Recent trends in high-performance computing and deep learning have led to the proliferation of studies on large-scale deep neural network training. However, the frequent communication requirements among computation nodes drastically slows the overall training speeds, which causes bottlenecks in distributed training, particularly in clusters with limited network bandwidths. To mitigate the drawbacks of distributed communications, researchers have proposed various optimization strategies. In this paper, we provide a comprehensive survey of communication strategies from both an algorithm viewpoint and a computer network perspective. Algorithm optimizations focus on reducing the communication volumes used in distributed training, while network optimizations focus on accelerating the communications between distributed devices. At the algorithm level, we describe how to reduce the number of communication rounds and transmitted bits per round. In addition, we elucidate how to overlap computation and communication. At the network level, we discuss the effects caused by network infrastructures, including logical communication schemes and network protocols. Finally, we extrapolate the potential future challenges and new research directions to accelerate communications for distributed deep neural network training.
Large-scale distributed training is increasingly becoming communication bound. Many gradient compression algorithms have been proposed to reduce the communication overhead and improve scalability. However, it has been observed that in some cases gradient compression may even harm the performance of distributed training. In this paper, we propose MergeComp, a compression scheduler to optimize the scalability of communication-efficient distributed training. It automatically schedules the compression operations to optimize the performance of compression algorithms without the knowledge of model a…
Graph Neural Networks(GNNs) are a family of neural models tailored for graph-structure data and have shown superior performance in learning representations for graph-structured data. However, training GNNs on large graphs remains challenging and a promising direction is distributed GNN training, which is to partition the input graph and distribute the workload across multiple machines. The key bottleneck of the existing distributed GNNs training framework is the across-machine communication induced by the dependency on the graph data and aggregation operator of GNNs. In this paper, we study th…
For high-dimensional data, there are huge communication costs for distributed GBDT because the communication volume of GBDT is related to the number of features. To overcome this problem, we propose a novel gradient boosting algorithm, the Gradient Boosting Untrained Neural Network(GBUN). GBUN ensembles the untrained randomly generated neural network that softly distributes data samples to multiple neuron outputs and dramatically reduces the communication costs for distributed learning. To avoid creating huge neural networks for high-dimensional data, we extend Simhash algorithm to mimic forwa…
In the paper notion of communication dualism id formalized and explained in Petri net interpretation. We consider communication dualism a basic property of communication in distributed systems. The formalization is done in the Integrated Model of Distributed Systems (IMDS) where synchronous communication, as well as asynchronous message-passing and variable-sharing are modeled in a common framework. In the light of this property, communication in distributed systems can be seen as a two-dimensional phenomenon with passing being its spatial dimension and sharing its temporal dimension. Any dist…
H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology（乔治亚理工学院H. Milton Stewart工业与系统工程学院）；Department of Electrical Engineering, Technion（以色列理工学院电气工程系）
Distributed optimization has gained a surge of interest in recent years. In this paper we propose a distributed projection free algorithm named Distributed Conditional Gradient Sliding(DCGS). Compared to the state-of-the-art distributed Frank-Wolfe algorithm, our algorithm attains the same communication complexity under much more realistic assumptions. In contrast to the consensus based algorithm, DCGS is based on the primal-dual algorithm, yielding a modular analysis that can be exploited to improve linear oracle complexity whenever centralized Frank-Wolfe can be improved. We demonstrate this…
Large neural network models present a hefty communication challenge to distributed Stochastic Gradient Descent (SGD), with a communication complexity of O(n) per worker for a model of n parameters. Many sparsification and quantization techniques have been proposed to compress the gradients, some reducing the communication complexity to O(k), where k << n. In this paper, we introduce a strategy called two-level gradient averaging (A2SGD) to consolidate all gradients down to merely two local averages per worker before the computation of two global averages for an updated model. A2SGD also retain…
In this paper, a distributed convex optimization algorithm, termed \emph{distributed coordinate dual averaging} (DCDA) algorithm, is proposed. The DCDA algorithm addresses the scenario of a large distributed optimization problem with limited communication among nodes in the network. Currently known distributed subgradient methods, such as the distributed dual averaging or the distributed alternating direction method of multipliers algorithms, assume that nodes can exchange messages of large cardinality. Such network communication capabilities are not valid in many scenarios of practical releva…
