configurations. In this paper, we characterize deep learning training workloads from Platform of Artificial Intelligence (PAI) in Alibaba. We establish an analytical framework to investigate detailed execution time breakdown of various workloads using different training architectures, to identify performance bottleneck. Results show that weight/gradient communication during training takes almost 62% of the total execution time among all our workloads on average. The computation part, involving both GPU computing and memory access, are not the biggest bottleneck based on collective behavior of the workloads. We further evaluate attainable performance of the workloads on various potential software/hardware mappings, and explore implications on software architecture selection and hardware configurations. We identify that 60% of PS/Worker workloads can be potentially sped up when ported to the AllReduce architecture exploiting the high-speed NVLink for GPU interconnect, and on average 1.7X speedup can be achieved when Ethernet bandwidth is upgraded from 25 Gbps to 100 Gbps.

cs.PF↗

arXiv2019

DL2: A Deep Learning-driven Scheduler for Deep Learning Clusters

More and more companies have deployed machine learning (ML) clusters, where deep learning (DL) models are trained for providing various AI-driven services. Efficient resource scheduling is essential for maximal utilization of expensive DL clusters. Existing cluster schedulers either are agnostic to ML workload characteristics, or use scheduling heuristics based on operators' understanding of particular ML framework and workload, which are less efficient or not general enough. In this paper, we show that DL techniques can be adopted to design a generic and efficient scheduler. DL2 is a DL-driven scheduler for DL clusters, targeting global training job expedition by dynamically resizing resources allocated to jobs. DL2 advocates a joint supervised learning and reinforcement learning approach: a neural network is warmed up via offline supervised learning based on job traces produced by the existing cluster scheduler; then the neural network is plugged into the live DL cluster, fine-tuned by reinforcement learning carried out throughout the training progress of the DL jobs, and used for deciding job resource allocation in an online fashion. By applying past decisions made by the existing cluster scheduler in the preparatory supervised learning phase, our approach enables a smooth transition from existing scheduler, and renders a high-quality scheduler in minimizing average training completion time. We implement DL2 on Kubernetes and enable dynamic resource scaling in DL jobs on MXNet. Extensive evaluation shows that DL2 outperforms fairness scheduler (i.e., DRF) by 44.1% and expert heuristic scheduler (i.e., Optimus) by 17.5% in terms of average job completion time.

cs.LG↗

arXiv2018

Normalized Direction-preserving Adam

Adaptive optimization algorithms, such as Adam and RMSprop, have shown better optimization performance than stochastic gradient descent (SGD) in some scenarios. However, recent studies show that they often lead to worse generalization performance than SGD, especially for training deep neural networks (DNNs). In this work, we identify the reasons that Adam generalizes worse than SGD, and develop a variant of Adam to eliminate the generalization gap. The proposed method, normalized direction-preserving Adam (ND-Adam), enables more precise control of the direction and step size for updating weight vectors, leading to significantly improved generalization performance. Following a similar rationale, we further improve the generalization performance in classification tasks by regularizing the softmax logits. By bridging the gap between SGD and Adam, we also hope to shed light on why certain optimization algorithms generalize better than others.

cs.LG↗

arXiv2018

NFVactor: A Resilient NFV System using the Distributed Actor Model

Resilience functionality, including failure resilience and flow migration, is of pivotal importance in practical network function virtualization (NFV) systems. However, existing failure recovery procedures incur high packet processing delay due to heavyweight process-checkpointing, while flow migration has poor performance due to centralized control. This paper proposes NFVactor, a novel NFV system that aims to provide lightweight failure resilience and high-performance flow migration. NFVactor enables these by using actor model to provide a per-flow execution environment, so that each flow can replicate and migrate itself with improved parallelism, while the efficiency of the actor model is guaranteed by a carefully designed runtime system. Moreover, NFVactor achieves transparent resilience: once a new NF is implemented for NFVactor, the NF automatically acquires resilience support. Our evaluation result shows that NFVactor achieves 10Gbps packet processing, flow migration completion time that is 197 times faster than existing system, and packet processing delay stablized at around 20 microseconds during replication.

