Behind the scenes at Labs
Showing results for 
Search instead for 
Did you mean: 

The Deep Learning Cookbook

Natalia Vassilieva and Sergey SerebryakovNatalia Vassilieva and Sergey SerebryakovBy Natalia Vassilieva and Sergey Serebryakov

Deep learning is a key enabling technology behind the recent revival of artificial intelligence. Already adopted by a handful of tech giants and industry leaders, deep learning has the potential to change the way we work, across industries.

We see a lot of interest in this technology from our customers, but many of them don’t have the expertise and resources of tech giants to make informed decisions on optimal hardware and software configurations to run deep learning workloads efficiently. To remedy that, we have created the Deep Learning Cookbook.

It is a common wisdom today, that to start a deep learning exploration one needs a GPU-based system and one of the existing open sources deep learning libraries. But which GPU box to choose? How many of them to put in a cluster? Which library to pick? There are so many of them out there! If a particular GPU system is enough for one deep learning workload, will it be enough for a different one?

Answers to these questions are not obvious. That’s why we decided to create the first “book of recipes” for deep learning workloads.

The Deep Learning Cookbook is based on a massive collection of performance results for different deep learning workloads on different hardware/software (HW/SW) stacks, and analytical performance models. This combination enables us to estimate the performance of a given workload and to recommend an optimal HW/SW stack for that workload.

Currently, our Cookbook is based on the extensive benchmarking of eleven workloads with eight deep learning frameworks and six hardware systems. These include the most popular open source frameworks such as TensorFlow, Caffe, and Caffe2, and hardware systems from different HPE product lines, including HPE Apollo 6500, HPE Apollo 6000, and HPE Edgeline. And this list keeps on growing.

The results of the benchmarks are incorporated into our models to predict performance for all possible system configurations, even those which we have not yet had the chance to evaluate.

We believe the Deep Learning Cookbook will make it easier to employ deep learning in real world vertical applications. Additionally, we use the Cookbook to detect bottlenecks in deep learning workloads and to guide the design of future HPE systems for artificial intelligence and deep learning.

About the Author


Managing Editor, Hewlett Packard Labs


Thanks for this useful information. Where can I get a copy of the Deep Learning Cookbook?




Yes, we will make HPE Deep Learning Cookbook available. Stay tuned!

Our cookbook is a toolset, which consists of several key assets:

  • benchmarking suite: automated benchmarking tool. It will be open sourced and available on GitHub by Discover Madrid time (the very end of November).
  • performance reporting tool: a web-based tool which provides access to a knowledgebase of benchmarking results. It will enable querying and analysis of measured results and performance prediction based on analytical performance models. It is planned to be released in the beginning of 2018.
  • reference architectures: hardware/software recipes for selected workloads. These will be available on HPE web site.

Thank you!

Doug Legge


What research is underway to utilise AI (Machine learning) within our network platforms, for instance, to provide autonomous/semi-autonomous QoS management?






Karthik VR

Hi.. Just checking if a copy of the cookbook is available. Please update. Thanks in advance. 


We do have a few research threads using AI/ML at our labs to improve network management platforms. Here are a few of these that have been published externally:

  • Earlier this summer we presented a paper describing our work on using Machine Learning to flexibly allocate resources in NFV ecosystems at USENIX HotCloud'17.  "ENVI: Elastic resource flexing for Network function Virtualization". The paper is available from
  • Another related work was presented at ACM CoNext Conference  2016 about QoE (Quality-of-Experience) technology based on ML (Machine Learning) and DL (Deep Learning).  We have demo'ed this at various HPE Discover events. The paper is available from:

We will be happy to discuss any questions you might have about this or related work.

Thanks for your interest,





Puneet Sharma, Ph.D.
IEEE Fellow, ACM Distinguished Scientist
Head, Networked Systems Group & Distinguished Technologist

Networking, IoT and Mobility Lab, Hewlett Packard Labs
Palo Alto, CA

Duarte Guerra

Can you please inform me as to when the Cookbook will be available?

Thank you

For those inquiring if the Deep Learning Cookbook is available - not yet, but our teams are working on it now. We'll have an update by the end of the year, stay tuned!


Martina Trucco, HPE

Hi Natalia and Sergey, 

are the Github Deep learning resources avalaible now at Github to download ? Where can I find it ?

Thanks in advance



Great blog posting. I really like the Cookbook on Github