LAAS-CNRS（LAAS-法国国家科学研究中心）；Université de Toulouse（图卢兹大学）；CNRS（法国国家科学研究中心）
We describe several features of parallel or distributed asynchronous iterative algorithms such as unbounded delays, possible out of order messages or flexible communication. We concentrate on the concept of macroiteration sequence which was introduced in order to study the convergence or termination of asynchronous iterations. A survey of asynchronous iterations for convex optimization problems is also presented. Finally, a new result of convergence for parallel or distributed asynchronous iterative algorithms with flexible communication for convex optimization problems and machine learning is…
School of Statistics and Mathematics（统计与数学学院）；Nanjing Audit University（南京审计大学）
Communication efficiency and robustness are two major issues in modern distributed learning framework. This is due to the practical situations where some computing nodes may have limited communication power or may behave adversarial behaviors. To address the two issues simultaneously, this paper develops two communication-efficient and robust distributed learning algorithms for convex problems. Our motivation is based on surrogate likelihood framework and the median and trimmed mean operations. Particularly, the proposed algorithms are provably robust against Byzantine failures, and also achie…
In distributed statistical learning, $N$ samples are split across $m$ machines and a learner wishes to use minimal communication to learn as well as if the examples were on a single machine. This model has received substantial interest in machine learning due to its scalability and potential for parallel speedup. However, in high-dimensional settings, where the number examples is smaller than the number of features ("dimension"), the speedup afforded by distributed learning may be overshadowed by the cost of communicating a single example. This paper investigates the following question: When i…
Rensselaer Polytechnic Institute（伦斯勒理工学院）；University of California, Los Angeles（加州大学洛杉矶分校）
Stochastic gradient descent (SGD) has taken the stage as the primary workhorse for large-scale machine learning. It is often used with its adaptive variants such as AdaGrad, Adam, and AMSGrad. This paper proposes an adaptive stochastic gradient descent method for distributed machine learning, which can be viewed as the communication-adaptive counterpart of the celebrated Adam method - justifying its name CADA. The key components of CADA are a set of new rules tailored for adaptive stochastic gradients that can be implemented to save communication upload. The new algorithms adaptively reuse the…
University of California, Berkeley（加州大学伯克利分校）；Duke University（杜克大学）
Task-based distributed frameworks (e.g., Ray, Dask, Hydro) have become increasingly popular for distributed applications that contain asynchronous and dynamic workloads, including asynchronous gradient descent, reinforcement learning, and model serving. As more data-intensive applications move to run on top of task-based systems, collective communication efficiency has become an important problem. Unfortunately, traditional collective communication libraries (e.g., MPI, Horovod, NCCL) are an ill fit, because they require the communication schedule to be known before runtime and they do not pro…
Moscow Institute of Physics and Technology（莫斯科物理技术学院）；HSE University（高等经济大学）；IITP RAS（俄罗斯科学院信息传输问题研究所）；Adyghe State University（阿迪格国立大学）
Variational inequalities are an important tool, which includes minimization, saddles, games, fixed-point problems. Modern large-scale and computationally expensive practical applications make distributed methods for solving these problems popular. Meanwhile, most distributed systems have a basic problem - a communication bottleneck. There are various techniques to deal with it. In particular, in this paper we consider a combination of two popular approaches: compression and data similarity. We show that this synergy can be more effective than each of the approaches separately in solving distri…
University of Massachusetts, Dartmouth（马萨诸塞大学达特茅斯分校）；Lovelace Respiratory Research Institute（洛夫莱斯呼吸研究所）
The last decades have seen a surge of interests in distributed computing thanks to advances in clustered computing and big data technology. Existing distributed algorithms typically assume {\it all the data are already in one place}, and divide the data and conquer on multiple machines. However, it is increasingly often that the data are located at a number of distributed sites, and one wishes to compute over all the data with low communication overhead. For spectral clustering, we propose a novel framework that enables its computation over such distributed data, with "minimal" communications …
Communication-Efficient Distributed Deep Learning: A Comprehensive Survey
Hong Kong Baptist University（香港浸会大学）；Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））；The Hong Kong University of Science and Technology（香港科技大学）；The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））
Penn State Behrend（宾夕法尼亚州立大学贝伦德分校）；University of California, Davis（加州大学戴维斯分校）；University of Iowa（爱荷华大学）