cs.NI↗

arXiv2018

Online Job Scheduling in Distributed Machine Learning Clusters

Nowadays large-scale distributed machine learning systems have been deployed to support various analytics and intelligence services in IT firms. To train a large dataset and derive the prediction/inference model, e.g., a deep neural network, multiple workers are run in parallel to train partitions of the input dataset, and update shared model parameters. In a shared cluster handling multiple training jobs, a fundamental issue is how to efficiently schedule jobs and set the number of concurrent workers to run for each job, such that server resources are maximally utilized and model training can be completed in time. Targeting a distributed machine learning system using the parameter server framework, we design an online algorithm for scheduling the arriving jobs and deciding the adjusted numbers of concurrent workers and parameter servers for each job over its course, to maximize overall utility of all jobs, contingent on their completion times. Our online algorithm design utilizes a primal-dual framework coupled with efficient dual subroutines, achieving good long-term performance guarantees with polynomial time complexity. Practical effectiveness of the online algorithm is evaluated using trace-driven simulation and testbed experiments, which demonstrate its outperformance as compared to commonly adopted scheduling algorithms in today's cloud systems.

cs.DC↗

arXiv2017

Optimal Posted Prices for Online Cloud Resource Allocation

We study online resource allocation in a cloud computing platform, through a posted pricing mechanism: The cloud provider publishes a unit price for each resource type, which may vary over time; upon arrival at the cloud system, a cloud user either takes the current prices, renting resources to execute its job, or refuses the prices without running its job there. We design pricing functions based on the current resource utilization ratios, in a wide array of demand-supply relationships and resource occupation durations, and prove worst-case competitive ratios of the pricing functions in terms of social welfare. In the basic case of a single-type, non-recycled resource (i.e., allocated resources are not later released for reuse), we prove that our pricing function design is optimal, in that any other pricing function can only lead to a worse competitive ratio. Insights obtained from the basic cases are then used to generalize the pricing functions to more realistic cloud systems with multiple types of resources, where a job occupies allocated resources for a number of time slots till completion, upon which time the resources are returned back to the cloud resource pool.

cs.NI↗

arXiv2017

Dynamic Scaling of Virtualized, Distributed Service Chains: A Case Study of IMS

The emerging paradigm of network function virtualization advocates deploying virtualized network functions (VNF) on standard virtualization platforms for significant cost reduction and management flexibility. There have been system designs for managing dynamic deployment and scaling of VNF service chains within one cloud data center. Many real-world network services involve geo-distributed service chains, with prominent examples of mobile core networks and IMSs (IP Multimedia Subsystems). Virtualizing these service chains requires efficient coordination of VNF deployment across different geo-distributed data centers over time, calling for new management system design. This paper designs a dynamic scaling system for geo-distributed VNF service chains, using the case of an IMS. IMSs are widely used subsystems for delivering multimedia services among mobile users in a 3G/4G network, whose virtualization has been broadly advocated in the industry for reducing cost, improving network usage efficiency and enabling dynamic network topology reconfiguration for performance optimization. Our scaling system design caters to key control-plane and data-plane service chains in an IMS, combining proactive and reactive approaches for timely, cost-effective scaling of the service chains. We evaluate our system design using real-world experiments on both emulated platforms and geo-distributed clouds.

cs.NI↗

arXiv2017

An Auction Approach to Spectrum Management in HetNets

The growing demand in mobile Internet access calls for high capacity and energy efficient cellular access with better cell coverage. The in-band relaying solution, proposed in LTE-Advanced, improves coverage without requiring additional spectrum for backhauling, making its deployment more economical and practical. However, in-band relay without careful management incurs low spectrum utilization and reduces the system capacity. We propose auction-based solutions that aim at dynamic spectrum resource sharing, maximizing the utilization of precious spectrum resources. We first present a truthful auction that ensures a theoretical performance guarantee in terms of social welfare. Then in an extended system model that focuses on addressing the heterogeneity of resource blocks, we design a more practical auction mechanism. We implement our proposed auctions under large scale real-world settings. Simulation results verify the efficacy of proposed auctions, showing improvements in both cell coverage and spectrum efficiency.

