skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: spotDNN: Provisioning Spot Instances for Predictable Distributed DNN Training in the Cloud
Distributed Deep Neural Network (DDNN) training on cloud spot instances is increasingly compelling as it can significantly save the user budget. To handle unexpected instance revocations, provisioning a heterogeneous cluster using the asynchronous parallel mechanism becomes the dominant method for DDNN training with spot instances. However, blindly provisioning a cluster of spot instances can easily result in unpredictable DDNN training performance, mainly because bottlenecks occur on the parameter server network bandwidth and PCIe bandwidth resources, as well as the inadequate cluster heterogeneity. To address the challenges above, we propose spotDNN, a heterogeneity-aware spot instance provisioning framework that provides predictable performance for DDNN training in the cloud. By explicitly considering the contention for bottle-neck resources, we first build an analytical performance model of DDNN training in heterogeneous clusters. It leverages the weighted average batch size and convergence coefficient to quantify the DDNN training loss in heterogeneous clusters. Through a lightweight workload profiling, we further design a cost-efficient instance provisioning strategy which incorporates the bounds calculation and sliding window techniques to effectively guarantee the training performance service level objectives (SLOs). We have implemented a prototype of spotDNN and conducted extensive experiments on Amazon EC2. Experiment results show that spotDNN can deliver predictable DDNN training performance while reducing the monetary cost by up to 68:1% compared to the existing solutions, yet with acceptable runtime overhead.  more » « less
Award ID(s):
2019511
PAR ID:
10431833
Author(s) / Creator(s):
Date Published:
Journal Name:
Proceedings of 31st IEEE/ACM International Symposium on Quality of Service (IWQoS 2023)
Volume:
1
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; (, VLDB DISPA Workshop 2020)
    null (Ed.)
    Cloud providers offer instances with similar compute capabilities (for example, instances with different generations of GPUs like K80s, P100s, V100s) across many regions, availability zones, and on-demand and spot markets, with prices governed independently by individual supplies and demands. In this paper, using machine learning model training as an example application, we explore the potential cost reductions possible by leveraging this cross-cloud instance market. We present quantitative results on how the prices of cloud instances change with time, and how total costs can be decreased by considering this dynamic pricing market. Our preliminary experiments show that a) the optimal instance choice for a model is dependent on both the objective (e.g., cost, time, or combination) and the model’s performance characteristics, b) the cost of moving training jobs between instances is cheap, c) jobs do not need to be preempted more frequently than once a day to leverage the benefits from spot instance price variations, and d) the cost of training a model can be decreased by as much as 3.5× compared to a static policy. We also look at contexts where users specify higherlevel objectives over collections of jobs, show examples of policies for these contexts, and discuss additional challenges involved in making these cost reductions viable. 
    more » « less
  2. ; ; ; (, 2023 ACM/SPEC International Conference on Performance Engineering)
    Many organizations maintain and operate large shared computing clusters, since they can substantially reduce computing costs by leveraging statistical multiplexing to amortize it across all users. Importantly, such shared clusters are generally not free to use, but have an internal pricing model that funds their operation. Since employees at many large organizations, especially Universities, have some budgetary autonomy over purchase decisions, internal shared clusters are increasingly competing for users with cloud platforms, which may offer lower costs and better performance. As a result, many organizations are shifting their shared clusters to operate on cloud resources. This paper empirically analyzes the user incentives for shared cloud clusters under two different pricing models using an 8-year job trace from a large shared cluster for a large University system. Our analysis shows that, with either pricing model, a large fraction of users have little financial incentive to participate in a shared cloud cluster compared to directly acquiring resources from a cloud platform. While shared cloud clusters can provide some limited reductions in cost by leveraging reserved instances at a discount, due to bursty workloads, realizing these reductions generally requires imposing long job waiting times, which for many users are likely not worth the cost reduction. In particular, we show that, assuming users defect from the shared cluster if their wait time is greater than 15x their average job runtime, over 80% of the users would defect, which increases the price of the remaining users such that it eliminates any incentive to participate in a shared cluster. Thus, while shared cloud clusters may provide users other benefits, their financial incentives are weak. 