Due to the big size of data and limited data storage volume of a single computer or a single server, data are often stored in a distributed manner. Thus, performing large-scale machine learning operations with the distributed datasets through communication networks is often required. In this paper, we study the convergence rate of the distributed dual coordinate ascent for distributed machine learning problems in a general tree-structured network. Since a tree network model can be understood as the generalization of a star network model, our algorithm can be thought of as the generalization of…
Distributed deep learning is becoming increasingly popular due to the expanding demand for computing resources for deep learning models with a larger amount of parameters. Different from traditional training approaches, data-parallel training allows multiple compute nodes to train large deep learning models simultaneously in order to boost the training efficiency. In this paper, we present and compare six strategies for data-parallel training using PyTorch on the language model GPT-2 with 100M parameters using a qualitative approach. These strategies are Single GPU, Single Parameter Server, Di…
Zhejiang University（浙江大学）；Rensselaer Polytechnic Institute（伦斯勒理工学院）；University of Minnesota, Twin Cities（明尼苏达大学双城分校）
The present paper develops a novel aggregated gradient approach for distributed machine learning that adaptively compresses the gradient communication. The key idea is to first quantize the computed gradients, and then skip less informative quantized gradient communications by reusing outdated gradients. Quantizing and skipping result in `lazy' worker-server communications, which justifies the term Lazily Aggregated Quantized gradient that is henceforth abbreviated as LAQ. Our LAQ can provably attain the same linear convergence rate as the gradient descent in the strongly convex case, while ef…
Using the matrix factorization technique in machine learning is very common mainly in areas like recommender systems. Despite its high prediction accuracy and its ability to avoid over-fitting of the data, the Bayesian Probabilistic Matrix Factorization algorithm (BPMF) has not been widely used on large scale data because of the prohibitive cost. In this paper, we propose a distributed high-performance parallel implementation of the BPMF using Gibbs sampling on shared and distributed architectures. We show by using efficient load balancing using work stealing on a single node, and by using asy…
We consider communication-efficient weighted and unweighted (uniform) random sampling from distributed data streams presented as a sequence of mini-batches of items. This is a natural model for distributed streaming computation, and our goal is to showcase its usefulness. We present and analyze fully distributed, communication-efficient algorithms for both versions of the problem. An experimental evaluation of weighted reservoir sampling on up to 256 nodes (5120 processors) shows good speedups, while theoretical analysis promises further scaling to much larger machines.
Amazon Web Services（亚马逊云科技）；Hong Kong University of Science and Technology（香港科技大学）
Communication overhead is a major bottleneck hampering the scalability of distributed machine learning systems. Recently, there has been a surge of interest in using gradient compression to improve the communication efficiency of distributed neural network training. Using 1-bit quantization, signSGD with majority vote achieves a 32x reduction on communication cost. However, its convergence is based on unrealistic assumptions and can diverge in practice. In this paper, we propose a general distributed compressed SGD with Nesterov's momentum. We consider two-way compression, which compresses the…
Pennsylvania State University（宾夕法尼亚州立大学）；University of Virginia（弗吉尼亚大学）
Due to the explosion in the size of the training datasets, distributed learning has received growing interest in recent years. One of the major bottlenecks is the large communication cost between the central server and the local workers. While error feedback compression has been proven to be successful in reducing communication costs with stochastic gradient descent (SGD), there are much fewer attempts in building communication-efficient adaptive gradient methods with provable guarantees, which are widely used in training large-scale machine learning models. In this paper, we propose a new com…
For parallel breadth first search (BFS) algorithm on large-scale distributed memory systems, communication often costs significantly more than arithmetic and limits the scalability of the algorithm. In this paper we sufficiently reduce the communication cost in distributed BFS by compressing and sieving the messages. First, we leverage a bitmap compression algorithm to reduce the size of messages before communication. Second, we propose a novel distributed directory algorithm, cross directory, to sieve the redundant data in messages. Experiments on a 6,144-core SMP cluster show our algorithm o…
This paper investigates Nash equilibrium (NE) seeking problems for noncooperative games over multi-players networks with finite bandwidth communication. A distributed quantized algorithm is presented, which consists of local gradient play, distributed decision estimating, and adaptive quantization. Exponential convergence of the algorithm is established, and a relationship between the convergence rate and the bandwidth is quantitatively analyzed. Finally, a simulation of an energy consumption game is presented to validate the proposed results.
