DeepSim: cluster level behavioural simulation model for deep learning
by Yuankun Shi; Kevin J. Long; Kaushik Balasubramanian; Zhaojuan Bian; Adam Procter; Ramesh Illikkal
International Journal of Big Data Intelligence (IJBDI), Vol. 6, No. 3/4, 2019

Abstract: We are witnessing an explosion of AI use cases driving the computer industry, and especially datacentre and server architectures. As Intel faces fierce competition in this emerging technology space, it is critical that architecture definitions and directions are driven with data from proper tools and methodologies, and insights are drawn from end-to-end holistic analysis at the datacentre levels. In this paper, we introduce DeepSim, a cluster-level behavioural simulation model for deep learning. DeepSim, which is based on the Intel CoFluent simulation framework, uses timed behavioural models to simulate complex interworking between compute nodes, networking, and storage at the datacentre level, providing a realistic performance model of real-world image recognition applications based on the popular deep learning framework Caffe. The end-to-end simulation data from DeepSim provides insights which can be used for architecture analysis driving future datacentre architecture directions. DeepSim enables scalable system design, deployment, and capacity planning through accurate performance insights. Results from preliminary scaling studies (e.g., node scaling and network scaling) and what-if analyses (e.g., Xeon with HBM and Xeon Phi with dual OPA) are presented in this paper. The simulation results are correlated well with empirical measurements, achieving an accuracy of 95%.

Online publication date: Fri, 19-Jul-2019

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Big Data Intelligence (IJBDI):
Login with your Inderscience username and password:

Username:        Password:          Forgotten your password?

Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com