Deep Learning (DL) models have achieved superior performance. Meanwhile, computing hardware like NVIDIA GPUs also demonstrated strong computing scaling trends with 2x throughput and memory bandwidth for each generation. With such strong computing scaling of GPUs, multi-tenant deep learning inference by co-locating multiple DL models onto the same GPU becomes widely deployed to improve resource utilization, enhance serving throughput, reduce energy cost, etc. However, achieving efficient multi-tenant DL inference is challenging which requires thorough full-stack system optimization. This survey aims to summarize and categorize the emerging challenges and optimization opportunities for multi-tenant DL inference on GPU. By overviewing the entire optimization stack, summarizing the multi-tenant computing innovations, and elaborating the recent technological advances, we hope that this survey could shed light on new optimization perspectives and motivate novel works in future large-scale DL system optimization.
GPU technology has been improving at an expedited pace in terms of size and performance, empowering HPC and AI/ML researchers to advance the scientific discovery process. However, this also leads to inefficient resource usage, as most GPU workloads, including complicated AI/ML models, are not able to utilize the GPU resources to their fullest extent -- encouraging support for GPU multi-tenancy. We propose MISO, a technique to exploit the Multi-Instance GPU (MIG) capability on the latest NVIDIA datacenter GPUs (e.g., A100, H100) to dynamically partition GPU resources among co-located jobs. MISO…
We study how to support elasticity, i.e., the ability to dynamically adjust the parallelism (number of GPUs), for deep neural network (DNN) training. Elasticity can benefit multi-tenant GPU cluster management in many ways, e.g., achieving various scheduling objectives (e.g., job throughput, job completion time, GPU efficiency) according to cluster load variations, maximizing the use of transient idle resources, performance profiling, job migration, and straggler mitigation. However, existing parallelism adjustment strategies incur high overheads, which hinder many applications from making effe…
With widespread advances in machine learning, a number of large enterprises are beginning to incorporate machine learning models across a number of products. These models are typically trained on shared, multi-tenant GPU clusters. Similar to existing cluster computing workloads, scheduling frameworks aim to provide features like high efficiency, resource isolation, fair sharing across users, etc. However Deep Neural Network (DNN) based workloads, predominantly trained on GPUs, differ in two significant ways from traditional big data analytics workloads. First, from a cluster utilization perspe…
Graphics Processing Units (GPUs) are becoming popular accelerators in modern High-Performance Computing (HPC) clusters. Installing GPUs on each node of the cluster is not efficient resulting in high costs and power consumption as well as underutilisation of the accelerator. The research reported in this paper is motivated towards the use of few physical GPUs by providing cluster nodes access to remote GPUs on-demand for a financial risk application. We hypothesise that sharing GPUs between several nodes, referred to as multi-tenancy, reduces the execution time and energy consumed by an applica…
University of Toronto（多伦多大学）；National University of Singapore（新加坡国立大学）；SenseTime（商汤科技）
Multi-tenant machine learning services have become emerging data-intensive workloads in data centers with heavy usage of GPU resources. Due to the large scale, many tuning parameters and heavy resource usage, it is usually impractical to evaluate and benchmark those machine learning services on real clusters. In this demonstration, we present AnalySIM, a cluster simulator that allows efficient design explorations for multi-tenant machine learning services. Specifically, by trace-driven cluster workload simulation, AnalySIM can easily test and analyze various scheduling policies in a number of …
With the fast development of deep neural networks (DNNs), many real-world applications are adopting multiple models to conduct compound tasks, such as co-running classification, detection, and segmentation models on autonomous vehicles. Such multi-tenant DNN inference cases greatly exacerbate the computational complexity and call for comprehensive collaboration for graph-level operator scheduling, runtime-level resource awareness, as well as hardware scheduler support. However, the current scheduling support for such multi-tenant inference is still relatively backward. In this work, we propose…
Microsoft Research（微软研究院）；University of Texas at Austin（德克萨斯大学奥斯汀分校）；VMware Research（威睿研究院）
Training Deep Neural Networks (DNNs) is a widely popular workload in both enterprises and cloud data centers. Existing schedulers for DNN training consider GPU as the dominant resource, and allocate other resources such as CPU and memory proportional to the number of GPUs requested by the job. Unfortunately, these schedulers do not consider the impact of a job's sensitivity to allocation of CPU, memory, and storage resources. In this work, we propose Synergy, a resource-sensitive scheduler for shared GPU clusters. Synergy infers the sensitivity of DNNs to different resources using optimistic p…
Multi-accelerator servers are increasingly being deployed in shared multi-tenant environments (such as in cloud data centers) in order to meet the demands of large-scale compute-intensive workloads. In addition, these accelerators are increasingly being inter-connected in complex topologies and workloads are exhibiting a wider variety of inter-accelerator communication patterns. However, existing allocation policies are ill-suited for these emerging use-cases. Specifically, this work identifies that multi-accelerator workloads are commonly fragmented leading to reduced bandwidth and increased …
BandPilot: Toward Performance- and Contention-Aware GPU Dispatching in AI Clusters
The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））；National University of Defense Technology（国防科技大学）；Southern University of Science and Technology（南方科技大学）；Sun Yat-sen University（中山大学）