Distributed learning platforms for processing large scale data-sets are becoming increasingly prevalent. In typical distributed implementations, a centralized master node breaks the data-set into smaller batches for parallel processing across distributed workers to achieve speed-up and efficiency. Several computational tasks are of sequential nature, and involve multiple passes over the data. At each iteration over the data, it is common practice to randomly re-shuffle the data at the master node, assigning different batches for each worker to process. This random re-shuffling operation comes …
Collective Communication for Distributed LLM Systems: Planning, Runtime Adaptation, and Computation Coordination
School of Software, Beihang University（北京航空航天大学软件学院）；State Key Laboratory of Internet Architecture, Tsinghua University（清华大学互联网架构国家重点实验室）；Infrawaves（英伟拉维斯）
Distributed Ledger Technology (DLT) is promising to become the foundation of many decentralised systems. However, the unbalanced and unregulated network layout contributes to the inefficiency of DLT especially in the Internet of Things (IoT) environments, where nodes connect to only a limited number of peers. The data communication speed globally is unbalanced and does not live up to the constraints of efficient real-time distributed systems. In this paper, we introduce a new communication protocol, which enables nodes to calculate the tradeoff between connecting/disconnecting a peer in a comp…
CondenseGraph: Communication-Efficient Distributed GNN Training via On-the-Fly Graph Condensation
University of Michigan（密歇根大学）；University of Southern California（南加州大学）；New York University（纽约大学）；Rice University（莱斯大学）；University of California, Berkeley（加利福尼亚大学伯克利分校）
This paper discusses some generic approach for developing grid-based framework for enabling establishment of workflows comprising existing software in computational sciences areas. We highlight the main requirements addressed the developing of such framework. Some strategies for enabling interoperability between convenient computation software in the grid environment has been shown. The UML based instruments of graphical description of workflows for the developing system has been suggested.
We investigate fast and communication-efficient algorithms for the classic problem of minimizing a sum of strongly convex and smooth functions that are distributed among $n$ different nodes, which can communicate using a limited number of bits. Most previous communication-efficient approaches for this problem are limited to first-order optimization, and therefore have \emph{linear} dependence on the condition number in their communication complexity. We show that this dependence is not inherent: communication-efficient methods can in fact have sublinear dependence on the condition number. For …
Leiden University（莱顿大学）；University of Amsterdam（阿姆斯特丹大学）
Nowadays, many AI applications utilizing resource-constrained edge devices (e.g., small mobile robots, tiny IoT devices, etc.) require Convolutional Neural Network (CNN) inference on a distributed system at the edge due to limited resources of a single edge device to accommodate and execute a large CNN. There are four main partitioning strategies that can be utilized to partition a large CNN model and perform distributed CNN inference on multiple devices at the edge. However, to the best of our knowledge, no research has been conducted to investigate how these four partitioning strategies affe…
Privacy issues and communication cost are both major concerns in distributed optimization. There is often a trade-off between them because the encryption methods required for privacy-preservation often incur expensive communication bandwidth. To address this issue, we, in this paper, propose a quantization-based approach to achieve both communication efficient and privacy-preserving solutions in the context of distributed optimization. By deploying an adaptive differential quantization scheme, we allow each node in the network to achieve its optimum solution with a low communication cost while…
This paper presents distributed conjugate gradient algorithms for distributed parameter estimation and spectrum estimation over wireless sensor networks. In particular, distributed conventional conjugate gradient (CCG) and modified conjugate gradient (MCG) are considered, together with incremental and diffusion adaptive solutions. The distributed CCG and MCG algorithms have an improved performance in terms of mean square error as compared with least--mean square (LMS)--based algorithms and a performance that is close to recursive least--squares (RLS) algorithms. In comparison with existing cen…
How can we optimally trade extra computing power to reduce the communication load in distributed computing? We answer this question by characterizing a fundamental tradeoff between computation and communication in distributed computing, i.e., the two are inversely proportional to each other. More specifically, a general distributed computing framework, motivated by commonly used structures like MapReduce, is considered, where the overall computation is decomposed into computing a set of "Map" and "Reduce" functions distributedly across multiple computing nodes. A coded scheme, named "Coded Dis…
Infinite games with imperfect information are known to be undecidable unless the information flow is severely restricted. One fundamental decidable case occurs when there is a total ordering among players, such that each player has access to all the information that the following ones receive. In this paper we consider variations of this hierarchy principle for synchronous games with perfect recall, and identify new decidable classes for which the distributed synthesis problem is solvable with finite-state strategies. In particular, we show that decidability is maintained when the information …
