We are featuring one of the seven Summit tracks—AI / Machine Learning, HPC. Get your Summit tickets for free and add these sessions to your Summit calendar!

image

The Open Infrastructure Summit, held virtually for the first time, takes place October 19-23 and includes more than 100 sessions around infrastructure use cases like cloud computing, edge computing, hardware enablement, and security. Thousands of attendees are expected to participate, representing 30+ open source communities and more than 110 countries.

Today, we are featuring one of the seven Summit tracks—AI / Machine Learning, HPC. Get your Summit tickets for free and don’t forget to add these sessions to your Summit calendar!

Docker for Machine Learning!

Enhancement and Optimization of New Heterogeneous Accelerators Based on Cyborg

  • Presented by Brin Zhang & Wenping Song from Inspur, Shaohe Feng from Intel.
  • What can you expect to learn?
    • Cyborg architecture
    • Nova, Cyborg interaction mechanism
    • Accelerator performance optimization
    • Cyborg’s benefits to the cloud platform
  • Add this session to your Summit calendar!

Federated Digital Research Infrastructures for Health Data Research

  • Presented by John Taylor from StackHPC Ltd.
  • What can you expect to learn?
    • Requirements on data governance in handling sensitive data.
    • What platforms are used for analysis.
    • What computation techniques are being used in exploring data such as HPC, AI, etc.
    • Requirements for federation in order to bring the “Researcher to the Data”.
    • How current OpenStack infrastructures are being used to address these aspects.
  • Add this session to your Summit calendar!

How to split a cake? GPUs with OpenStack

  • Presented by Sylvain Bauza from Red Hat.
  • What can you expect to learn?
    • Operators will know how to use Nova for providing instances with virtual GPUs. They will also know about the current supported features and the next ones.
  • Add this session to your Summit calendar!

Lessons Learnt Building Cambridge University’s CSD3 Supercomputer with OpenStack

  • Presented by John Garbutt from StackHPC, Paul Browne from the University of Cambridge.
  • What can you expect to learn?
    • Find out about how Cambridge University manages around 1000 nodes using OpenStack to deliver a mix of Baremetal, VM and Container based workloads. The presenters will include details of how the full lifecycle of the hardware is automated using OpenStack, instead of xCAT. There is a mix of workloads across AI, HPDA, HPC and HTC all sharing a single pool of diverse hardware. Using OpenStack helps break the hardcoding between infrastructure resources and the science platform used to access those resources, and also eases the transition towards more portal science platforms through the adoption of infrastructure as code tooling.
    • Routine re-imaging of Slurm is integrated with OpenStack. A new networking-generic-switch driver is used to dynamically provision the Cumulus switches.
  • Add this session to your Summit calendar!

Machine Learning at Edge Cloud

  • Presented by Prakash Ramchandran from Dell, Vivek Hariharan from Indeed.
  • What can you expect to learn?
    • Attendees will learn how to set up an optimal environment, both in AWS and with Open-Infra projects for running machine learning models and transmitting inferences with low latency. Application developers will be able to better understand application requirements when integrating machine learning models.
    • Ultimately, understanding the technical requirements to allow for machine learning to be effectively integrated with Edge Cloud infrastructure will help attendees innovate, whether it be with AWS or with Open-Infra projects.
  • Add this session to your Summit calendar!

Standalone yet at scale: Ironic take on High Performance Computing

  • Presented by Timothy Randles from Los Alamos National Laboratory, Julia Kreger & Jacob Anders from Red Hat.
  • What can you expect to learn?
    • In this presentation, the presenters will focus on the unique combination of ease of deployment and powerful capabilities that Ironic standalone brings to High Performance Computing (HPC).
  • Add this session to your Summit calendar!

The Coral Reef Cloud: Autoscaling, Reservation and Preemption

  • Presented by John Garbutt & Steve Brasier & Pierre Riteau from StackHPC.
  • What can you expect to learn?
    • In a fixed capacity cloud, it can be hard to manage a fair share of resources between lots of groups of users. The presenters will explore how Blazar can help with these problems.
    • When you first use quotas, you dedicate resources to specific groups. Soon it becomes clear more people want some quota, and the existing users have not yet used all they asked for. The obvious next step is to hand out quota for more resources than you have in your cloud. Eventually the cloud becomes full, instance builds go into an ERROR state with “no valid host”. Then you find people start leaving VMs running, and using rebuild instead of delete, in case they lose access to their “hard won” slice of resources.
    • The presenters will look at how to manage this resource allocation better through the use of Blazar reservations, and how to avoid under-utilization by using preemptible instances.
  • Add this session to your Summit calendar!

Don’t miss these Summit sessions and get your Summit ticket for free!

Still on the fence?

View the full Summit schedule and check out some most anticipated Summit sessions that you might love!

Participate:

Follow the #OpenInfraSummit hashtag on Twitter, Facebook, LinkedIn and make sure to subscribe to the OpenStack Foundation (OSF) YouTube channel to get exclusive behind-the-scenes content on how the Summit is being organized!

Participate in the conversation on

Twittertwitter.com/OpenStack

Facebookfacebook.com/OpenStack

YouTube: youtube.com/user/OpenStackFoundation

WeChat ID: OpenStack

Superuser