    more » « less
  3. ; ; ; ; ; ; ; ; (, IEEE)
    Deep neural networks (DNNs) are increasingly popular owing to their ability to solve complex problems such as image recognition, autonomous driving, and natural language processing. Their growing complexity coupled with the use of larger volumes of training data (to achieve acceptable accuracy) has warranted the use of GPUs and other accelerators. Such accelerators are typically expensive, with users having to pay a high upfront cost to acquire them. For infrequent use, users can, instead, leverage the public cloud to mitigate the high acquisition cost. However, with the wide diversity of hardware instances (particularly GPU instances) available in public cloud, it becomes challenging for a user to make an appropriate choice from a cost/performance standpoint. In this work, we try to address this problem by (i) introducing a comprehensive distributed deep learning (DDL) profiler Stash, which determines the various execution stalls that DDL suffers from, and (ii) using Stash to extensively characterize various public cloud GPU instances by running popular DNN models on them. Specifically, it estimates two types of communication stalls, namely, interconnect and network stalls, that play a dominant role in DDL execution time. Stash is implemented on top of prior work, DS-analyzer, that computes only the CPU and disk stalls. Using our detailed stall characterization, we list the advantages and shortcomings of public cloud GPU instances for users to help them make an informed decision(s). Our characterization results indicate that the more expensive GPU instances may not be the most performant for all DNN models and that AWS can sometimes sub-optimally allocate hardware interconnect resources. Specifically, the intra-machine interconnect can introduce communication overheads of up to 90% of DNN training time and the network-connected instances can suffer from up to 5× slowdown compared to training on a single instance. Furthermore, (iii) we also model the impact of DNN macroscopic features such as the number of layers and the number of gradients on communication stalls, and finally, (iv) we briefly discuss a cost comparison with existing work. 
    more » « less
  4. ; ; ; ; ; ; ; ; (, Proceedings of the International Conference on Distributed Computing Systems)
    Deep neural networks (DNNs) are increasingly popular owing to their ability to solve complex problems such as image recognition, autonomous driving, and natural language processing. Their growing complexity coupled with the use of larger volumes of training data (to achieve acceptable accuracy) has warranted the use of GPUs and other accelerators. Such accelerators are typically expensive, with users having to pay a high upfront cost to acquire them. For infrequent use, users can, instead, leverage the public cloud to mitigate the high acquisition cost. However, with the wide diversity of hardware instances (particularly GPU instances) available in public cloud, it becomes challenging for a user to make an appropriate choice from a cost/performance standpoint. In this work, we try to address this problem by (i) introducing a comprehensive distributed deep learning (DDL) profiler Stash, which determines the various execution stalls that DDL suffers from, and (ii) using Stash to extensively characterize various public cloud GPU instances by running popular DNN models on them. Specifically, it estimates two types of communication stalls, namely, interconnect and network stalls, that play a dominant role in DDL execution time. Stash is implemented on top of prior work, DS-analyzer, that computes only the CPU and disk stalls. Using our detailed stall characterization, we list the advantages and shortcomings of public cloud GPU instances for users to help them make an informed decision(s). Our characterization results indicate that the more expensive GPU instances may not be the most performant for all DNN models and that AWS can sometimes sub-optimally allocate hardware interconnect resources. Specifically, the intra-machine interconnect can introduce communication overheads of up to 90% of DNN training time and the network-connected instances can suffer from up to 5× slowdown compared to training on a single instance. Furthermore, (iii) we also model the impact of DNN macroscopic features such as the number of layers and the number of gradients on communication stalls, and finally, (iv) we briefly discuss a cost comparison with existing work. 
    more » « less
  5. ; ; ; ; (, IEEE)
    With the wide adoption of deep neural network (DNN) models for various applications, enterprises, and cloud providers have built deep learning clusters and increasingly deployed specialized accelerators, such as GPUs and TPUs, for DNN training jobs. To arbitrate cluster resources among multi-user jobs, existing schedulers fall short, either lacking fine-grained heterogeneity awareness or hardly generalizable to various scheduling policies. To fill this gap, we propose a novel design of a task-level heterogeneity-aware scheduler, Hadar, based on an online optimization framework that can express other scheduling algorithms. Hadar leverages the performance traits of DNN jobs on a heterogeneous cluster, characterizes the task-level performance heterogeneity in the optimization problem, and makes scheduling decisions across both spatial and temporal dimensions. The primal-dual framework is employed, with our design of a dual subroutine, to solve the optimization problem and guide the scheduling design. Extensive trace-driven simulations with representative DNN models have been conducted to demonstrate that Hadar improves the average job completion time (JCT) by 3× over an Apache YARN-based resource manager used in production. Moreover, Hadar outperforms Gavel[1], the state-of-the-art heterogeneity-aware scheduler, by 2.5× for the average JCT, and shortens the queuing delay by 13% and improve FTF (Finish-Time-Fairness) by 1.5%. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email