College of Computer, National University of Defense Technology（国防科技大学计算机学院）
Distributed stochastic gradient descent (SGD) approach has been widely used in large-scale deep learning, and the gradient collective method is vital to ensure the training scalability of the distributed deep learning system. Collective communication such as AllReduce has been widely adopted for the distributed SGD process to reduce the communication time. However, AllReduce incurs large bandwidth resources while most gradients are sparse in many cases since many gradient values are zeros and should be efficiently compressed for bandwidth saving. To reduce the sparse gradient communication ove…
We give lower bounds on the communication complexity required to solve several computational problems in a distributed-memory parallel machine, namely standard matrix multiplication, stencil computations, comparison sorting, and the Fast Fourier Transform. We revisit the assumptions under which preceding results were derived and provide new lower bounds which use much weaker and appropriate hypotheses. Our bounds rely on a mild assumption on work distribution, and strengthen previous results which require either the computation to be balanced among the processors, or specific initial distribut…
We study efficient distributed algorithms for the fundamental problem of principal component analysis and leading eigenvector computation on the sphere, when the data are randomly distributed among a set of computational nodes. We propose a new quantized variant of Riemannian gradient descent to solve this problem, and prove that the algorithm converges with high probability under a set of necessary spherical-convexity properties. We give bounds on the number of bits transmitted by the algorithm under common initialization schemes, and investigate the dependency on the problem dimension in eac…
Model-based development is a widely-used method to describe complex systems that enables the rapid prototyping. Advances in the science of distributed systems has led to the development of large scale statechart models which are distributed among multiple locations. Taking medicine for example, models of best-practice guidelines during rural ambulance transport are distributed across hospital settings from a rural hospital, to an ambulance, to a central tertiary hospital. Unfortunately, these medical models require continuous and real-time communication across individual medical models in phys…
When the data is distributed across multiple servers, lowering the communication cost between the servers (or workers) while solving the distributed learning problem is an important problem and is the focus of this paper. In particular, we propose a fast, and communication-efficient decentralized framework to solve the distributed machine learning (DML) problem. The proposed algorithm, Group Alternating Direction Method of Multipliers (GADMM) is based on the Alternating Direction Method of Multipliers (ADMM) framework. The key novelty in GADMM is that it solves the problem in a decentralized t…
We study distributed estimation of a Gaussian mean under communication constraints in a decision theoretical framework. Minimax rates of convergence, which characterize the tradeoff between the communication costs and statistical accuracy, are established in both the univariate and multivariate settings. Communication-efficient and statistically optimal procedures are developed. In the univariate case, the optimal rate depends only on the total communication budget, so long as each local machine has at least one bit. However, in the multivariate case, the minimax rate depends on the specific a…
To process and transfer large amounts of data in emerging wireless services, it has become increasingly appealing to exploit distributed data communication and learning. Specifically, edge learning (EL) enables local model training on geographically disperse edge nodes and minimizes the need for frequent data exchange. However, the current design of separating EL deployment and communication optimization does not yet reap the promised benefits of distributed signal processing, and sometimes suffers from excessive signalling overhead, long processing delay, and unstable learning convergence. In…
The overall performance of a distributed system is highly dependent on the communication efficiency of the system. Although network resources (links, bandwidth) are becoming increasingly more available, the communication performance of data transfers involving large volumes of data does not necessarily improve at the same rate. This is due to the inefficient usage of the available network resources. A solution to this problem consists of data transfer scheduling techniques, which manage and allocate the network resources in an efficient manner. In this paper we present several online and offli…
Motivated by the needs of resiliency, scalability, and plug-and-play operation, distributed decision-making is becoming increasingly prevalent. The problem of achieving consensus in a multi-agent system is at the core of distributed decision-making. In this article, we study the problem of achieving average consensus over a directed multi-agent network when the communication links are corrupted with noise. We propose an algorithm where each agent updates its estimates based on the local mixing of information and adds its weighted noise-free initial information to its updates during every itera…