Powered by advances in deep learning (DL) techniques, machine learning and artificial intelligence have achieved astonishing successes. However, the rapidly growing needs for DL also led to communication- and resource-intensive distributed training jobs for large-scale DL training, which are typically deployed over GPU clusters. To sustain the ever-increasing demand for DL training, the so-called "ring-all-reduce" (RAR) technologies have recently emerged as a favorable computing architecture to efficiently process network communication and computation load in GPU clusters. The most salient fea…
Network management on multi-tenant container-based data centers has critical impact on performance. Tenants encapsulate applications in containers abstracting away details on hosting infrastructures, and entrust data centers management framework with the provisioning of network QoS requirements. In this paper, we propose a network-aware multi-criteria container scheduler to jointly process containers and network requirements. We introduce a new Mixed Integer Linear Programming formulation for network-aware scheduling encompassing both tenants and providers metrics. We describe two GPU-accelera…
Deep learning models are increasingly used for end-user applications, supporting both novel features such as facial recognition, and traditional features, e.g. web search. To accommodate high inference throughput, it is common to host a single pre-trained Convolutional Neural Network (CNN) in dedicated cloud-based servers with hardware accelerators such as Graphics Processing Units (GPUs). However, GPUs can be orders of magnitude more expensive than traditional Central Processing Unit (CPU) servers. These resources could also be under-utilized facing dynamic workloads, which may result in infl…
While providing low latency is a fundamental requirement in deploying recommendation services, achieving high resource utility is also crucial in cost-effectively maintaining the datacenter. Co-locating multiple workers of a model is an effective way to maximize query-level parallelism and server throughput, but the interference caused by concurrent workers at shared resources can prevent server queries from meeting its SLA. Hera utilizes the heterogeneous memory requirement of multi-tenant recommendation models to intelligently determine a productive set of co-located models and its resource …
Causal Inference for Quantifying Noisy Neighbor Effects in Multi-Tenant Cloud Environments
Federal University of Espírito Santo（圣灵联邦大学）；Federal University of Ceará（塞阿拉联邦大学）；Universidade Federal de Minas Gerais（米纳斯吉拉斯联邦大学）；University of São Paulo（圣保罗大学）；University of Minho（米尼奥大学）
With the rise of machine learning, inference on deep neural networks (DNNs) has become a core building block on the critical path for many cloud applications. Applications today rely on isolated ad-hoc deployments that force users to compromise on consistent latency, elasticity, or cost-efficiency, depending on workload characteristics. We propose to elevate DNN inference to be a first class cloud primitive provided by a shared multi-tenant system, akin to cloud storage, and cloud databases. A shared system enables cost-efficient operation with consistent performance across the full spectrum o…
GPUs are essential to accelerating the latency-sensitive deep neural network (DNN) inference workloads in cloud datacenters. To fully utilize GPU resources, spatial sharing of GPUs among co-located DNN inference workloads becomes increasingly compelling. However, GPU sharing inevitably brings severe performance interference among co-located inference workloads, as motivated by an empirical measurement study of DNN inference on EC2 GPU instances. While existing works on guaranteeing inference performance service level objectives (SLOs) focus on either temporal sharing of GPUs or reactive GPU re…
In cloud machine learning (ML) inference systems, providing low latency to end-users is of utmost importance. However, maximizing server utilization and system throughput is also crucial for ML service providers as it helps lower the total-cost-of-ownership. GPUs have oftentimes been criticized for ML inference usages as its massive compute and memory throughput is hard to be fully utilized under low-batch inference scenarios. To address such limitation, NVIDIA's recently announced Ampere GPU architecture provides features to "reconfigure" one large, monolithic GPU into multiple smaller "GPU p…
Function-as-a-Service (FaaS) is emerging as an important cloud computing service model as it can improve the scalability and usability of a wide range of applications, especially Machine-Learning (ML) inference tasks that require scalable resources and complex software configurations. These inference tasks heavily rely on GPUs to achieve high performance; however, support for GPUs is currently lacking in the existing FaaS solutions. The unique event-triggered and short-lived nature of functions poses new challenges to enabling GPUs on FaaS, which must consider the overhead of transferring data…
Serving deep neural networks in latency critical interactive settings often requires GPU acceleration. However, the small batch sizes typical in online inference results in poor GPU utilization, a potential performance gap which GPU resource sharing can address. In this paper, we explore several techniques to leverage both temporal and spatial multiplexing to improve GPU utilization for deep learning inference workloads. We evaluate the performance trade-offs of each approach with respect to resource-efficiency, latency predictability, and isolation when compared with conventional batched infe…
Edge computing has emerged as a popular paradigm for supporting mobile and IoT applications with low latency or high bandwidth needs. The attractiveness of edge computing has been further enhanced due to the recent availability of special-purpose hardware to accelerate specific compute tasks, such as deep learning inference, on edge nodes. In this paper, we experimentally compare the benefits and limitations of using specialized edge systems, built using edge accelerators, to more traditional forms of edge and cloud computing. Our experimental study using edge-based AI workloads shows that tod…