cs.NI↗

arXiv2016

Online Scaling of NFV Service Chains across Geo-distributed Datacenters

Network Function Virtualization (NFV) is an emerging paradigm that turns hardware-dependent implementation of network functions (i.e., middleboxes) into software modules running on virtualized platforms, for significant cost reduction and ease of management. Such virtual network functions (VNFs) commonly constitute service chains, to provide network services that traffic flows need to go through. Efficient deployment of VNFs for network service provisioning is key to realize the NFV goals. Existing efforts on VNF placement mostly deal with offline or one-time placement, ignoring the fundamental, dynamic deployment and scaling need of VNFs to handle practical time-varying traffic volumes. This work investigates dynamic placement of VNF service chains across geo-distributed datacenters to serve flows between dispersed source and destination pairs, for operational cost minimization of the service chain provider over the entire system span. An efficient online algorithm is proposed, which consists of two main components: (1) A regularization-based approach from online learning literature to convert the offline optimal deployment problem into a sequence of one-shot regularized problems, each to be efficiently solved in one time slot; (2) An online dependent rounding scheme to derive feasible integer solutions from the optimal fractional solutions of the one-shot problems, and to guarantee a good competitive ratio of the online algorithm over the entire time span. We verify our online algorithm with solid theoretical analysis and trace-driven simulations under realistic settings.

cs.NI↗

arXiv2016

A Truthful $(1-ε)$-Optimal Mechanism for On-demand Cloud Resource Provisioning

On-demand resource provisioning in cloud computing provides tailor-made resource packages (typically in the form of VMs) to meet users' demands. Public clouds nowadays provide more and more elaborated types of VMs, but have yet to offer the most flexible dynamic VM assembly, which is partly due to the lack of a mature mechanism for pricing tailor-made VMs on the spot. This work proposes an efficient randomized auction mechanism based on a novel application of smoothed analysis and randomized reduction, for dynamic VM provisioning and pricing in geo-distributed cloud data centers. This auction, to the best of our knowledge, is the first one in literature that achieves (i) truthfulness in expectation, (ii) polynomial running time in expectation, and (iii) $(1-ε)$-optimal social welfare in expectation for resource allocation, where $ε$ can be arbitrarily close to 0. Our mechanism consists of three modules: (1) an exact algorithm to solve the NP-hard social welfare maximization problem, which runs in polynomial time in expectation, (2) a perturbation-based randomized resource allocation scheme which produces a VM provisioning solution that is $(1-ε)$-optimal, and (3) an auction mechanism that applies the perturbation-based scheme for dynamic VM provisioning and prices the customized VMs using a randomized VCG payment, with a guarantee in truthfulness in expectation. We validate the efficacy of the mechanism through careful theoretical analysis and trace-driven simulations.

cs.GT↗

arXiv2016

Online Influence Maximization in Non-Stationary Social Networks

Social networks have been popular platforms for information propagation. An important use case is viral marketing: given a promotion budget, an advertiser can choose some influential users as the seed set and provide them free or discounted sample products; in this way, the advertiser hopes to increase the popularity of the product in the users' friend circles by the world-of-mouth effect, and thus maximizes the number of users that information of the production can reach. There has been a body of literature studying the influence maximization problem. Nevertheless, the existing studies mostly investigate the problem on a one-off basis, assuming fixed known influence probabilities among users, or the knowledge of the exact social network topology. In practice, the social network topology and the influence probabilities are typically unknown to the advertiser, which can be varying over time, i.e., in cases of newly established, strengthened or weakened social ties. In this paper, we focus on a dynamic non-stationary social network and design a randomized algorithm, RSB, based on multi-armed bandit optimization, to maximize influence propagation over time. The algorithm produces a sequence of online decisions and calibrates its explore-exploit strategy utilizing outcomes of previous decisions. It is rigorously proven to achieve an upper-bounded regret in reward and applicable to large-scale social networks. Practical effectiveness of the algorithm is evaluated using both synthetic and real-world datasets, which demonstrates that our algorithm outperforms previous stationary methods under non-stationary conditions.

cs.SI↗